v5.3.0

This takes the place of #line directives

.github/workflows/run-tests.yml

@@ -0,0 +1,41 @@
+name: Run Propane Tests
+
+on:
+  push:
+    branches:
+    - master
+  pull_request:
+
+jobs:
+  test:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-latest, macos-latest]
+        ruby-version: ['3.4']
+
+    steps:
+    - name: Install dependencies (Linux)
+      if: runner.os == 'Linux'
+      run: sudo apt-get update && sudo apt-get install -y gcc gdc ldc valgrind
+
+    - name: Install dependencies (macOS)
+      if: runner.os == 'macOS'
+      run: brew install gcc ldc
+
+    - name: Checkout repository
+      uses: actions/checkout@v4
+
+    - name: Set up Ruby
+      uses: ruby/setup-ruby@v1
+      with:
+        ruby-version: ${{ matrix.ruby-version }}
+
+    - name: Set up Rust
+      uses: dtolnay/rust-toolchain@stable
+
+    - name: Install dependencies
+      run: bundle install
+
+    - name: Run tests
+      run: rake all

.gitignore

@@ -1,9 +1,11 @@
-imbecile
-tags
-*.o
-.*.swp
-*.dep
-tmpl.*
-tests/*/itest.cc
-tests/*/itest.h
-tests/*/test
+/.bundle/
+/.yardoc
+/_yardoc/
+/coverage/
+/dist/
+/gen/
+/pkg/
+/spec/reports/
+/tmp/
+/.rspec_status
+/spec/run/

.gitmodules

@@ -1,3 +0,0 @@
-[submodule "refptr"]
-	path = refptr
-	url = http://github.com/holtrop/refptr.git

.rspec

@@ -0,0 +1,2 @@
+--color
+--require spec_helper

CHANGELOG.md

@@ -0,0 +1,306 @@
+## v5.3.0
+
+### New Features
+
+- Warn about unused rules (rules not reachable from any start rule). The
+  warnings are written to the log file and, with `-w`, are printed to stderr
+  and treated as fatal errors.
+
+## v5.2.0
+
+### New Features
+
+- Tree node handles can now be used directly as a boolean condition (e.g.
+  `if (node)`) in C++ (an `explicit operator bool()`) and D (an
+  `opCast!bool`), equivalent to testing `valid()` / `valid`. C and Rust do not
+  support a user-defined conversion to a boolean condition; use
+  `p_node_valid(node)` and `node.valid()` there.
+
+## v5.1.0
+
+### New Features
+
+- Add a `node_id()` accessor to the C++ and D tree node handle types, for node
+  identity comparison. This matches the existing `p_node_id()` macro (C) and
+  `node_id()` method (Rust).
+
+## v5.0.0
+
+### New Features
+
+- Add Rust target language output.
+- Add Rust language detection in propane.vim.
+
+### API Changes
+
+- The matched text argument passed to lexer user code blocks is now named
+  `match_text` instead of `match`, since `match` is a keyword in Rust. The
+  `match_length` argument (C and C++) is unchanged.
+- Tree generation mode now stores all tree nodes in a compact arena owned by
+  the parser context (a flat node array plus a shared child-link array).
+  This replaces the previous design of one heap allocation per node with
+  layout-punned typed structs.
+- Tree nodes are now referenced by lightweight handles rather than pointers.
+  `p_result()` and the field accessors now return handle values in tree
+  generation mode.
+- The whole tree is freed together with the context by `p_context_delete()`.
+  The `p_tree_delete()` / `p_tree_delete_XXX()` functions have been removed;
+  tree node handles are only valid while the context is alive.
+- Tree node field access changed per target language:
+  - C: per-field accessor functions (e.g. `p_Start_pItems(node)`) plus tree
+    walk macros (e.g. `p_tree_walk_Start(node, pItems, pItem, pToken1, token)`),
+    and generic accessors `p_node_valid()`, `p_node_position()`,
+    `p_node_end_position()`, `p_node_n_fields()`, `p_node_data()`, `p_node_id()`.
+  - C++: handle methods called with `()` (e.g. `node.pItems().pToken1().token()`),
+    plus the same C-style functions/macros for convenience.
+  - D: `@property` accessors preserving the previous field-access syntax
+    (e.g. `node.pItems.pToken1.token`); null checks use `.valid` instead of
+    `is null`.
+- Tree-mode parser rule user code: `$$` and `$1` etc. now yield node handles.
+  Reference child fields through the target-language accessors described above
+  rather than through struct pointer members.
+
+### Improvements
+
+- Improve D language detection in propane.vim
+- Speed up specs
+
+## v4.8.1
+
+### Fixes
+
+- Fix tree node struct type forward-declarations for C/C++
+
+## v4.8.0
+
+### New Features
+
+- Add `p_parse_inner_XXX()` APIs that accept a caller-provided set of follow
+  tokens. These behave the same as `p_parse_XXX()` by parsing starting at the
+  given start rule, but instead of expecting the rest of the input to match
+  the start rule they allow specifying a set of tokens that may follow the
+  start rule.
+- Add `p_set_position()` API to set the current text position stored in the
+  context. Useful for setting the initial text position to something other
+  than `(1, 1)` for a nested parse operation.
+- Add `p_input_index()` API to get the current input text byte offset.
+- Add `p_set_input_index()` API to set the current input text byte offset.
+  Useful together with `p_set_position()` to rewind the input part-way through
+  a parse in order to re-read an earlier section of the input.
+
+## v4.7.0
+
+### New Features
+
+- Support parser rule user code blocks in tree generation mode.
+
+### Fixes
+
+- propane.vim: do not highlight rule components as propane keywords
+
+## v4.6.0
+
+### New Features
+
+- Add lexer user code API to access matched input text positions
+- Track rule component text positions and add parser user code API to access
+
+### Fixes
+
+- Fixed a few user guide and source comments related to text input positions
+
+## v4.5.0
+
+### New Features
+
+- Add `noline` grammar statement to skip emitting `#line` directives
+- Attempt to autodetect target language (D/C++) in extra/vim/syntax/propane.vim
+
+### Fixes
+
+- Fix #line reset directives
+- Update keyword list in extra/vim/syntax/propane.vim
+- Fix propane.vim keyword detection
+
+## v4.4.0
+
+### New Features
+
+- Add p_value_get() / p_value_get_XXX() accessors
+
+## v4.3.0
+
+### New Features
+
+- Use #line for user code blocks to report input grammar position for errors.
+
+## v4.2.0
+
+### New Features
+
+- Add support for a custom lex function.
+
+## v4.1.0
+
+### New Features
+
+- Add `p_context_delete()` and `p_tree_delete()` for D targets.
+
+## v4.0.0
+
+### New Features
+
+- Add `context_user_fields` statement to allow custom context user fields.
+- Add `token_user_fields` statement to allow custom token user fields.
+- Add `on_token_node` statement to allow custom code when constructing token nodes.
+- Add `free_token_node` statement to allow custom code when freeing token nodes.
+- Add `p_context_delete()`.
+- Allow `drop` patterns to execute lexer user code blocks.
+
+### Breaking Changes
+
+- Replace `p_context_init()` with `p_context_new()` and `p_context_delete()`.
+- Renamed `p_free_tree()` to `p_tree_delete()`.
+- The `free_token_node` statement now takes a user code block instead of a
+  function name parameter.
+
+## v3.0.0
+
+### New Features
+
+- Add support for multiple starting rules (#38)
+- Add `p_free_tree()` functions to reclaim generated tree memory
+- Add `free_token_node` grammar statement to reclaim user-allocated memory stored in a Token tree node `pvalue` field
+- Add valgrind memory leak tests to unit tests
+- Fix build issues for C++ to officially support C++ target output
+
+### Improvements
+
+- Document `p_lex()` and `p_token_info_t` in user guide (#37)
+
+### Breaking Changes
+
+- Rename AST generation mode to tree generation mode (see [UPGRADING.md](UPGRADING.md))
+
+## v2.3.0
+
+### New Features
+
+- Add \D, \S, \w, \W special character classes
+
+### Improvements
+
+- Include line numbers for pattern errors
+- Improve performance in a few places
+- Parallelize parser table generation on Linux hosts
+- Add github workflow to run unit tests
+
+### Fixes
+
+- Fix a couple clang warnings for C backend
+- Fix C backend not fully initializing pvalues when multiple ptypes are used with different sizes.
+- Fix some user guide examples
+
+## v2.2.1
+
+### Fixes
+
+- Fix GC issue for D backend when AST is enabled (#36)
+
+## v2.2.0
+
+### Improvements
+
+- Allow multiple lexer modes to be specified for a lexer pattern (#35)
+- Document p_decode_code_point() API function (#34)
+
+## v2.1.1
+
+### Fixes
+
+- Field aliases for AST node fields could alias incorrect field when multiple rule alternatives present for one rule set (#33)
+
+## v2.1.0
+
+### Improvements
+
+- Report rule name and line number for conflicting AST node field positions errors (#32)
+
+## v2.0.0
+
+### Improvements
+
+- Log conflicting rules on reduce/reduce conflict (#31)
+- Use 1-based row and column values for position values (#30)
+
+### Fixes
+
+- Fix named optional rules (#29)
+
+### Upgrading
+
+- Adjust all uses of p_position_t row and col values to expect 1-based instead
+of 0-based values.
+
+## v1.5.1
+
+### Improvements
+
+- Improve performance (#28)
+
+## v1.5.0
+
+### New Features
+
+- Track start and end text positions for tokens and rules in AST node structures (#27)
+- Add warnings for shift/reduce conflicts to log file (#25)
+- Add -w command line switch to treat warnings as errors and output to stderr (#26)
+- Add rule field aliases (#24)
+
+### Improvements
+
+- Show line numbers of rules on conflict (#23)
+
+## v1.4.0
+
+### New Features
+
+- Allow user to specify AST node name prefix or suffix
+- Allow specifying the start rule name
+- Allow rule terms to be marked as optional
+
+### Improvements
+
+- Give a better error message when a referenced ptype has not been declared
+
+## v1.3.0
+
+### New Features
+
+- Add AST generation (#22)
+
+## v1.2.0
+
+### New Features
+
+- Allow one line user code blocks (#21)
+- Add backslash escape codes (#19)
+- Add API to access unexpected token found (#18)
+- Add token_names API (#17)
+- Add D example to user guide for p_context_init() (#16)
+- Allow user termination from lexer code blocks (#15)
+
+### Fixes
+
+- Fix generator hang when state transition cycle is present (#20)
+
+## v1.1.0
+
+### New Features
+
+- Add user parser terminations (#13)
+- Document generated parser API in user guide (#14)
+
+## v1.0.0
+
+- Initial release

Gemfile

@@ -0,0 +1,9 @@
+source "https://rubygems.org"
+
+gem "base64"
+gem "rake"
+gem "rspec"
+gem "rdoc"
+gem "redcarpet"
+gem "syntax"
+gem "simplecov"

Gemfile.lock

@@ -0,0 +1,66 @@
+GEM
+  remote: https://rubygems.org/
+  specs:
+    base64 (0.3.0)
+    diff-lcs (1.6.2)
+    erb (6.0.7)
+    logger (1.7.0)
+    prism (2.1.0)
+    prism (2.1.0-aarch64-linux-gnu)
+    prism (2.1.0-aarch64-linux-musl)
+    prism (2.1.0-arm-linux-gnu)
+    prism (2.1.0-arm-linux-musl)
+    prism (2.1.0-arm64-darwin)
+    prism (2.1.0-x86_64-darwin)
+    prism (2.1.0-x86_64-linux-gnu)
+    prism (2.1.0-x86_64-linux-musl)
+    rake (13.4.2)
+    rbs (4.2.0)
+      logger
+      prism (>= 1.6.0)
+      tsort
+    rdoc (8.1.0)
+      erb
+      prism (>= 1.6.0)
+      rbs (>= 4.0.0)
+      tsort
+    redcarpet (3.6.1)
+    rspec (3.13.2)
+      rspec-core (~> 3.13.0)
+      rspec-expectations (~> 3.13.0)
+      rspec-mocks (~> 3.13.0)
+    rspec-core (3.13.6)
+      rspec-support (~> 3.13.0)
+    rspec-expectations (3.13.5)
+      diff-lcs (>= 1.2.0, < 2.0)
+      rspec-support (~> 3.13.0)
+    rspec-mocks (3.13.8)
+      diff-lcs (>= 1.2.0, < 2.0)
+      rspec-support (~> 3.13.0)
+    rspec-support (3.13.7)
+    simplecov (1.3.2)
+    syntax (1.2.2)
+    tsort (0.2.0)
+
+PLATFORMS
+  aarch64-linux-gnu
+  aarch64-linux-musl
+  arm-linux-gnu
+  arm-linux-musl
+  arm64-darwin
+  ruby
+  x86_64-darwin
+  x86_64-linux-gnu
+  x86_64-linux-musl
+
+DEPENDENCIES
+  base64
+  rake
+  rdoc
+  redcarpet
+  rspec
+  simplecov
+  syntax
+
+BUNDLED WITH
+  4.0.14

LICENSE.txt

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2010-2026 Josh Holtrop
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

Makefile

@@ -1,61 +0,0 @@
-
-TARGET   := imbecile
-CXXOBJS  := $(patsubst %.cc,%.o,$(wildcard *.cc)) tmpl.o
-CXXDEPS  := $(patsubst %.o,.%.dep,$(CXXOBJS))
-CXXFLAGS := -O2
-DEPS     := $(CXXDEPS)
-OBJS     := $(CXXOBJS)
-LDFLAGS  := -lpcre
-CPPFLAGS := -I$(shell pwd)/refptr
-
-all: submodule_check tmpl.h $(TARGET)
-
-.PHONY: submodule_check
-submodule_check:
-	@if [ ! -e refptr/refptr.h ]; then \
-		echo Error: \"refptr\" folder is not populated.; \
-		echo Perhaps you forgot to do \"git checkout --recursive\"?; \
-		echo You can remedy the situation with \"git submodule update --init\".; \
-		exit 1; \
-	fi
-
-$(TARGET): $(OBJS)
-	$(CXX) -o $@ $^ $(LDFLAGS)
-
-# Object file rules
-%.o: %.cc
-	$(CXX) -c -o $@ $(CPPFLAGS) $(CXXFLAGS) $<
-
-# Make dependency files
-.%.dep: %.c
-	@set -e; rm -f $@; \
-	  $(CC) -MM $(CPPFLAGS) $< | sed 's,\($*\)\.o[ :]*,\1.o $@ : ,g' > $@
-
-.%.dep: %.cc tmpl.h
-	@set -e; rm -f $@; \
-	  $(CXX) -MM $(CPPFLAGS) $< | sed 's,\($*\)\.o[ :]*,\1.o $@ : ,g' > $@
-
-tmpl.cc: $(wildcard tmpl/*)
-	echo -n > $@
-	for f in $*/*; \
-		do xxd -i $$f >> $@; \
-	done
-
-tmpl.h: tmpl.cc
-	echo '#ifndef $*_h' > $@
-	echo '#define $*_h' >> $@
-	grep '$*_' $^ | sed -e 's/^/extern /' -e 's/ =.*/;/' >> $@
-	echo '#endif' >> $@
-
-.PHONY: tests
-tests: PATH := $(shell pwd):$(PATH)
-tests: all
-	$(MAKE) -C $@
-
-tests-clean:
-	$(MAKE) -C tests clean
-
-clean: tests-clean
-	-rm -f $(TARGET) *.o .*.dep tmpl.cc tmpl.h
-
--include $(CXXDEPS)

Parser.cc

@@ -1,423 +0,0 @@
-
-#include <stdio.h>
-#include <string.h>
-#include <pcre.h>
-#include <ctype.h>                  /* toupper() */
-
-#include <iostream>
-#include <fstream>
-#include <string>
-#include <map>
-
-#include "Parser.h"
-#include "TokenDefinition.h"
-#include "RuleDefinition.h"
-#include "tmpl.h"
-
-using namespace std;
-
-#define DEBUG
-
-Parser::Parser()
-    : m_classname("Parser"), m_namespace(""), m_extension("cc"),
-    m_token_data(new string()), m_token_code(new string()),
-    m_defines(new string())
-{
-}
-
-void Parser::makeDefine(const string & defname, const string & definition)
-{
-    *m_defines += string("#define ") + defname + " " + definition + "\n";
-}
-
-bool Parser::write(const string & fname)
-{
-    if (m_tokens.size() < 1 || m_rules.size() < 1)
-        return false;
-
-    string header_fname = fname + ".h";
-    string body_fname = fname + "." + m_extension;
-
-    ofstream header(header_fname.c_str());
-    ofstream body(body_fname.c_str());
-
-    /* process data */
-    refptr<string> token_classes = new string();
-    refptr<string> token_classes_code = new string();
-    int i = 0;
-    for (list<TokenDefinitionRef>::const_iterator it = m_tokens.begin();
-            it != m_tokens.end();
-            it++)
-    {
-        char buff[20];
-        sprintf(buff, "%d", i++);
-        makeDefine((*it)->getIdentifier(), buff);
-        *token_classes += (*it)->getClassDefinition();
-        *token_classes_code += (*it)->getProcessMethod();
-    }
-    if (m_namespace != "")
-    {
-        makeDefine("I_NAMESPACE", m_namespace);
-    }
-    makeDefine("I_CLASSNAME", m_classname);
-
-    /* set up replacements */
-    setReplacement("token_list", buildTokenList());
-    setReplacement("buildToken", buildBuildToken());
-    setReplacement("header_name",
-            new string(string("\"") + header_fname + "\""));
-    setReplacement("token_code", m_token_code);
-    setReplacement("token_data", m_token_data);
-    setReplacement("defines", m_defines);
-    setReplacement("token_classes", token_classes);
-    setReplacement("token_classes_code", token_classes_code);
-
-    /* write the header */
-    writeTmpl(header, (char *) tmpl_parser_h, tmpl_parser_h_len);
-
-    /* write the body */
-    writeTmpl(body, (char *) tmpl_parser_cc, tmpl_parser_cc_len);
-
-    header.close();
-    body.close();
-
-    return true;
-}
-
-bool Parser::writeTmpl(std::ostream & out, char * dat, int len)
-{
-    char * newline;
-    char * data = dat;
-    const char * errptr;
-    int erroffset;
-    data[len-1] = '\n';
-    const int ovec_size = 6;
-    int ovector[ovec_size];
-    pcre * replace = pcre_compile("{%(\\w+)%}", 0, &errptr, &erroffset, NULL);
-    while (data < (dat + len) && (newline = strstr(data, "\n")) != NULL)
-    {
-        if (pcre_exec(replace, NULL, data, newline - data,
-                    0, 0, ovector, ovec_size) >= 0)
-        {
-            if (ovector[0] > 0)
-            {
-                out.write(data, ovector[0]);
-            }
-            out << *getReplacement(string(data, ovector[2],
-                        ovector[3] - ovector[2]));
-            if (ovector[1] < newline - data)
-            {
-                out.write(data + ovector[1], newline - data - ovector[1]);
-            }
-        }
-        else
-        {
-            out.write(data, newline - data);
-        }
-        out << '\n';
-        data = newline + 1;
-    }
-}
-
-refptr<std::string> Parser::getReplacement(const std::string & name)
-{
-    if (m_replacements.find(name) != m_replacements.end())
-    {
-        return m_replacements[name];
-    }
-#ifdef DEBUG
-    cerr << "No replacement found for \"" << name << "\"" << endl;
-#endif
-    return new string("");
-}
-
-refptr<string> Parser::buildTokenList()
-{
-    refptr<string> tokenlist = new string();
-    for (list<TokenDefinitionRef>::const_iterator t = m_tokens.begin();
-            t != m_tokens.end();
-            t++)
-    {
-        if (t != m_tokens.begin())
-            *tokenlist += "    ";
-        *tokenlist += "{ \"" + (*t)->getName() + "\", \""
-            + (*t)->getCString() + "\", "
-            + ((*t)->getProcessFlag() ? "true" : "false") + " }";
-        if (({typeof(t) tmp = t; ++tmp;}) != m_tokens.end())
-            *tokenlist += ",\n";
-    }
-    return tokenlist;
-}
-
-refptr<string> Parser::buildBuildToken()
-{
-    refptr<string> buildToken = new string();
-    for (list<TokenDefinitionRef>::const_iterator t = m_tokens.begin();
-            t != m_tokens.end();
-            t++)
-    {
-        *buildToken += "case " + (*t)->getIdentifier() + ":\n";
-        *buildToken += "    token = new " + (*t)->getClassName() + "();\n";
-        *buildToken += "    break;\n";
-    }
-    return buildToken;
-}
-
-bool Parser::parseInputFile(char * buff, int size)
-{
-    typedef pcre * pcre_ptr;
-    enum { none, tokens, rules };
-    pcre_ptr empty, comment, section_name, token, rule,
-             data_begin, data_end, code_begin, code_end;
-    struct { pcre_ptr * re; const char * pattern; } exprs[] = {
-        {&empty,        "^\\s*$"},
-        {&comment,      "^\\s*#"},
-        {&section_name, "^\\s*\\[([^\\]]+?)\\]\\s*$"},
-        {&token,        "^\\s*"                     /* possible leading ws */
-                        "([a-zA-Z_][a-zA-Z_0-9]*)"  /* 1: token name */
-                        "\\s+"                      /* required whitespace */
-                        "((?:[^\\\\\\s]|\\\\.)+)"}, /* 2: token RE */
-        {&rule,         "^\\s*(\\S+)\\s*:=(.*)$"},
-        {&data_begin,   "^\\s*\\${"},
-        {&data_end,     "\\$}"},
-        {&code_begin,   "^\\s*%{"},
-        {&code_end,     "%}"}
-    };
-    const int ovec_size = 3 * 10;
-    int ovector[ovec_size];
-    int lineno = 0;
-    char * newline;
-    char * input = buff;
-    string current_section_name;
-    map<string, int> sections;
-    sections["none"] = none;
-    sections["tokens"] = tokens;
-    sections["rules"] = rules;
-    int section = none;
-    string line;
-    bool append_line = false;
-    bool gathering_data = false;
-    bool gathering_code = false;
-    string gather;
-    bool continue_line = false;
-    TokenDefinitionRef current_token;
-
-    for (int i = 0; i < sizeof(exprs)/sizeof(exprs[0]); i++)
-    {
-        const char * errptr;
-        int erroffset;
-        *exprs[i].re = pcre_compile(exprs[i].pattern, 0,
-                &errptr, &erroffset, NULL);
-        if (*exprs[i].re == NULL)
-        {
-            cerr << "Error compiling regex '" << exprs[i].pattern <<
-                "': " << errptr << " at position " << erroffset << endl;
-            return false;
-        }
-    }
-
-    for (;;)
-    {
-        if (continue_line)
-        {
-            continue_line = false;
-        }
-        else
-        {
-            if ((newline = strstr(input, "\n")) == NULL)
-                break;
-            int line_length = newline - input;
-            if (line_length >= 1 && newline[-1] == '\r')
-            {
-                newline[-1] = '\n';
-                line_length--;
-            }
-            lineno++;
-
-            if (append_line)
-            {
-                line += string(input, line_length);
-            }
-            else
-            {
-                line = string(input, line_length);
-            }
-            input = newline + 1;        /* set up for next loop iteration */
-        }
-
-        if ( (pcre_exec(empty, NULL, line.c_str(), line.size(),
-                  0, 0, ovector, ovec_size) >= 0)
-          || (pcre_exec(comment, NULL, line.c_str(), line.size(),
-                  0, 0, ovector, ovec_size) >= 0)
-           )
-        {
-            /* skip empty or comment lines */;
-            continue;
-        }
-
-        if (! (gathering_code || gathering_data) )
-        {
-            if (line.size() > 0 && line[line.size()-1] == '\\')
-            {
-                line[line.size()-1] = ' ';
-                append_line = true;
-                continue;
-            }
-            else
-            {
-                append_line = false;
-            }
-
-            if (pcre_exec(section_name, NULL, line.c_str(), line.size(),
-                        0, 0, ovector, ovec_size) >= 0)
-            {
-                current_section_name
-                    = string(line, ovector[2], ovector[3] - ovector[2]);
-                if (sections.find(current_section_name) != sections.end())
-                {
-                    section = sections[current_section_name];
-                }
-                else
-                {
-                    cerr << "Unknown section name '" << current_section_name
-                        << "'!" << endl;
-                    return false;
-                }
-                continue;
-            }
-        }
-
-        switch (section)
-        {
-            case none:
-                cerr << "Unrecognized input on line " << lineno << endl;
-                return false;
-            case tokens:
-                if      (gathering_data)
-                {
-                    if (pcre_exec(data_end, NULL, line.c_str(), line.size(),
-                                0, 0, ovector, ovec_size) >= 0)
-                    {
-                        gather += string(line, 0, ovector[0]) + "\n";
-                        gathering_data = false;
-                        line = string(line, ovector[1]);
-                        continue_line = true;
-                        if (current_token.isNull())
-                        {
-                            *m_token_data += gather;
-                        }
-                        else
-                        {
-                            current_token->addData(gather);
-                        }
-                    }
-                    else
-                    {
-                        gather += line + "\n";
-                    }
-                    continue;
-                }
-                else if (gathering_code)
-                {
-                    if (pcre_exec(code_end, NULL, line.c_str(), line.size(),
-                                0, 0, ovector, ovec_size) >= 0)
-                    {
-                        gather += string(line, 0, ovector[0]) + "\n";
-                        gathering_code = false;
-                        line = string(line, ovector[1]);
-                        continue_line = true;
-                        if (current_token.isNull())
-                        {
-                            *m_token_code += gather;
-                        }
-                        else
-                        {
-                            current_token->addCode(gather);
-                        }
-                    }
-                    else
-                    {
-                        gather += line + "\n";
-                    }
-                    continue;
-                }
-                else if (pcre_exec(data_begin, NULL, line.c_str(), line.size(),
-                            0, 0, ovector, ovec_size) >= 0)
-                {
-                    gathering_data = true;
-                    gather = "";
-                    line = string(line, ovector[1]);
-                    continue_line = true;
-                    continue;
-                }
-                else if (pcre_exec(code_begin, NULL, line.c_str(), line.size(),
-                            0, 0, ovector, ovec_size) >= 0)
-                {
-                    gathering_code = true;
-                    gather = "";
-                    line = string(line, ovector[1]);
-                    continue_line = true;
-                    continue;
-                }
-                else if (pcre_exec(token, NULL, line.c_str(), line.size(),
-                            0, 0, ovector, ovec_size) >= 0)
-                {
-                    string name(line, ovector[2], ovector[3] - ovector[2]);
-                    string definition(line,
-                            ovector[4], ovector[5] - ovector[4]);
-                    current_token = new TokenDefinition();
-                    if (current_token->create(name, definition))
-                    {
-                        addTokenDefinition(current_token);
-                    }
-                    else
-                    {
-                        cerr << "Error in token definition ending on line "
-                            << lineno << endl;
-                        return false;
-                    }
-                    line = string(line, ovector[1]);
-                    continue_line = true;
-                    continue;
-                }
-                else
-                {
-                    cerr << "Unrecognized input on line " << lineno << endl;
-                    return false;
-                }
-                break;
-            case rules:
-                if (pcre_exec(rule, NULL, line.c_str(), line.size(),
-                            0, 0, ovector, ovec_size) >= 0)
-                {
-                    string name(line, ovector[2], ovector[3] - ovector[2]);
-                    string definition(line,
-                            ovector[4], ovector[5] - ovector[4]);
-                    refptr<RuleDefinition> rd = new RuleDefinition();
-                    if (rd->create(name, definition))
-                    {
-                        addRuleDefinition(rd);
-                    }
-                    else
-                    {
-                        cerr << "Error in rule definition ending on line "
-                            << lineno << endl;
-                        return false;
-                    }
-                }
-                else
-                {
-                    cerr << "Unrecognized input on line " << lineno << endl;
-                    return false;
-                }
-                break;
-        }
-    }
-
-    for (int i = 0; i < sizeof(exprs)/sizeof(exprs[0]); i++)
-    {
-        pcre_free(*exprs[i].re);
-    }
-
-    return true;
-}

Parser.h

@@ -1,61 +0,0 @@
-
-#ifndef PARSER_H
-#define PARSER_H
-
-#include <vector>
-#include <string>
-#include <list>
-#include <map>
-
-#include "refptr.h"
-#include "TokenDefinition.h"
-#include "RuleDefinition.h"
-
-class Parser
-{
-    public:
-        Parser();
-        void addTokenDefinition(refptr<TokenDefinition> td)
-        {
-            m_tokens.push_back(td);
-        }
-        void addRuleDefinition(refptr<RuleDefinition> rd)
-        {
-            m_rules.push_back(rd);
-        }
-        bool write(const std::string & fname);
-        bool parseInputFile(char * buff, int size);
-
-        void setClassName(const std::string & cn) { m_classname = cn; }
-        std::string getClassName() { return m_classname; }
-
-        void setNamespace(const std::string & ns) { m_namespace = ns; }
-        std::string getNamespace() { return m_namespace; }
-
-        void setExtension(const std::string & e) { m_extension = e; }
-        std::string getExtension() { return m_extension; }
-
-    protected:
-        refptr<std::string> buildTokenList();
-        refptr<std::string> buildBuildToken();
-        bool writeTmpl(std::ostream & out, char * dat, int len);
-        refptr<std::string> getReplacement(const std::string & name);
-        void setReplacement(const std::string & name, refptr<std::string> val)
-        {
-            m_replacements[name] = val;
-        }
-        void makeDefine(const std::string & defname,
-                const std::string & definition);
-
-        std::list<TokenDefinitionRef> m_tokens;
-        std::vector< refptr< RuleDefinition > > m_rules;
-        std::string m_classname;
-        std::string m_namespace;
-        std::string m_extension;
-        std::map< std::string, refptr<std::string> > m_replacements;
-        refptr<std::string> m_token_data;
-        refptr<std::string> m_token_code;
-        refptr<std::string> m_defines;
-};
-
-#endif

README

@@ -1,5 +0,0 @@
-Imbecile is a bottom-up parser generator. It targets C++ and automatically
-generates a class heirarchy for interacting with the parser.
-
-Imbecile generates both a lexer and a parser based on the rules given to
-it in the input file.

README.md

@@ -0,0 +1,116 @@
+# The Propane Parser Generator
+
+Propane is a LALR Parser Generator (LPG) which:
+
+  * accepts LR(0), SLR, and LALR grammars
+  * generates a built-in lexer to tokenize input
+  * supports UTF-8 lexer inputs
+  * generates a table-driven shift/reduce parser to parse input in linear time
+  * targets C, C++, D, or Rust language outputs
+  * optionally supports automatic full parse tree generation
+  * supports starting parsing from multiple start rules
+  * tracks input text start and end positions for all matched tokens/rules
+  * is MIT-licensed
+  * is distributable as a standalone Ruby script
+
+## Installation
+
+Propane is designed to be distributed as a stand-alone single file script that
+can be copied into and versioned in a project's source tree.
+The only requirement to run Propane is that the system has a Ruby interpreter
+installed.
+The latest release can be downloaded from [https://github.com/holtrop/propane/releases](https://github.com/holtrop/propane/releases).
+
+Simply copy the `propane` executable script into the desired location within
+the project to be built (typically the root of the repository) and mark it
+executable.
+
+## Usage
+
+### Command Line Interface
+
+Propane is typically invoked from the command-line as `./propane`.
+
+    Usage: ./propane [options] <input-file> <output-file>
+    Options:
+      -h, --help  Show this usage and exit.
+      --log LOG   Write log file. This will show all parser states and their
+                  associated shifts and reduces. It can be helpful when
+                  debugging a grammar.
+      --version   Show program version and exit.
+      -w          Treat warnings as errors. This option will treat shift/reduce
+                  conflicts and unused rules as fatal errors and will print
+                  them to stderr in addition to the log file.
+
+The user must specify the path to a Propane input grammar file and a path to an
+output file.
+The generated source code will be written to the output file.
+If a log file path is specified, Propane will write a log file containing
+detailed information about the parser states and transitions.
+
+### Propane Grammar File
+
+A Propane grammar file provides Propane with the patterns, tokens, grammar
+rules, and user code blocks from which to build the generated lexer and parser.
+
+Example grammar file:
+
+```
+<<
+import std.math;
+>>
+
+# Parser values are unsigned integers.
+ptype ulong;
+
+# A few basic arithmetic operators.
+token plus /\+/;
+token times /\*/;
+token power /\*\*/;
+token integer /\d+/ <<
+  ulong v;
+  foreach (c; match_text)
+  {
+    v *= 10;
+    v += (c - '0');
+  }
+  $$ = v;
+>>
+token lparen /\(/;
+token rparen /\)/;
+# Drop whitespace.
+drop /\s+/;
+
+Start -> E1 << $$ = $1; >>
+E1 -> E2 << $$ = $1; >>
+E1 -> E1 plus E2 << $$ = $1 + $3; >>
+E2 -> E3 << $$ = $1; >>
+E2 -> E2 times E3 << $$ = $1 * $3; >>
+E3 -> E4 << $$ = $1; >>
+E3 -> E3 power E4 <<
+  $$ = pow($1, $3);
+>>
+E4 -> integer << $$ = $1; >>
+E4 -> lparen E1 rparen << $$ = $2; >>
+```
+
+Grammar files can contain comment lines beginning with `#` which are ignored.
+White space in the grammar file is also ignored.
+
+It is convention to use the extension `.propane` for the Propane grammar file,
+however any file name is accepted by Propane.
+
+See [https://holtrop.github.io/propane/index.html](https://holtrop.github.io/propane/index.html) for the full User Guide.
+
+## Development
+
+After checking out the repository, run `bundle install` to install dependencies.
+Run `rake spec` to execute tests.
+
+## Contributing
+
+Bug reports and pull requests are welcome on GitHub at https://github.com/holtrop/propane.
+
+## License
+
+Propane is available as open source under the terms of the [MIT License](https://opensource.org/licenses/MIT).

Rakefile.rb

@@ -0,0 +1,59 @@
+require "fileutils"
+require "rake/clean"
+require "rspec/core/rake_task"
+require "simplecov"
+require "stringio"
+
+CLEAN.include %w[spec/run gen .yardoc yard coverage dist]
+
+task :build_dist do
+  sh "ruby rb/build_dist.rb"
+end
+
+RSpec::Core::RakeTask.new(:spec, :example_pattern) do |task, args|
+  if args.example_pattern
+    ENV["partial_specs"] = "1"
+    task.rspec_opts = %W[-e "#{args.example_pattern}" -f documentation]
+  else
+    FileUtils.rm_rf("coverage")
+  end
+end
+task :spec do |task, args|
+  unless ENV["dist_specs"]
+    original_stdout = $stdout
+    sio = StringIO.new
+    $stdout = sio
+    SimpleCov.collate Dir["coverage/parts/*/.resultset.json"]
+    $stdout = original_stdout
+    sio.string.lines.each do |line|
+      $stdout.write(line) unless line =~ /Coverage report generated for/
+    end
+  end
+end
+
+task :valgrind do
+  begin
+    ENV["spec-valgrind"] = "1"
+    Rake::Task[:spec].execute
+  ensure
+    ENV.delete("spec-valgrind")
+  end
+end
+
+# dspec task is useful to test the distributable release script, but is not
+# useful for coverage information.
+desc "Dist Specs"
+task :dspec, [:example_string] => :build_dist do |task, args|
+  ENV["dist_specs"] = "1"
+  Rake::Task["spec"].execute(args)
+  ENV.delete("dist_specs")
+end
+
+task :default => :spec
+
+desc "Build user guide"
+task :user_guide do
+  system("ruby", "-Ilib", "rb/gen_user_guide.rb")
+end
+
+task :all => [:valgrind, :dspec, :user_guide]

RuleDefinition.cc

@@ -1,9 +0,0 @@
-
-#include "RuleDefinition.h"
-
-using namespace std;
-
-bool RuleDefinition::create(const string & name, const string & definition)
-{
-    m_name = name;
-}

RuleDefinition.h

@@ -1,16 +0,0 @@
-
-#ifndef RULEDEFINITION_H
-#define RULEDEFINITION_H
-
-#include <string>
-
-class RuleDefinition
-{
-    public:
-        bool create(const std::string & name, const std::string & definition);
-
-    protected:
-        std::string m_name;
-};
-
-#endif

TokenDefinition.cc

@@ -1,125 +0,0 @@
-
-#include <pcre.h>
-
-#include <iostream>
-#include <string>
-#include <vector>
-
-#include "TokenDefinition.h"
-#include "refptr.h"
-
-using namespace std;
-
-#define WHITESPACE " \n\r\t\v"
-
-static string trim(string s)
-{
-    size_t lastpos = s.find_last_not_of(WHITESPACE);
-    if (lastpos == string::npos)
-        return "";
-    s.erase(lastpos + 1);
-    s.erase(0, s.find_first_not_of(WHITESPACE));
-    return s;
-}
-
-static refptr< vector<string> > split(const string & delim, string str)
-{
-    refptr< vector<string> > ret = new vector<string>();
-    size_t pos;
-    while ( (pos = str.find(delim)) != string::npos )
-    {
-        string t = str.substr(0, pos);
-        ret->push_back(t);
-        str.erase(0, pos + 1);
-    }
-    if (str != "")
-        ret->push_back(str);
-    return ret;
-}
-
-static string c_escape(const string & orig)
-{
-    string result;
-    for (string::const_iterator it = orig.begin(); it != orig.end(); it++)
-    {
-        if (*it == '\\' || *it == '"')
-            result += '\\';
-        result += *it;
-    }
-    return result;
-}
-
-
-TokenDefinition::TokenDefinition()
-    : m_process(false)
-{
-}
-
-bool TokenDefinition::create(const string & name,
-        const string & definition)
-{
-    const char * errptr;
-    int erroffset;
-    pcre * re = pcre_compile(definition.c_str(), 0, &errptr, &erroffset, NULL);
-    if (re == NULL)
-    {
-        cerr << "Error compiling regular expression '" << definition
-            << "' at position " << erroffset << ": " << errptr << endl;
-        return false;
-    }
-    m_name = name;
-    m_definition = definition;
-    pcre_free(re);
-
-#if 0
-    refptr< vector< string > > parts = split(",", flags);
-    for (int i = 0, sz = parts->size(); i < sz; i++)
-    {
-        (*parts)[i] = trim((*parts)[i]);
-        string & s = (*parts)[i];
-        if (s == "p")
-        {
-            m_process = true;
-        }
-        else
-        {
-            cerr << "Unknown token flag \"" << s << "\"" << endl;
-            return false;
-        }
-    }
-#endif
-
-    return true;
-}
-
-string TokenDefinition::getCString() const
-{
-    return c_escape(m_definition);
-}
-
-string TokenDefinition::getClassDefinition() const
-{
-    string ret = "class "+ getClassName() + " : public Token {\n";
-    ret += "public:\n";
-    if (m_process)
-    {
-        ret += "    virtual void process(const Matches & matches);\n";
-    }
-    ret += "\n";
-    ret += "protected:\n";
-    ret += m_data + "\n";
-    ret += "};\n";
-    return ret;
-}
-
-string TokenDefinition::getProcessMethod() const
-{
-    string ret;
-    if (m_code != "")
-    {
-        ret += "void " + getClassName() + "::process(const Matches & matches) {\n";
-        ret += m_code + "\n";
-        ret += "}\n";
-    }
-    return ret;
-}

TokenDefinition.h

@@ -1,37 +0,0 @@
-
-#ifndef TOKENDEFINITION_H
-#define TOKENDEFINITION_H
-
-#include <string>
-#include "refptr.h"
-
-class TokenDefinition
-{
-    public:
-        TokenDefinition();
-        bool create(const std::string & name,
-                const std::string & definition);
-        std::string getCString() const;
-        std::string getName() const { return m_name; }
-        bool getProcessFlag() const { return m_process; }
-        void setProcessFlag(bool p) { m_process = p; }
-        void addData(const std::string & d) { m_data += d; }
-        std::string getData() const { return m_data; }
-        void addCode(const std::string & c) { m_code += c; m_process = true; }
-        std::string getCode() const { return m_code; }
-        std::string getClassDefinition() const;
-        std::string getProcessMethod() const;
-        std::string getIdentifier() const { return "TK_" + m_name; }
-        std::string getClassName() const { return "Tk" + m_name; }
-
-    protected:
-        std::string m_name;
-        std::string m_definition;
-        bool m_process;
-        std::string m_data;
-        std::string m_code;
-};
-
-typedef refptr<TokenDefinition> TokenDefinitionRef;
-
-#endif

UPGRADING.md

@@ -0,0 +1,110 @@
+## v5.0.0
+
+The generated API for tree generation mode (`tree;`) has been changed
+significantly for this version.
+Aside from the lexer user code block matched text rename described below, the
+lexer/parser value APIs for non-tree grammars are unchanged.
+
+### Lexer user code block matched text
+
+The matched text argument passed to lexer user code blocks has been renamed
+from `match` to `match_text` for all target languages.
+
+- C, C++, and D: rename references to `match` in lexer user code blocks to
+  `match_text` (for example `$$ = match[0];` becomes `$$ = match_text[0];`).
+
+The `match_length` argument (C, C++) is unchanged.
+
+### Tree memory management
+
+- Remove all calls to `p_tree_delete()` / `p_tree_delete_XXX()`. Tree nodes now
+  live in the parser context and are freed by `p_context_delete()`.
+- Tree node handles (returned by `p_result()` and the field accessors) are only
+  valid while the context is alive. Do not use them after `p_context_delete()`.
+
+### Tree node field access
+
+Tree nodes are now referenced by handle values instead of pointers, and field
+access differs per target language:
+
+- C: replace `node->field` with the accessor function `p_TYPE_field(node)`, or
+  use the tree walk macro `p_tree_walk_TYPE(node, field1, field2, ...)`. Replace
+  `x != NULL` / `x == NULL` node checks with `p_node_valid(x)` /
+  `!p_node_valid(x)`. Read positions with `p_node_position(node)` /
+  `p_node_end_position(node)`, token payload with `p_TYPE_token(node)` /
+  `p_TYPE_pvalue(node)` or `p_node_data(node)->field`, and compare node identity
+  with `p_node_id(a) == p_node_id(b)`.
+- C++: replace `node->field` with the handle method `node.field()`. Use
+  `node.valid()`, `node.position()`, `node.token()`, `node.pvalue()`, and
+  `node.data()->field` for user token fields. (The C-style functions and macros
+  above are also available.)
+- D: replace pointer declarations (`Start * s`) with value handles (`Start s`)
+  and replace `x !is null` / `x is null` with `x.valid` / `!x.valid`. Field
+  access syntax (`node.field.field`) is otherwise unchanged.
+
+### Tree-mode parser rule user code
+
+In tree generation mode `$$` and `$1`, `$2`, ... now expand to node handles.
+Reference child fields through the target-language accessors above (for example
+`$$->pA->pToken1->pvalue` becomes `p_tree_walk_Start($$, pA, pToken1, pvalue)`
+in C, `$$.pA().pToken1().pvalue()` in C++, and `$$.pA.pToken1.pvalue` in D).
+
+### Pointers into tree node storage
+
+Tree nodes previously each had their own allocation, so a pointer to a node
+stayed valid for the life of the tree. They are now held in a single array
+which is reallocated as it grows, so a pointer or reference into that array may
+be invalidated whenever a new node is created.
+
+New nodes are created while parsing, so this matters for a pointer taken in a
+tree-mode parser rule user code block, which runs before the parse has
+finished. Keep the node handle instead, which stores a node ID rather than an
+address and stays valid, and obtain the pointer from it when it is needed.
+
+For example, replace a saved pointer:
+
+```
+context_user_fields <<
+    p_node_data_t * saved;
+>>
+Items -> Items a << ${context.saved} = p_node_data($$); >>
+```
+
+with a saved handle:
+
+```
+context_user_fields <<
+    Items saved_node;
+>>
+Items -> Items a << ${context.saved_node} = $$; >>
+```
+
+```
+p_node_data_t * data = p_node_data(context->saved_node);
+```
+
+Once parsing has finished, no further nodes are created, so a pointer obtained
+after `p_parse()` returns stays valid until the context is deleted, as long as
+no further parsing is performed with the same context.
+
+## v4.0.0
+
+### API Changes
+
+- Replace any calls to `p_context_init()` with `p_context_new()`.
+- Replace any references to the address of a statically allocated context
+  structure with the pointer returned from `p_context_init()` (e.g. `&context`
+  -> `context`).
+- Add a call to `p_context_delete()` (for C or C++) after lexing/parsing to
+  reclaim context memory.
+- Rename `p_free_tree()` calls to `p_tree_delete()`.
+- Change `free_token_node` statement calls from taking a function name argument
+  to taking a user code block.
+
+## v3.0.0
+
+### Grammar Changes
+
+- Rename `ast;` statement to `tree;`.
+- Rename `ast_prefix;` statement to `tree_prefix;`.
+- Rename `ast_suffix;` statement to `tree_suffix;`.

assets/parser.h.erb

@@ -0,0 +1,254 @@
+/**
+ * @file
+ *
+ * This file is generated by Propane.
+ */
+
+#pragma once
+
+#include <stdint.h>
+#include <stddef.h>
+<% if @cpp %>
+#include <vector>
+<% end %>
+
+/**************************************************************************
+ * Public types
+ *************************************************************************/
+
+/* Result codes. */
+#define <%= @grammar.prefix.upcase %>SUCCESS 0u
+#define <%= @grammar.prefix.upcase %>DECODE_ERROR 1u
+#define <%= @grammar.prefix.upcase %>UNEXPECTED_INPUT 2u
+#define <%= @grammar.prefix.upcase %>UNEXPECTED_TOKEN 3u
+#define <%= @grammar.prefix.upcase %>DROP 4u
+#define <%= @grammar.prefix.upcase %>EOF 5u
+#define <%= @grammar.prefix.upcase %>USER_TERMINATED 6u
+
+/** Token type. */
+typedef <%= get_type_for(@grammar.terminate_token_id) %> <%= @grammar.prefix %>token_t;
+
+/** Token IDs. */
+<% @grammar.tokens.each_with_index do |token, index| %>
+#define TOKEN_<%= token.code_name %> <%= index %>u
+<%   unless token.id == index %>
+<%     raise "Token ID (#{token.id}) does not match index (#{index}) for token #{token.name}!" %>
+<%   end %>
+<% end %>
+#define INVALID_TOKEN_ID <%= @grammar.invalid_token_id %>u
+#define TERMINATE_TOKEN_ID <%= @grammar.terminate_token_id %>u
+
+/** Code point type. */
+typedef uint32_t <%= @grammar.prefix %>code_point_t;
+
+/**
+ * A structure to keep track of input position.
+ *
+ * This is useful for reporting errors, etc...
+ */
+typedef struct
+{
+    /** Input text row (1-based). */
+    uint32_t row;
+
+    /** Input text column (1-based). */
+    uint32_t col;
+} <%= @grammar.prefix %>position_t;
+
+/** Return whether the position is valid. */
+#define <%= @grammar.prefix %>position_valid(p) ((p).row != 0u)
+
+/** User header code blocks. */
+<%= @grammar.code_blocks.fetch("header", "") %>
+
+<% if @grammar.tree %>
+/** Parser values type. */
+typedef <%= @grammar.ptype %> <%= @grammar.prefix %>value_t;
+<% else %>
+/** Parser values type(s). */
+typedef union
+{
+<%   @grammar.ptypes.each do |name, typestring| %>
+    <%= typestring %> v_<%= name %>;
+<%   end %>
+} <%= @grammar.prefix %>value_t;
+
+/** Parser value constructor(s). */
+<%   @grammar.ptypes.each do |name, typestring| %>
+static inline <%= @grammar.prefix %>value_t <%= @grammar.prefix %>value<%= name == "default" ? "" : "_#{name}" %>(<%= typestring %> v)
+{
+    return (<%= @grammar.prefix %>value_t){.v_<%= name %> = v};
+}
+<%   end %>
+
+/** Parser value accessor(s). */
+<%   @grammar.ptypes.each do |name, typestring| %>
+static inline <%= typestring %> <%= @grammar.prefix %>value_get<%= name == "default" ? "" : "_#{name}" %>(<%= @grammar.prefix %>value_t const * pvalue)
+{
+    return pvalue->v_<%= name %>;
+}
+<%   end %>
+<% end %>
+
+<% if @grammar.tree %>
+/** Tree node ID type (index into the context node arena). ID 0 is null. */
+typedef uint32_t <%= @grammar.prefix %>node_id_t;
+
+/**
+ * Tree node record.
+ *
+ * All tree nodes are stored contiguously in the context node arena. Child
+ * links are stored in a shared children array: a node's children
+ * occupy children[child_offset .. child_offset + n_fields]. Token payload
+ * fields (token, pvalue, and any user fields) are only meaningful when
+ * is_token is nonzero.
+ */
+typedef struct
+{
+    <%= @grammar.prefix %>position_t position;
+    <%= @grammar.prefix %>position_t end_position;
+    <%= @grammar.prefix %>node_id_t child_offset;
+    uint16_t n_fields;
+    uint8_t is_token;
+    <%= @grammar.prefix %>token_t token;
+    <%= @grammar.prefix %>value_t pvalue;
+<%= @grammar.token_user_fields %>
+} <%= @grammar.prefix %>node_data_t;
+<% end %>
+
+/** Lexed token information. */
+typedef struct
+{
+    /** Text position of first code point in token. */
+    <%= @grammar.prefix %>position_t position;
+
+    /** Text position of last code point in token. */
+    <%= @grammar.prefix %>position_t end_position;
+
+    /** Number of input bytes used by the token. */
+    size_t length;
+
+    /** Token that was lexed. */
+    <%= @grammar.prefix %>token_t token;
+
+    /** Parser value associated with the token. */
+    <%= @grammar.prefix %>value_t pvalue;
+} <%= @grammar.prefix %>token_info_t;
+
+typedef struct <%= @grammar.prefix %>context_s <%= @grammar.prefix %>context_t;
+
+<% if @grammar.tree %>
+<%= c_tree_handle_types_header %>
+
+<% end %>
+/**
+ * Lexer and parser context.
+ *
+ * The user must allocate an instance of this structure and pass it to any
+ * public API function.
+ */
+struct <%= @grammar.prefix %>context_s
+{
+    /* Lexer context data. */
+
+    /** Input text. */
+    uint8_t const * input;
+
+    /** Input text length. */
+    size_t input_length;
+
+    /** Input text index (byte offset). */
+    size_t input_index;
+
+    /** Input text position (row/column). */
+    <%= @grammar.prefix %>position_t text_position;
+
+    /** Current lexer mode. */
+    size_t mode;
+
+    /* Parser context data. */
+
+    /** Parse result value. */
+<% if @grammar.tree %>
+    <%= @grammar.prefix %>node_id_t parse_result;
+
+<%   if @cpp %>
+    /** Tree node arena. Node ID 0 is reserved as the null node. */
+    std::vector<<%= @grammar.prefix %>node_data_t> <%= @grammar.prefix %>tree_nodes;
+
+    /** Shared tree child links. */
+    std::vector<<%= @grammar.prefix %>node_id_t> <%= @grammar.prefix %>tree_children;
+<%   else %>
+    /** Tree node arena. Node ID 0 is reserved as the null node. */
+    <%= @grammar.prefix %>node_data_t * <%= @grammar.prefix %>tree_nodes;
+    size_t <%= @grammar.prefix %>tree_nodes_length;
+    size_t <%= @grammar.prefix %>tree_nodes_capacity;
+
+    /** Shared tree child links. */
+    <%= @grammar.prefix %>node_id_t * <%= @grammar.prefix %>tree_children;
+    size_t <%= @grammar.prefix %>tree_children_length;
+    size_t <%= @grammar.prefix %>tree_children_capacity;
+<%   end %>
+<% else %>
+    <%= @grammar.prefix %>value_t parse_result;
+<% end %>
+
+    /** Unexpected token received. */
+    <%= @grammar.prefix %>token_t token;
+
+    /** User terminate code. */
+    size_t user_terminate_code;
+
+<%= @grammar.context_user_fields %>
+};
+
+<% if @grammar.tree %>
+<%= c_tree_types_header %>
+<% end %>
+
+/**************************************************************************
+ * Public data
+ *************************************************************************/
+
+/** Token names. */
+extern const char * <%= @grammar.prefix %>token_names[];
+
+<%= @grammar.prefix %>context_t * <%= @grammar.prefix %>context_new(uint8_t const * input, size_t input_length);
+
+void <%= @grammar.prefix %>context_delete(<%= @grammar.prefix %>context_t * context);
+
+size_t <%= @grammar.prefix %>decode_code_point(uint8_t const * input, size_t input_length,
+        <%= @grammar.prefix %>code_point_t * out_code_point, uint8_t * out_code_point_length);
+
+size_t <%= @grammar.prefix %>lex(<%= @grammar.prefix %>context_t * context, <%= @grammar.prefix %>token_info_t * out_token_info);
+
+size_t <%= @grammar.prefix %>parse(<%= @grammar.prefix %>context_t * context);
+<%   @grammar.start_rules.each_with_index do |start_rule, i| %>
+size_t <%= @grammar.prefix %>parse_<%= start_rule %>(<%= @grammar.prefix %>context_t * context);
+size_t <%= @grammar.prefix %>parse_inner_<%= start_rule %>(<%= @grammar.prefix %>context_t * context,
+        <%= @grammar.prefix %>token_t const * follow_tokens, size_t n_follow_tokens);
+<%   end %>
+
+<% if @grammar.tree %>
+<%= h_type(@grammar.start_rules[0]) %> <%= @grammar.prefix %>result(<%= @grammar.prefix %>context_t * context);
+<%   @grammar.start_rules.each_with_index do |start_rule, i| %>
+<%= h_type(start_rule) %> <%= @grammar.prefix %>result_<%= start_rule %>(<%= @grammar.prefix %>context_t * context);
+<%   end %>
+<% else %>
+<%= start_rule_type[1] %> <%= @grammar.prefix %>result(<%= @grammar.prefix %>context_t * context);
+<%   @grammar.start_rules.each_with_index do |start_rule, i| %>
+<%= start_rule_type(i)[1] %> <%= @grammar.prefix %>result_<%= start_rule %>(<%= @grammar.prefix %>context_t * context);
+<%   end %>
+<% end %>
+
+<%= @grammar.prefix %>position_t <%= @grammar.prefix %>position(<%= @grammar.prefix %>context_t * context);
+
+void <%= @grammar.prefix %>set_position(<%= @grammar.prefix %>context_t * context, <%= @grammar.prefix %>position_t position);
+
+size_t <%= @grammar.prefix %>input_index(<%= @grammar.prefix %>context_t * context);
+
+void <%= @grammar.prefix %>set_input_index(<%= @grammar.prefix %>context_t * context, size_t input_index);
+
+size_t <%= @grammar.prefix %>user_terminate_code(<%= @grammar.prefix %>context_t * context);
+
+<%= @grammar.prefix %>token_t <%= @grammar.prefix %>token(<%= @grammar.prefix %>context_t * context);

bin/propane

@@ -0,0 +1,5 @@
+#!/usr/bin/env ruby
+
+require "propane"
+
+exit Propane::CLI.run(ARGV.dup)

extra/vim/ftdetect/propane.vim

@@ -0,0 +1 @@
+au BufNewFile,BufRead *.propane set filetype=propane

extra/vim/syntax/propane.vim

@@ -0,0 +1,85 @@
+" Vim syntax file for Propane
+" Language: propane
+" Maintainer: Josh Holtrop
+" URL: https://github.com/holtrop/propane
+
+if exists("b:current_syntax")
+  finish
+endif
+
+" Guess the language of the user code blocks from their contents so that the
+" matching syntax file can be included below. b:propane_subtype may also be set
+" before this file is sourced to select the language explicitly.
+if !exists("b:propane_subtype")
+  " Rust markers. Each keyword requires the syntax that follows it in Rust so
+  " that a plain identifier of the same name in another language does not match
+  " (`int fn = 3;' in C, for example). Type names are only accepted within a
+  " `ptype' statement for the same reason.
+  let s:rust = '\<let\s\+\%(mut\s\+\)\?\w'
+  let s:rust .= '\|\<fn\s\+\w\+\s*('
+  let s:rust .= '\|&mut\>\|\<pub\s\+\w\|\<impl\s\+\w'
+  let s:rust .= '\|#\[\|\<use\s\+\%(std\|core\)::'
+  let s:rust .= '\|\<ptype\>[^;]*\<\%(isize\|usize\|i8\|i16\|i32\|i64\|i128'
+  let s:rust .= '\|u8\|u16\|u32\|u64\|u128\|f32\|f64\|String\)\>'
+  " D markers. These are spellings that have no valid C, C++, or Rust
+  " equivalent, so `import' is deliberately not among them: it is a D keyword
+  " but is also a C++20 module declaration.
+  let s:d = '\<foreach\%(_reverse\)\?\s*([^)]*;'
+  let s:d .= '\|\~=\|\<static\s\+if\s*(\|\<version\s*(\s*\w\+\s*)'
+  let s:d .= '\|\<scope\s*(\s*\%(exit\|failure\|success\)\s*)'
+  let s:d .= '\|\<\%(unittest\|mixin\|immutable\|__gshared\|invariant\)\>'
+  let s:d .= '\|\<alias\s\+\w\+\s*=\|\<enum\s\+\w\+\s*='
+  let s:d .= '\|@\%(property\|safe\|trusted\|system\|nogc\|disable\)\>'
+  let s:d .= '\|\<is\s\+null\>\|\<cast\s*(\s*\w\+\s*)'
+  let s:d .= '\|\<write\%(ln\|fln\|f\)\s*('
+  let s:d .= '\|\<\%(dchar\|dstring\|wstring\|cent\|ucent\)\>'
+  " A module import on its own is ambiguous between D and C++20, so only take
+  " it as D when nothing else in the file looks like C++.
+  let s:import = '\<import\s\+[A-Za-z_][A-Za-z0-9_.]*\s*;'
+  let s:cpp = '::\|\<template\s*<\|\<namespace\>\|\<nullptr\>\|#include\s*[<"]'
+  if search(s:rust, 'nw') > 0
+    let b:propane_subtype = "rust"
+  elseif search(s:d, 'nw') > 0
+    let b:propane_subtype = "d"
+  elseif search(s:import, 'nw') > 0 && search(s:cpp, 'nw') == 0
+    let b:propane_subtype = "d"
+  else
+    let b:propane_subtype = "cpp"
+  endif
+  unlet s:rust s:d s:import s:cpp
+endif
+
+exe "syn include @propaneTarget syntax/".b:propane_subtype.".vim"
+
+syn region propaneTarget matchgroup=propaneDelimiter start="<<" end=">>$" contains=@propaneTarget keepend
+
+syn match propaneComment "#.*"
+syn match propaneFieldAlias ":[a-zA-Z0-9_]\+" contains=propaneFieldOperator
+syn match propaneFieldOperator ":" contained
+syn match propaneOperator "?"
+" The right-hand side of a rule (after '->' up to '<<' or ';') lists symbol
+" names that may coincide with propane keywords (e.g. 'start', 'token',
+" 'tree'). Wrap it in a region that excludes keyword matches so those names
+" are not highlighted as keywords. The '<<' is left unconsumed so the
+" propaneTarget region can still match it.
+syn region propaneRuleRhs matchgroup=propaneOperator start="->" end="\ze<<" end=";" contains=propaneFieldAlias,propaneRuleOperator,propaneComment keepend
+syn match propaneRuleOperator "?" contained
+" Keywords that introduce a user-defined name. The name is consumed by
+" propaneName via nextgroup so a name matching a keyword (e.g. 'token start')
+" is not highlighted as a keyword. These must be a match (not syn keyword)
+" because a syn keyword always wins over a contained nextgroup match.
+syn match propaneNameDecl "\<\%(tokenid\|token\|lex_fn\|module\|start\|tree_prefix\|tree_suffix\)\>" nextgroup=propaneName skipwhite
+syn match propaneName "\<\h\w*\>" contained
+syn match propaneKeyword "\<\%(context_user_fields\|drop\|free_token_node\|noline\|on_token_node\|prefix\|ptype\|token_user_fields\|tree\)\>"
+
+syn region propaneRegex start="/" end="/" skip="\v\\\\|\\/"
+
+hi def link propaneComment Comment
+hi def link propaneKeyword Keyword
+hi def link propaneNameDecl Keyword
+hi def link propaneRegex String
+hi def link propaneOperator Operator
+hi def link propaneRuleOperator Operator
+hi def link propaneFieldOperator Operator
+hi def link propaneDelimiter Delimiter
+hi def link propaneFieldAlias Identifier

imbecile.cc

@@ -1,101 +0,0 @@
-
-#include <getopt.h>
-
-#include <iostream>
-#include <fstream>
-
-#include "refptr.h"
-#include "Parser.h"
-
-using namespace std;
-
-string buildOutputFilename(string & input_fname);
-
-int main(int argc, char * argv[])
-{
-    int longind = 1;
-    int opt;
-    Parser p;
-    string outfile;
-
-    static struct option longopts[] = {
-        /* name, has_arg, flag, val */
-        { "classname", required_argument, NULL, 'c' },
-        { "extension", required_argument, NULL, 'e' },
-        { "namespace", required_argument, NULL, 'n' },
-        { "outfile", required_argument, NULL, 'o' },
-        { NULL, 0, NULL, 0 }
-    };
-
-    while ((opt = getopt_long(argc, argv, "", longopts, &longind)) != -1)
-    {
-        switch (opt)
-        {
-            case 'c':   /* classname */
-                p.setClassName(optarg);
-                break;
-            case 'e':   /* extension */
-                p.setExtension(optarg);
-                break;
-            case 'n':   /* namespace */
-                p.setNamespace(optarg);
-                break;
-            case 'o':   /* outfile */
-                outfile = optarg;
-                break;
-        }
-    }
-
-    if (optind >= argc)
-    {
-        cerr << "Usage: imbecile [options] <input-file>" << endl;
-        return 1;
-    }
-
-    string input_fname = argv[optind];
-    ifstream ifs;
-    ifs.open(input_fname.c_str(), ios::binary);
-    if (!ifs.is_open())
-    {
-        cerr << "Error opening input file: '" << input_fname << "'";
-        return 2;
-    }
-    ifs.seekg(0, ios_base::end);
-    int size = ifs.tellg();
-    ifs.seekg(0, ios_base::beg);
-    char * buff = new char[size];
-    ifs.read(buff, size);
-    ifs.close();
-
-    if (outfile == "")
-        outfile = buildOutputFilename(input_fname);
-
-    if (!p.parseInputFile(buff, size))
-    {
-        cerr << "Error parsing " << input_fname << endl;
-        return 3;
-    }
-    if (!p.write(outfile))
-    {
-        cerr << "Error processing " << input_fname << endl;
-        return 4;
-    }
-
-    delete[] buff;
-    return 0;
-}
-
-string buildOutputFilename(string & input_fname)
-{
-    string outfile;
-    size_t len = input_fname.length();
-    if (len > 2 && input_fname.substr(len - 2) == ".I")
-    {
-        outfile = input_fname.substr(0, len - 2);
-    }
-    else
-    {
-        outfile = input_fname;
-    }
-    return outfile;
-}

lib/propane.rb

@@ -0,0 +1,48 @@
+require "erb"
+require "set"
+require "stringio"
+require_relative "propane/assets"
+require_relative "propane/cli"
+require_relative "propane/code_point_range"
+require_relative "propane/fa"
+require_relative "propane/fa/state"
+require_relative "propane/fa/state/transition"
+require_relative "propane/generator"
+require_relative "propane/grammar"
+require_relative "propane/lexer"
+require_relative "propane/lexer/dfa"
+require_relative "propane/parser"
+require_relative "propane/parser/item"
+require_relative "propane/parser/item_set"
+require_relative "propane/pattern"
+require_relative "propane/regex"
+require_relative "propane/regex/nfa"
+require_relative "propane/regex/unit"
+require_relative "propane/rule_set"
+require_relative "propane/rule"
+require_relative "propane/token"
+require_relative "propane/util"
+require_relative "propane/version"
+
+class Propane
+
+  class Error < RuntimeError
+  end
+
+  class << self
+
+    def run(input_file, output_file, log_file, options)
+      begin
+        grammar = Grammar.new(File.read(input_file), input_file)
+        generator = Generator.new(grammar, output_file, log_file, options)
+        generator.generate
+      rescue Error => e
+        $stderr.puts e.message
+        return 2
+      end
+      return 0
+    end
+
+  end
+
+end

lib/propane/assets.rb

@@ -0,0 +1,10 @@
+class Propane
+  module Assets
+    class << self
+      def get(name)
+        path = File.join(File.dirname(File.expand_path(__FILE__)), "../../assets/#{name}")
+        File.binread(path)
+      end
+    end
+  end
+end

lib/propane/cli.rb

@@ -0,0 +1,62 @@
+class Propane
+  module CLI
+
+    USAGE = <<EOF
+Usage: #{$0} [options] <input-file> <output-file>
+Options:
+  -h, --help  Show this usage and exit.
+  --log LOG   Write log file. This will show all parser states and their
+              associated shifts and reduces. It can be helpful when
+              debugging a grammar.
+  --version   Show program version and exit.
+  -w          Treat warnings as errors. This option will treat shift/reduce
+              conflicts and unused rules as fatal errors and will print
+              them to stderr in addition to the log file.
+EOF
+
+    class << self
+
+      def run(args)
+        params = []
+        options = {}
+        log_file = nil
+        i = 0
+        while i < args.size
+          arg = args[i]
+          case arg
+          when "--log"
+            if i + 1 < args.size
+              i += 1
+              log_file = args[i]
+            end
+          when "--version"
+            puts "propane version #{VERSION}"
+            return 0
+          when "-h", "--help"
+            puts USAGE
+            return 0
+          when "-w"
+            options[:warnings_as_errors] = true
+          when /^-/
+            $stderr.puts "Error: unknown option #{arg}"
+            return 1
+          else
+            params << arg
+          end
+          i += 1
+        end
+        if params.size != 2
+          $stderr.puts "Error: specify input and output files"
+          return 1
+        end
+        unless File.readable?(params[0])
+          $stderr.puts "Error: cannot read #{params[0]}"
+          return 2
+        end
+        Propane.run(*params, log_file, options)
+      end
+
+    end
+
+  end
+end

lib/propane/code_point_range.rb

@@ -0,0 +1,84 @@
+class Propane
+  class CodePointRange
+
+    MAX_CODE_POINT = 0x7FFFFFFF
+
+    attr_reader :first
+    attr_reader :last
+
+    include Comparable
+
+    # Build a CodePointRange
+    def initialize(first, last = nil)
+      @first = first.ord
+      if last
+        @last = last.ord
+        if @last < @first
+          raise "Invalid CodePointRange: last code point must be > first code point"
+        end
+      else
+        @last = @first
+      end
+    end
+
+    def <=>(other)
+      if self.first != other.first
+        @first <=> other.first
+      else
+        @last <=> other.last
+      end
+    end
+
+    def include?(v)
+      if v.is_a?(CodePointRange)
+        @first <= v.first && v.last <= @last
+      else
+        @first <= v && v <= @last
+      end
+    end
+
+    def size
+      @last - @first + 1
+    end
+
+    class << self
+
+      def invert_ranges(code_point_ranges)
+        new_ranges = []
+        last_cp = -1
+        code_point_ranges.sort.each do |code_point_range|
+          if code_point_range.first > (last_cp + 1)
+            new_ranges << CodePointRange.new(last_cp + 1, code_point_range.first - 1)
+            last_cp = code_point_range.last
+          else
+            last_cp = [last_cp, code_point_range.last].max
+          end
+        end
+        if last_cp < MAX_CODE_POINT
+          new_ranges << CodePointRange.new(last_cp + 1, MAX_CODE_POINT)
+        end
+        new_ranges
+      end
+
+      def first_subrange(code_point_ranges)
+        code_point_ranges.sort.reduce do |result, code_point_range|
+          if code_point_range.include?(result.first)
+            if code_point_range.last < result.last
+              code_point_range
+            else
+              result
+            end
+          else
+            if code_point_range.first <= result.last
+              CodePointRange.new(result.first, code_point_range.first - 1)
+            else
+              result
+            end
+          end
+        end
+      end
+
+    end
+
+  end
+end

lib/propane/fa.rb

@@ -0,0 +1,61 @@
+class Propane
+
+  class FA
+
+    attr_reader :start_state
+
+    def initialize
+      @start_state = State.new
+    end
+
+    def to_s
+      chr = lambda do |value|
+        if value < 32 || value > 127
+          "{#{value}}"
+        else
+          value.chr
+        end
+      end
+      rv = ""
+      states = enumerate
+      states.each do |state, id|
+        accepts_s = state.accepts ? " #{state.accepts}" : ""
+        rv += "#{id}#{accepts_s}:\n"
+        state.transitions.each do |transition|
+          if transition.nil?
+            range_s = "nil"
+          else
+            range_s = chr[transition.code_point_range.first]
+            if transition.code_point_range.size > 1
+              range_s += "-" + chr[transition.code_point_range.last]
+            end
+          end
+          accepts_s = transition.destination.accepts ? " #{transition.destination.accepts}" : ""
+          rv += "  #{range_s} => #{states[transition.destination]}#{accepts_s}\n"
+        end
+      end
+      rv
+    end
+
+    def enumerate
+      @_enumerated ||=
+        begin
+          id = 0
+          states = {}
+          visit = lambda do |state|
+            unless states.include?(state)
+              states[state] = id
+              id += 1
+              state.transitions.each do |transition|
+                visit[transition.destination]
+              end
+            end
+          end
+          visit[@start_state]
+          states
+        end
+    end
+
+  end
+
+end

lib/propane/fa/state.rb

@@ -0,0 +1,51 @@
+class Propane
+  class FA
+
+    class State
+
+      attr_accessor :accepts
+      attr_reader :transitions
+
+      def initialize
+        @transitions = []
+      end
+
+      def add_transition(code_point_range, destination)
+        @transitions << Transition.new(code_point_range, destination)
+      end
+
+      # Determine the set of states that can be reached by nil transitions.
+      # from this state.
+      #
+      # @return [Set<NFA::State>]
+      #   Set of states.
+      def nil_transition_states
+        states = Set[self]
+        analyze_state = lambda do |state|
+          state.nil_transitions.each do |transition|
+            unless states.include?(transition.destination)
+              states << transition.destination
+              analyze_state[transition.destination]
+            end
+          end
+        end
+        analyze_state[self]
+        states
+      end
+
+      def nil_transitions
+        @transitions.select do |transition|
+          transition.nil?
+        end
+      end
+
+      def cp_transitions
+        @transitions.reject do |transition|
+          transition.nil?
+        end
+      end
+
+    end
+
+  end
+end

lib/propane/fa/state/transition.rb

@@ -0,0 +1,23 @@
+class Propane
+  class FA
+    class State
+
+      class Transition
+
+        attr_reader :code_point_range
+        attr_reader :destination
+
+        def initialize(code_point_range, destination)
+          @code_point_range = code_point_range
+          @destination = destination
+        end
+
+        def nil?
+          @code_point_range.nil?
+        end
+
+      end
+
+    end
+  end
+end

lib/propane/generator.rb

@@ -0,0 +1,880 @@
+class Propane
+
+  class Generator
+
+    def initialize(grammar, output_file, log_file, options)
+      @grammar = grammar
+      @output_file = output_file
+      if log_file
+        @log = File.open(log_file, "wb")
+      else
+        @log = StringIO.new
+      end
+      @language =
+        if output_file.end_with?(".d")
+          "d"
+        elsif output_file.end_with?(".c")
+          "c"
+        elsif output_file =~ %r{\.(cc|cpp|cxx)$}
+          @cpp = true
+          "c"
+        elsif output_file.end_with?(".rs")
+          "rust"
+        else
+          raise Error.new("Could not determine target language from output file name (#{output_file})")
+        end
+      @options = options
+      process_grammar!
+    end
+
+    def generate
+      extensions = [nil]
+      if @language == "c"
+        extensions += %w[h]
+      end
+      extensions.each do |extension|
+        template_language = @language == "rust" ? "rs" : @language
+        template = Assets.get("parser.#{extension || template_language}.erb")
+        if extension
+          output_file = @output_file.sub(%r{\.[a-z]+$}, ".#{extension}")
+        else
+          output_file = @output_file
+        end
+        erb = ERB.new(template, trim_mode: "<>")
+        # Rust has no #line directive support. For a Rust target the directives
+        # that the grammar embeds around user code blocks are replaced with
+        # comments naming the grammar file and line number the code came from,
+        # so that the origin of a section of user code can still be found by
+        # reading up from a compiler diagnostic pointing into the generated
+        # module.
+        user_code_origin = nil
+        result = erb.result(binding.clone).lines.each_with_index.map do |line, i|
+          if @language == "rust"
+            if md = line.match(/^#line (\d+) "([^"]*)"/)
+              user_code_origin = "#{md[2]} line #{md[1]}"
+              line.sub(/^#line \d+ "[^"]*"/, %[/* Begin user code from #{user_code_origin}. */])
+            elsif line == "#linereset\n"
+              %[/* End user code from #{user_code_origin}. */\n]
+            else
+              line
+            end
+          elsif line == "#linereset\n"
+            %[#line #{i + 2} "#{output_file}"\n]
+          else
+            line
+          end
+        end.join
+        File.open(output_file, "wb") do |fh|
+          fh.write(result)
+        end
+      end
+      @log.close
+    end
+
+    private
+
+    def process_grammar!
+      # Assign default pattern mode to patterns without a mode assigned.
+      found_default = false
+      @grammar.patterns.each do |pattern|
+        if pattern.modes.empty?
+          pattern.modes << "default"
+          found_default = true
+        end
+        pattern.ptypename ||= "default"
+      end
+      unless found_default
+        raise Error.new("No patterns found for default mode")
+      end
+      check_ptypes!
+      # Add EOF token.
+      @grammar.tokens << Token.new("$EOF", nil, nil)
+      tokens_by_name = {}
+      @grammar.tokens.each_with_index do |token, token_id|
+        # Assign token ID.
+        token.id = token_id
+        # Set default ptypename if none given.
+        token.ptypename ||= "default"
+        # Check for token name conflicts.
+        if tokens_by_name.include?(token.name)
+          raise Error.new("Duplicate token name #{token.name.inspect}")
+        end
+        tokens_by_name[token.name] = token
+      end
+      # Create real start rule(s).
+      real_start_rules = @grammar.start_rules.map do |start_rule|
+        unless @grammar.rules.find {|rule| rule.name == start_rule}
+          raise Error.new("Start rule `#{start_rule}` not found")
+        end
+        Rule.new("$#{start_rule}", [start_rule, "$EOF"], nil, nil, nil)
+      end
+      # Add real start rules before user-given rules.
+      @grammar.rules = real_start_rules + @grammar.rules
+      # Generate and add rules for optional components.
+      generate_optional_component_rules!(tokens_by_name)
+      # Build rule sets.
+      rule_sets = {}
+      rule_set_id = @grammar.tokens.size
+      @grammar.rules.each_with_index do |rule, rule_id|
+        # Assign rule ID.
+        rule.id = rule_id
+        # Check for token/rule name conflict.
+        if tokens_by_name.include?(rule.name)
+          raise Error.new("Rule name collides with token name #{rule.name.inspect}")
+        end
+        # Build rule sets of all rules with the same name.
+        unless rule_sets[rule.name]
+          rule_sets[rule.name] = RuleSet.new(rule.name, rule_set_id)
+          rule_set_id += 1
+        end
+        rule_set = rule_sets[rule.name]
+        if rule_set.ptypename && rule.ptypename && rule_set.ptypename != rule.ptypename
+          raise Error.new("Conflicting ptypes for rule #{rule.name}")
+        end
+        rule_set.ptypename ||= rule.ptypename
+        rule.rule_set = rule_set
+        rule_set << rule
+      end
+      rule_sets.each do |name, rule_set|
+        rule_set.ptypename ||= "default"
+        # Assign rule set ptypenames back to rules.
+        rule_set.rules.each do |rule|
+          rule.ptypename = rule_set.ptypename
+        end
+      end
+      # Generate lexer user code IDs for lexer patterns with user code blocks.
+      @grammar.patterns.select do |pattern|
+        pattern.code
+      end.each_with_index do |pattern, code_id|
+        pattern.code_id = code_id
+      end
+      # Map rule components from names to Token/RuleSet objects.
+      @grammar.rules.each do |rule|
+        rule.components.map! do |component|
+          if tokens_by_name[component]
+            tokens_by_name[component]
+          elsif rule_sets[component]
+            rule_sets[component]
+          else
+            raise Error.new("Symbol #{component} not found")
+          end
+        end
+      end
+      determine_possibly_empty_rulesets!(rule_sets)
+      rule_sets.each do |name, rule_set|
+        rule_set.finalize(@grammar)
+      end
+      # Generate the lexer.
+      @lexer = Lexer.new(@grammar)
+      # Generate the parser.
+      @parser = Parser.new(@grammar, rule_sets, @log, @options)
+    end
+
+    # Check that any referenced ptypes have been defined.
+    def check_ptypes!
+      (@grammar.patterns + @grammar.tokens + @grammar.rules).each do |potor|
+        if potor.ptypename
+          unless @grammar.ptypes.include?(potor.ptypename)
+            raise Error.new("Error: Line #{potor.line_number}: ptype #{potor.ptypename} not declared. Declare with `ptype` statement.")
+          end
+        end
+      end
+    end
+
+    # Generate and add rules for any optional components.
+    def generate_optional_component_rules!(tokens_by_name)
+      optional_rules_added = Set.new
+      @grammar.rules.each do |rule|
+        rule.components.each do |component|
+          if component =~ /^(.*)\?$/
+            c = $1
+            unless optional_rules_added.include?(component)
+              # Create two rules for the optional component: one empty and
+              # one just matching the component.
+              # We need to find the ptypename for the optional component in
+              # order to copy it to the generated rules.
+              if tokens_by_name[c]
+                # The optional component is a token.
+                ptypename = tokens_by_name[c].ptypename
+              else
+                # The optional component must be a rule, so find any instance
+                # of that rule that specifies a ptypename.
+                ptypename = @grammar.rules.reduce(nil) do |result, rule|
+                  rule.name == c && rule.ptypename ? rule.ptypename : result
+                end
+              end
+              @grammar.rules << Rule.new(component, [], nil, ptypename, rule.line_number)
+              optcode = @grammar.tree ? nil : "$$ = $1;\n"
+              @grammar.rules << Rule.new(component, [c], optcode, ptypename, rule.line_number)
+              optional_rules_added << component
+            end
+          end
+        end
+      end
+    end
+
+    # Determine which grammar rules could expand to empty sequences.
+    #
+    # @param rule_sets [Hash]
+    #   RuleSets.
+    #
+    # @return [void]
+    def determine_possibly_empty_rulesets!(rule_sets)
+      begin
+        newly_discovered_empty_rulesets = false
+        rule_sets.each do |name, rule_set|
+          unless rule_set.could_be_empty?
+            if could_rule_set_be_empty?(rule_set)
+              newly_discovered_empty_rulesets = true
+              rule_set.could_be_empty = true
+            end
+          end
+        end
+      end while newly_discovered_empty_rulesets
+    end
+
+    # Determine whether a RuleSet could be empty.
+    #
+    # @param rule_set [RuleSet]
+    #   RuleSet to test.
+    #
+    # @return [Boolean]
+    #   Whether the RuleSet could be empty.
+    def could_rule_set_be_empty?(rule_set)
+      rule_set.rules.any? do |rule|
+        could_rule_be_empty?(rule)
+      end
+    end
+
+    # Determine whether a Rule could be empty.
+    #
+    # @param rule [Rule]
+    #   Rule to test.
+    #
+    # @return [Boolean]
+    #   Whether the Rule could be empty.
+    def could_rule_be_empty?(rule)
+      i = 0
+      loop do
+        if i == rule.components.size
+          return true
+        end
+        if rule.components[i].is_a?(Token)
+          return false
+        end
+        if !rule.components[i].could_be_empty?
+          return false
+        end
+        i += 1
+      end
+    end
+
+    # Expand expansions in user code block.
+    #
+    # @param code [String]
+    #   User code block.
+    # @param parser [Boolean]
+    #   Whether the user code is for the parser or lexer.
+    # @param rule [Rule, nil]
+    #   The Rule associated with the user code if user code is for the parser.
+    # @param pattern [Pattern, nil]
+    #   The Pattern associated with the user code if user code is for the lexer.
+    #
+    # @return [String]
+    #   Expanded user code block.
+    def expand_code(code, parser, rule, pattern)
+      code = code.gsub(/\$token\(([$\w]+)\)/) do |match|
+        "TOKEN_#{Token.code_name($1)}"
+      end
+      code = code.gsub(/\$terminate\((.*)\);/) do |match|
+        user_terminate_code = $1
+        retval = rule ? "P_USER_TERMINATED" : "TERMINATE_TOKEN_ID"
+        case @language
+        when "c"
+          "context->user_terminate_code = (#{user_terminate_code}); return #{retval};"
+        when "d"
+          "context.user_terminate_code = (#{user_terminate_code}); return #{retval};"
+        when "rust"
+          "context.user_terminate_code = (#{user_terminate_code}); return #{retval};"
+        end
+      end
+      code = code.gsub(/\$\{context\.(\w+)\}/) do |match|
+        fieldname = $1
+        case @language
+        when "c"
+          "context->#{fieldname}"
+        when "d"
+          "context.#{fieldname}"
+        when "rust"
+          "context.#{fieldname}"
+        end
+      end
+      code = code.gsub(/\$\{token\.(\w+)\}/) do |match|
+        fieldname = $1
+        case @language
+        when "c"
+          "token_tree_node->#{fieldname}"
+        when "d"
+          "token_tree_node.#{fieldname}"
+        when "rust"
+          "token_tree_node.#{fieldname}"
+        end
+      end
+      if parser
+        code = code.gsub(/\$\$/) do |match|
+          if @grammar.tree
+            typename = "#{@grammar.tree_prefix}#{rule.name}#{@grammar.tree_suffix}"
+            case @language
+            when "c"
+              tree_handle(typename, "_node_id")
+            when "d"
+              tree_handle(typename, "_node_id")
+            when "rust"
+              tree_handle(typename, "_node_id")
+            end
+          else
+            case @language
+            when "c"
+              "_pvalue->v_#{rule.ptypename}"
+            when "d"
+              "_pvalue.v_#{rule.ptypename}"
+            when "rust"
+              "(*_pvalue.v_#{rule.ptypename}_mut())"
+            end
+          end
+        end
+        code = code.gsub(/\$(\d+)/) do |match|
+          parser_component_reference(rule, $1.to_i)
+        end
+        code = code.gsub(/\$\{(\$|\d+)\.position\}/) do |match|
+          index = $1.to_i
+          "get_rule_position(statevalues, #{index}, n_states, false)"
+        end
+        code = code.gsub(/\$\{(\$|\d+)\.end_position\}/) do |match|
+          index = $1.to_i
+          "get_rule_position(statevalues, #{index}, n_states, true)"
+        end
+        code = code.gsub(/\$\{(\w+)\}/) do |match|
+          aliasname = $1
+          if index = rule.aliases[aliasname]
+            # Field aliases are just a named reference to a positional rule
+            # component, so reuse the same expansion as `$1', `$2', etc. Note
+            # that rule.aliases stores a 0-based component index, so add 1 to
+            # convert it to the 1-based index used for positional references.
+            parser_component_reference(rule, index + 1)
+          else
+            raise Error.new("Field alias '#{aliasname}' not found")
+          end
+        end
+      else
+        code = code.gsub(/\$\$/) do |match|
+          if @grammar.tree
+            case @language
+            when "c"
+              "out_token_info->pvalue"
+            when "d"
+              "out_token_info.pvalue"
+            when "rust"
+              "out_token_info.pvalue"
+            end
+          else
+            case @language
+            when "c"
+              "out_token_info->pvalue.v_#{pattern.ptypename}"
+            when "d"
+              "out_token_info.pvalue.v_#{pattern.ptypename}"
+            when "rust"
+              "(*out_token_info.pvalue.v_#{pattern.ptypename}_mut())"
+            end
+          end
+        end
+        code = code.gsub(/\$\{position\}/) do |match|
+          case @language
+          when "c"
+            "out_token_info->position"
+          when "d"
+            "out_token_info.position"
+          when "rust"
+            "out_token_info.position"
+          end
+        end
+        code = code.gsub(/\$\{end_position\}/) do |match|
+          case @language
+          when "c"
+            "out_token_info->end_position"
+          when "d"
+            "out_token_info.end_position"
+          when "rust"
+            "out_token_info.end_position"
+          end
+        end
+        code = code.gsub(/\$mode\(([a-zA-Z_][a-zA-Z_0-9]*)\)/) do |match|
+          mode_name = $1
+          mode_id = @lexer.mode_id(mode_name)
+          unless mode_id
+            raise Error.new("Lexer mode '#{mode_name}' not found")
+          end
+          case @language
+          when "c"
+            "context->mode = #{mode_id}u"
+          when "d"
+            "context.mode = #{mode_id}u"
+          when "rust"
+            "context.mode = #{mode_id}"
+          end
+        end
+      end
+      code
+    end
+
+    # Expand a positional reference to a parser rule component.
+    #
+    # This is used to expand `$1', `$2', etc. as well as field aliases (which
+    # are just named references to a positional rule component).
+    #
+    # @param rule [Rule]
+    #   The Rule containing the user code.
+    # @param index [Integer]
+    #   1-based index of the rule component to reference.
+    #
+    # @return [String]
+    #   Expanded rule component reference.
+    def parser_component_reference(rule, index)
+      component = rule.components[index - 1]
+      if @grammar.tree
+        # In tree mode a component reference yields a handle to that
+        # component's tree node. An optional component propagates its target
+        # node (or null), so use the optional target's node type.
+        if component.is_a?(RuleSet) && component.optional?
+          component = component.option_target
+        end
+        node_name = component.is_a?(Token) ? "Token" : component.name
+        typename = "#{@grammar.tree_prefix}#{node_name}#{@grammar.tree_suffix}"
+        case @language
+        when "c"
+          tree_handle(typename, "state_values_stack_index(statevalues, -1 - (int)n_states + #{index})->node_id")
+        when "d"
+          tree_handle(typename, "statevalues[$-1-n_states+#{index}].node_id")
+        when "rust"
+          tree_handle(typename, "statevalues[statevalues.len() - 1 - n_states + #{index}].node_id")
+        end
+      else
+        case @language
+        when "c"
+          "state_values_stack_index(statevalues, -1 - (int)n_states + #{index})->pvalue.v_#{component.ptypename}"
+        when "d"
+          "statevalues[$-1-n_states+#{index}].pvalue.v_#{component.ptypename}"
+        when "rust"
+          "statevalues[statevalues.len() - 1 - n_states + #{index}].pvalue.get_v_#{component.ptypename}()"
+        end
+      end
+    end
+
+    # Construct a tree node handle expression for the target language.
+    #
+    # A handle is a small value pairing the parser context with a node ID
+    # (an index into the context's node arena). All handle types share this
+    # layout; the distinct types exist for documentation and, in C, to drive
+    # the tree walk macro's type threading.
+    #
+    # @param typename [String]
+    #   Handle type name.
+    # @param id_expr [String]
+    #   Expression yielding the node ID.
+    # @param parenthesize [Boolean]
+    #   Whether to parenthesize the expression. Parentheses are required where
+    #   the expression is substituted into a user code block, since the
+    #   expression could be followed there by a field access or appear in a
+    #   position where a bare Rust struct literal is not accepted. They are
+    #   unnecessary where the expression stands alone, and Rust warns about
+    #   them there, so this can be disabled for those uses.
+    #
+    # @return [String]
+    #   Handle constructor expression.
+    def tree_handle(typename, id_expr, parenthesize = true)
+      if @cpp
+        "(#{typename}{context, #{id_expr}})"
+      elsif @language == "c"
+        "((#{typename}){context, #{id_expr}})"
+      elsif @language == "rust"
+        expr = "#{typename} { context, id: #{id_expr} }"
+        parenthesize ? "(#{expr})" : expr
+      else
+        "#{typename}(context, #{id_expr})"
+      end
+    end
+
+    # Get the list of non-optional, non-internal rule sets that get a tree node
+    # handle type generated for them.
+    #
+    # @return [Array<Propane::RuleSet>]
+    #   Rule sets with generated tree node handle types.
+    def tree_node_rule_sets
+      @parser.rule_sets.reject do |name, rule_set|
+        name.start_with?("$") || rule_set.optional?
+      end.map {|name, rule_set| rule_set}
+    end
+
+    # Maximum number of chained fields supported by a single C tree walk macro
+    # invocation. Deeper navigation can be expressed by nesting walk calls.
+    C_TREE_WALK_MAX = 16
+
+    # Get the tree node handle type name for a node name.
+    #
+    # @param name [String]
+    #   Rule set name, or "Token".
+    #
+    # @return [String]
+    #   Handle type name.
+    def h_type(name)
+      "#{@grammar.tree_prefix}#{name}#{@grammar.tree_suffix}"
+    end
+
+    # Get the list of all tree node handle type names (Token plus rule sets).
+    #
+    # @return [Array<String>]
+    #   Handle type names.
+    def tree_handle_types
+      [h_type("Token")] + tree_node_rule_sets.map {|rs| h_type(rs.name)}
+    end
+
+    # Enumerate the navigation fields of a rule set's tree node.
+    #
+    # @yield [rtype, field_name, child_type, slot]
+    #   Handle type name, field accessor name, child handle type, and child
+    #   slot index.
+    def each_tree_field(rule_set)
+      rtype = h_type(rule_set.name)
+      rule_set.tree_fields.each_with_index do |fields, slot|
+        fields.each do |field_name, child_type|
+          yield rtype, field_name, child_type, slot
+        end
+      end
+    end
+
+    # Generate the tree node handle type declarations for the header.
+    #
+    # These are emitted before the context structure definition so that a
+    # context_user_fields block can declare a field of a handle type.
+    #
+    # @return [String]
+    #   Handle type declarations.
+    def c_tree_handle_types_header
+      @cpp ? cpp_tree_handle_types_header : c_only_tree_handle_types_header
+    end
+
+    # Generate the remainder of the tree node section for the header.
+    #
+    # This is emitted after the context structure definition since it
+    # dereferences the context and so requires the complete type.
+    #
+    # @return [String]
+    #   Accessors, macros, and out-of-line handle method definitions.
+    def c_tree_types_header
+      @cpp ? cpp_tree_types_header : c_only_tree_types_header
+    end
+
+    # Generate the C (non-C++) tree node handle type section for the header.
+    def c_only_tree_handle_types_header
+      p = @grammar.prefix
+      out = []
+      out << "/** Tree node handle types. @{ */"
+      tree_handle_types.each do |t|
+        out << "typedef struct { #{p}context_t * __context; #{p}node_id_t __id; } #{t};"
+      end
+      out << ""
+      out.join("\n")
+    end
+
+    def c_only_tree_types_header
+      out = []
+      out << c_common_accessors_header
+      out << "/** @} */"
+      out.join("\n")
+    end
+
+    # Generate the C-style (function + macro) tree node accessors shared by the
+    # C and C++ headers. In C++ these are provided in addition to the handle
+    # methods so that C-style code (and the tree walk macros) also works.
+    def c_common_accessors_header
+      p = @grammar.prefix
+      out = []
+      out << "/** Generic tree node accessors (usable on any handle type). */"
+      out << "#define #{p}node_valid(h) ((h).__id != 0u)"
+      out << "#define #{p}node_id(h) ((h).__id)"
+      out << "#define #{p}node_data(h) (&(h).__context->#{p}tree_nodes[(h).__id])"
+      out << "#define #{p}node_position(h) ((h).__context->#{p}tree_nodes[(h).__id].position)"
+      out << "#define #{p}node_end_position(h) ((h).__context->#{p}tree_nodes[(h).__id].end_position)"
+      out << "#define #{p}node_n_fields(h) ((h).__id ? (h).__context->#{p}tree_nodes[(h).__id].n_fields : (uint16_t)0u)"
+      out << ""
+      out << "/** Tree node field accessor functions. */"
+      out << "#{p}token_t #{p}#{h_type("Token")}_token(#{h_type("Token")} node);"
+      out << "#{p}value_t #{p}#{h_type("Token")}_pvalue(#{h_type("Token")} node);"
+      tree_node_rule_sets.each do |rule_set|
+        each_tree_field(rule_set) do |rtype, field_name, child_type, slot|
+          out << "#{child_type} #{p}#{rtype}_#{field_name}(#{rtype} node);"
+        end
+      end
+      out << ""
+      out << c_tree_walk_macros
+      out.join("\n")
+    end
+
+    # Generate the C tree walk macro machinery.
+    def c_tree_walk_macros
+      p = @grammar.prefix
+      max = C_TREE_WALK_MAX
+      out = []
+      out << "/* Tree walk macros: p_tree_walk_<Type>(handle, field, ...). */"
+      out << "#define #{p}CAT_(a, b) a##b"
+      out << "#define #{p}CAT(a, b) #{p}CAT_(a, b)"
+      out << "#define #{p}TA(t, f) #{p}CAT(#{p}CAT(#{p}CAT(#{p}TYPEAFTER_, t), _), f)"
+      out << "#define #{p}ACC(t, f) #{p}CAT(#{p}CAT(#{p}CAT(#{p}, t), _), f)"
+      argn = (1..max).map {|i| "_#{i}"}.join(", ")
+      rseq = (0..max).to_a.reverse.join(", ")
+      out << "#define #{p}ARG_N(#{argn}, N, ...) N"
+      out << "#define #{p}NARG(...) #{p}ARG_N(__VA_ARGS__, #{rseq})"
+      (1..max).each do |n|
+        fparams = (1..n).map {|k| "f#{k}"}.join(", ")
+        call = "h"
+        (1..n).each do |k|
+          texpr = "R"
+          (1...k).each {|j| texpr = "#{p}TA(#{texpr}, f#{j})"}
+          call = "#{p}ACC(#{texpr}, f#{k})(#{call})"
+        end
+        out << "#define #{p}tree_walk_#{n}(R, h, #{fparams}) #{call}"
+      end
+      out << "#define #{p}tree_walk_dispatch(R, h, ...) #{p}CAT(#{p}tree_walk_, #{p}NARG(__VA_ARGS__))(R, h, __VA_ARGS__)"
+      # Type transition map (navigation fields only).
+      tree_node_rule_sets.each do |rule_set|
+        each_tree_field(rule_set) do |rtype, field_name, child_type, slot|
+          out << "#define #{p}TYPEAFTER_#{rtype}_#{field_name} #{child_type}"
+        end
+      end
+      # Per-handle-type walk entry points.
+      tree_handle_types.each do |t|
+        out << "#define #{p}tree_walk_#{t}(...) #{p}tree_walk_dispatch(#{t}, __VA_ARGS__)"
+      end
+      out.join("\n")
+    end
+
+    # Generate the C tree node accessor function definitions for the source.
+    #
+    # @return [String]
+    #   Accessor function definitions.
+    def c_tree_accessor_defs
+      p = @grammar.prefix
+      tt = h_type("Token")
+      out = []
+      out << "#{p}token_t #{p}#{tt}_token(#{tt} node)"
+      out << "{"
+      out << "    return node.__context->#{p}tree_nodes[node.__id].token;"
+      out << "}"
+      out << ""
+      out << "#{p}value_t #{p}#{tt}_pvalue(#{tt} node)"
+      out << "{"
+      out << "    return node.__context->#{p}tree_nodes[node.__id].pvalue;"
+      out << "}"
+      tree_node_rule_sets.each do |rule_set|
+        each_tree_field(rule_set) do |rtype, field_name, child_type, slot|
+          out << ""
+          out << "#{child_type} #{p}#{rtype}_#{field_name}(#{rtype} node)"
+          out << "{"
+          out << "    #{child_type} result;"
+          out << "    result.__context = node.__context;"
+          out << "    if (node.__id == 0u)"
+          out << "    {"
+          out << "        result.__id = 0u;"
+          out << "        return result;"
+          out << "    }"
+          out << "    result.__id = node.__context->#{p}tree_children[node.__context->#{p}tree_nodes[node.__id].child_offset + #{slot}u];"
+          out << "    return result;"
+          out << "}"
+        end
+      end
+      out.join("\n")
+    end
+
+    # Generate the C++ tree node handle class declarations for the header.
+    # Only valid() and node_id() are defined inline; every other method
+    # dereferences the context, which is still an incomplete type here, so
+    # those are declared and defined out of line once the context is
+    # complete.
+    def cpp_tree_handle_types_header
+      p = @grammar.prefix
+      out = []
+      out << "/** Tree node handle types. @{ */"
+      tree_handle_types.each {|t| out << "struct #{t};"}
+      out << ""
+      tt = h_type("Token")
+      out << "struct #{tt}"
+      out << "{"
+      out << "    #{p}context_t * __context;"
+      out << "    #{p}node_id_t __id;"
+      out << "    bool valid() const { return __id != 0u; }"
+      out << "    explicit operator bool() const { return __id != 0u; }"
+      out << "    #{p}node_id_t node_id() const { return __id; }"
+      out << "    #{p}node_data_t * data() const;"
+      out << "    #{p}position_t position() const;"
+      out << "    #{p}position_t end_position() const;"
+      out << "    uint16_t n_fields() const;"
+      out << "    #{p}token_t token() const;"
+      out << "    #{p}value_t pvalue() const;"
+      out << "};"
+      out << ""
+      tree_node_rule_sets.each do |rule_set|
+        rtype = h_type(rule_set.name)
+        out << "struct #{rtype}"
+        out << "{"
+        out << "    #{p}context_t * __context;"
+        out << "    #{p}node_id_t __id;"
+        out << "    bool valid() const { return __id != 0u; }"
+        out << "    explicit operator bool() const { return __id != 0u; }"
+        out << "    #{p}node_id_t node_id() const { return __id; }"
+        out << "    #{p}node_data_t * data() const;"
+        out << "    #{p}position_t position() const;"
+        out << "    #{p}position_t end_position() const;"
+        out << "    uint16_t n_fields() const;"
+        each_tree_field(rule_set) do |rt, field_name, child_type, slot|
+          out << "    #{child_type} #{field_name}() const;"
+        end
+        out << "};"
+        out << ""
+      end
+      out.join("\n")
+    end
+
+    # Generate the out-of-line C++ handle method definitions plus the C-style
+    # accessors. Emitted after the context structure definition.
+    def cpp_tree_types_header
+      p = @grammar.prefix
+      out = []
+      # Common node methods, now that the context type is complete.
+      ([h_type("Token")] + tree_node_rule_sets.map {|rs| h_type(rs.name)}).each do |ht|
+        out << "inline #{p}node_data_t * #{ht}::data() const { return &__context->#{p}tree_nodes[__id]; }"
+        out << "inline #{p}position_t #{ht}::position() const { return __context->#{p}tree_nodes[__id].position; }"
+        out << "inline #{p}position_t #{ht}::end_position() const { return __context->#{p}tree_nodes[__id].end_position; }"
+        out << "inline uint16_t #{ht}::n_fields() const { return __id ? __context->#{p}tree_nodes[__id].n_fields : (uint16_t)0u; }"
+      end
+      tt = h_type("Token")
+      out << "inline #{p}token_t #{tt}::token() const { return __context->#{p}tree_nodes[__id].token; }"
+      out << "inline #{p}value_t #{tt}::pvalue() const { return __context->#{p}tree_nodes[__id].pvalue; }"
+      out << ""
+      # Out-of-line navigation method bodies (all handle types now complete).
+      tree_node_rule_sets.each do |rule_set|
+        rtype = h_type(rule_set.name)
+        each_tree_field(rule_set) do |rt, field_name, child_type, slot|
+          out << "inline #{child_type} #{rtype}::#{field_name}() const"
+          out << "{"
+          out << "    if (__id == 0u)"
+          out << "    {"
+          out << "        return #{child_type}{__context, 0u};"
+          out << "    }"
+          out << "    return #{child_type}{__context, __context->#{p}tree_children[__context->#{p}tree_nodes[__id].child_offset + #{slot}u]};"
+          out << "}"
+        end
+      end
+      out << ""
+      out << "/*"
+      out << " * C-style function and macro accessors, provided in addition to the handle"
+      out << " * methods above so that C-style code and the tree walk macros also work."
+      out << " */"
+      out << c_common_accessors_header
+      out << "/** @} */"
+      out.join("\n")
+    end
+
+    # Rust keywords that must be escaped as raw identifiers when used as a
+    # generated identifier (e.g. a field alias named `type`).
+    RUST_KEYWORDS = %w[
+      as break const continue dyn else enum extern false fn for if impl in let
+      loop match mod move mut pub ref return static struct trait true type
+      unsafe use where while async await abstract become box do final macro
+      override priv typeof unsized virtual yield try gen
+    ]
+
+    # Escape a name as a Rust raw identifier if it is a reserved keyword.
+    #
+    # @param name [String]
+    #   Identifier name.
+    #
+    # @return [String]
+    #   Name, escaped as a raw identifier if necessary.
+    def rust_ident(name)
+      RUST_KEYWORDS.include?(name) ? "r##{name}" : name
+    end
+
+    # Map a ptype type string to a valid Rust type.
+    #
+    # The default ptype is a C "void *"; for Rust with no declared ptype we use
+    # the unit type instead.
+    #
+    # @param typestring [String]
+    #   ptype type string.
+    #
+    # @return [String]
+    #   Rust type string.
+    def rust_ptype(typestring)
+      typestring == "void *" ? "()" : typestring
+    end
+
+    # Get the lex function to use.
+    #
+    # @return [String]
+    #   Lex function to use.
+    def lex_fn
+      @grammar.lex_fn || "#{@grammar.prefix}lex"
+    end
+
+    # Get the parser value type for the start rule.
+    #
+    # @return [Array<String>]
+    #   Start rule parser value type name and type string.
+    def start_rule_type(start_rule_index = 0)
+      start_rule = @grammar.rules.find do |rule|
+        rule.name == @grammar.start_rules[start_rule_index]
+      end
+      [start_rule.ptypename, @grammar.ptypes[start_rule.ptypename]]
+    end
+
+    # Get an unsigned integer type that can hold the given maximum value.
+    #
+    # @param max [Integer]
+    #   Maximum value to store.
+    #
+    # @return [String]
+    #   Type.
+    def get_type_for(max)
+      if max <= 0xFF
+        case @language
+        when "c"
+          "uint8_t"
+        when "d"
+          "ubyte"
+        when "rust"
+          "u8"
+        end
+      elsif max <= 0xFFFF
+        case @language
+        when "c"
+          "uint16_t"
+        when "d"
+          "ushort"
+        when "rust"
+          "u16"
+        end
+      else
+        case @language
+        when "c"
+          "uint32_t"
+        when "rust"
+          "u32"
+        else
+          "uint"
+        end
+      end
+    end
+
+  end
+
+end

lib/propane/grammar.rb

@@ -0,0 +1,414 @@
+class Propane
+
+  class Grammar
+
+    # Reserve identifiers beginning with a double-underscore for internal use.
+    IDENTIFIER_REGEX = /(?:[a-zA-Z]|_[a-zA-Z0-9])[a-zA-Z_0-9]*/
+
+    attr_reader :context_user_fields
+    attr_reader :lex_fn
+    attr_reader :tree
+    attr_reader :tree_prefix
+    attr_reader :tree_suffix
+    attr_reader :free_token_node
+    attr_reader :modulename
+    attr_reader :patterns
+    attr_accessor :rules
+    attr_reader :start_rules
+    attr_reader :tokens
+    attr_reader :code_blocks
+    attr_reader :ptypes
+    attr_reader :prefix
+    attr_reader :on_token_node
+    attr_reader :token_user_fields
+
+    def initialize(input, filename)
+      @filename = filename
+      @patterns = []
+      @start_rules = []
+      @tokens = []
+      @rules = []
+      @code_blocks = {}
+      @line_number = 1
+      @next_line_number = @line_number
+      @modeline = nil
+      @input = input.gsub("\r\n", "\n")
+      @ptypes = {"default" => "void *"}
+      @prefix = "p_"
+      @tree = false
+      @tree_prefix = ""
+      @tree_suffix = ""
+      @free_token_node = ""
+      @context_user_fields = nil
+      @on_token_node = ""
+      @token_user_fields = nil
+      parse_grammar!
+      @start_rules << "Start" if @start_rules.empty?
+    end
+
+    def ptype
+      @ptypes["default"]
+    end
+
+    def invalid_token_id
+      @tokens.size
+    end
+
+    def terminate_token_id
+      @tokens.size + 1
+    end
+
+    def parser_user_code_used?
+      @rules.any? {|r| r.code}
+    end
+
+    private
+
+    def parse_grammar!
+      while @input.size > 0
+        parse_statement!
+      end
+    end
+
+    def parse_statement!
+      if parse_white_space!
+      elsif parse_comment_line!
+      elsif @modeline.nil? && parse_mode_label!
+      elsif parse_context_user_fields_statement!
+      elsif parse_lex_fn!
+      elsif parse_tree_statement!
+      elsif parse_tree_prefix_statement!
+      elsif parse_tree_suffix_statement!
+      elsif parse_free_token_node_statement!
+      elsif parse_module_statement!
+      elsif parse_on_token_node_statement!
+      elsif parse_token_user_fields_statement!
+      elsif parse_ptype_statement!
+      elsif parse_pattern_statement!
+      elsif parse_start_statement!
+      elsif parse_token_statement!
+      elsif parse_tokenid_statement!
+      elsif parse_drop_statement!
+      elsif parse_rule_statement!
+      elsif parse_code_block_statement!
+      elsif parse_prefix_statement!
+      elsif parse_noline_statement!
+      else
+        if @input.size > 25
+          @input = @input.slice(0..20) + "..."
+        end
+        raise Error.new("Unexpected grammar input at line #{@line_number}: #{@input.chomp}")
+      end
+    end
+
+    def parse_mode_label!
+      if md = consume!(/(#{IDENTIFIER_REGEX}(?:\s*,\s*#{IDENTIFIER_REGEX})*)\s*:/)
+        @modeline = md[1]
+      end
+    end
+
+    def parse_white_space!
+      consume!(/\s+/)
+    end
+
+    def parse_comment_line!
+      consume!(/#.*\n/)
+    end
+
+    def parse_context_user_fields_statement!
+      if md = consume!(/context_user_fields\b\s*/)
+        unless code = parse_code_block!
+          raise Error.new("Line #{@line_number}: expected code block")
+        end
+        @context_user_fields ||= ""
+        @context_user_fields += code
+      end
+    end
+
+    def parse_lex_fn!
+      if md = consume!(/lex_fn\b\s*(\w+)\s*;/)
+        @lex_fn = md[1]
+      end
+    end
+
+    def parse_tree_statement!
+      if consume!(/tree\s*;/)
+        @tree = true
+      end
+    end
+
+    def parse_tree_prefix_statement!
+      if md = consume!(/tree_prefix\s+(\w+)\s*;/)
+        @tree_prefix = md[1]
+      end
+    end
+
+    def parse_tree_suffix_statement!
+      if md = consume!(/tree_suffix\s+(\w+)\s*;/)
+        @tree_suffix = md[1]
+      end
+    end
+
+    def parse_module_statement!
+      if consume!(/module\s+/)
+        md = consume!(/([\w.]+)\s*/, "expected module name")
+        @modulename = md[1]
+        consume!(/;/, "expected `;'")
+        @modeline = nil
+        true
+      end
+    end
+
+    def parse_on_token_node_statement!
+      if md = consume!(/on_token_node\b\s*/)
+        unless code = parse_code_block!
+          raise Error.new("Line #{@line_number}: expected code block")
+        end
+        @on_token_node += code
+      end
+    end
+
+    def parse_token_user_fields_statement!
+      if md = consume!(/token_user_fields\b\s*/)
+        unless code = parse_code_block!
+          raise Error.new("Line #{@line_number}: expected code block")
+        end
+        @token_user_fields ||= ""
+        @token_user_fields += code
+      end
+    end
+
+    def parse_free_token_node_statement!
+      if md = consume!(/free_token_node\b\s*/)
+        unless code = parse_code_block!
+          raise Error.new("Line #{@line_number}: expected code block")
+        end
+        @free_token_node += code
+      end
+    end
+
+    def parse_ptype_statement!
+      if consume!(/ptype\s+/)
+        name = "default"
+        if md = consume!(/(#{IDENTIFIER_REGEX})\s*=\s*/)
+          if @tree
+            raise Error.new("Multiple ptypes are unsupported in tree mode")
+          end
+          name = md[1]
+        end
+        md = consume!(/([^;]+);/, "expected parser result type expression")
+        @ptypes[name] = md[1].strip
+      end
+    end
+
+    def parse_token_statement!
+      if consume!(/token\s+/)
+        md = consume!(/(#{IDENTIFIER_REGEX})\s*/, "expected token name")
+        name = md[1]
+        if md = consume!(/\((#{IDENTIFIER_REGEX})\)\s*/)
+          if @tree
+            raise Error.new("Multiple ptypes are unsupported in tree mode")
+          end
+          ptypename = md[1]
+        end
+        pattern = parse_pattern! || name
+        consume!(/\s+/)
+        unless code = parse_code_block!
+          consume!(/;/, "expected `;' or code block")
+        end
+        token = Token.new(name, ptypename, @line_number)
+        @tokens << token
+        pattern = Pattern.new(pattern: pattern, token: token, line_number: @line_number, code: code, modes: get_modes_from_modeline, ptypename: ptypename)
+        @patterns << pattern
+        @modeline = nil
+        true
+      end
+    end
+
+    def parse_tokenid_statement!
+      if md = consume!(/tokenid\s+/)
+        md = consume!(/(#{IDENTIFIER_REGEX})\s*/, "expected token name")
+        name = md[1]
+        if md = consume!(/\((#{IDENTIFIER_REGEX})\)\s*/)
+          if @tree
+            raise Error.new("Multiple ptypes are unsupported in tree mode")
+          end
+          ptypename = md[1]
+        end
+        consume!(/;/, "expected `;'");
+        token = Token.new(name, ptypename, @line_number)
+        @tokens << token
+        @modeline = nil
+        true
+      end
+    end
+
+    def parse_drop_statement!
+      if md = consume!(/drop\s+/)
+        pattern = parse_pattern!
+        unless pattern
+          raise Error.new("Line #{@line_number}: expected pattern to follow `drop'")
+        end
+        consume!(/\s+/)
+        unless code = parse_code_block!
+          consume!(/;/, "expected `;' or code block")
+        end
+        @patterns << Pattern.new(pattern: pattern, line_number: @line_number, code: code, modes: get_modes_from_modeline)
+        @modeline = nil
+        true
+      end
+    end
+
+    def parse_rule_statement!
+      if md = consume!(/(#{IDENTIFIER_REGEX})\s*(?:\((#{IDENTIFIER_REGEX})\))?\s*->\s*/)
+        rule_name, ptypename = *md[1, 2]
+        if @tree && ptypename
+          raise Error.new("Multiple ptypes are unsupported in tree mode")
+        end
+        md = consume!(/((?:#{IDENTIFIER_REGEX}\??(?::#{IDENTIFIER_REGEX})?\s*)*)\s*/, "expected rule component list")
+        components = md[1].strip.split(/\s+/)
+        unless code = parse_code_block!
+          consume!(/;/, "expected `;' or code block")
+        end
+        @rules << Rule.new(rule_name, components, code, ptypename, @line_number)
+        @modeline = nil
+        true
+      end
+    end
+
+    def parse_pattern_statement!
+      if pattern = parse_pattern!
+        consume!(/\s+/)
+        if md = consume!(/\((#{IDENTIFIER_REGEX})\)\s*/)
+          if @tree
+            raise Error.new("Multiple ptypes are unsupported in tree mode")
+          end
+          ptypename = md[1]
+        end
+        unless code = parse_code_block!
+          raise Error.new("Line #{@line_number}: expected code block to follow pattern")
+        end
+        @patterns << Pattern.new(pattern: pattern, line_number: @line_number, code: code, modes: get_modes_from_modeline, ptypename: ptypename)
+        @modeline = nil
+        true
+      end
+    end
+
+    def parse_start_statement!
+      if md = consume!(/start\s+([\w\s]*);/)
+        start_rules = md[1].split(/\s+/).map(&:strip)
+        start_rules.each do |start_rule|
+          @start_rules << start_rule unless @start_rules.include?(start_rule)
+        end
+      end
+    end
+
+    def parse_code_block_statement!
+      if md = consume!(/<<([a-z]*)(.*?)>>\n/m)
+        name, code = md[1..2]
+        code = code.chomp
+        unless @noline
+          if code.start_with?("\n")
+            code = %[#line #{@line_number + 1} "#{@filename}"#{code}\n#linereset\n]
+          else
+            code = %[#line #{@line_number} "#{@filename}"\n#{code}\n#linereset\n]
+          end
+        end
+        if @code_blocks[name]
+          @code_blocks[name] += code
+        else
+          @code_blocks[name] = code
+        end
+        @modeline = nil
+        true
+      end
+    end
+
+    def parse_prefix_statement!
+      if md = consume!(/prefix\s+(#{IDENTIFIER_REGEX})\s*;/)
+        @prefix = md[1]
+        true
+      end
+    end
+
+    def parse_noline_statement!
+      if md = consume!(/noline\s*;/)
+        @noline = true
+        true
+      end
+    end
+
+    def parse_pattern!
+      if md = consume!(%r{/})
+        pattern = ""
+        while !consume!(%r{/})
+          if consume!(%r{\\})
+            pattern += "\\"
+            if md = consume!(%r{(.)})
+              pattern += md[1]
+            else
+              raise Error.new("Line #{@line_number}: unterminated escape sequence")
+            end
+          elsif md = consume!(%r{(.)})
+            pattern += md[1]
+          elsif @input == "" || @input.start_with?("\n")
+            raise Error.new("Line #{@line_number}: Unterminated pattern; expected `/`")
+          end
+        end
+        pattern
+      end
+    end
+
+    def parse_code_block!
+      if md = consume!(/<<(.*?)>>\n/m)
+        code = md[1].chomp
+        unless @noline
+          if code.start_with?("\n")
+            code = %[#line #{@line_number + 1} "#{@filename}"#{code}\n#linereset\n]
+          else
+            code = %[#line #{@line_number} "#{@filename}"\n#{code}\n#linereset\n]
+          end
+        end
+        code
+      end
+    end
+
+    # Check if the input string matches the given regex.
+    #
+    # If so, remove the match from the input string, and update the line
+    # number. If the regex is not matched and an error message is provided,
+    # the error is raised.
+    #
+    # @param regex [Regexp]
+    #   Regex to attempt to match.
+    # @param error_message [String, nil]
+    #   Error message to display if the regex is not matched. If nil and the
+    #   regex is not matched, an error is not raised.
+    #
+    # @return [MatchData, nil]
+    #   MatchData for the given regex if it was matched and removed from the
+    #   input.
+    def consume!(regex, error_message = nil)
+      @line_number = @next_line_number
+      if md = @input.match(/\A#{regex}/)
+        @input.slice!(0, md[0].size)
+        @next_line_number += md[0].count("\n")
+        md
+      elsif error_message
+        raise Error.new("Line #{@line_number}: Error: #{error_message}")
+      else
+        false
+      end
+    end
+
+    def get_modes_from_modeline
+      if @modeline
+        Set[*@modeline.split(",").map(&:strip)]
+      else
+        Set.new
+      end
+    end
+
+  end
+
+end

lib/propane/lexer.rb

@@ -0,0 +1,74 @@
+class Propane
+  class Lexer
+
+    attr_reader :state_table
+    attr_reader :transition_table
+    attr_reader :mode_table
+
+    def initialize(grammar)
+      @grammar = grammar
+      build_tables!
+    end
+
+    # Get ID for a mode.
+    #
+    # @param mode_name [String]
+    #   Mode name.
+    #
+    # @return [Integer, nil]
+    #   Mode ID.
+    def mode_id(mode_name)
+      if mode_info = @modes[mode_name]
+        mode_info[:id]
+      end
+    end
+
+    private
+
+    def build_tables!
+      modenames = @grammar.patterns.reduce(Set.new) do |result, pattern|
+        result + pattern.modes
+      end
+      @modes = modenames.reduce({}) do |result, modename|
+        result[modename] = @grammar.patterns.select do |pattern|
+          pattern.modes.include?(modename)
+        end
+        result
+      end.transform_values do |patterns|
+        {dfa: DFA.new(patterns)}
+      end
+      @modes.each_with_index do |(mode_name, mode_info), index|
+        mode_info[:id] = index
+      end
+      @state_table = []
+      @transition_table = []
+      @mode_table = []
+      @modes.each do |mode_name, mode_info|
+        state_table_offset = @state_table.size
+        @mode_table << {
+          state_table_offset: state_table_offset,
+        }
+        states = mode_info[:dfa].enumerate
+        states.each do |state, id|
+          token = state.accepts && state.accepts.token && state.accepts.token.id
+          code_id = state.accepts && state.accepts.code_id && state.accepts.code_id
+          @state_table << {
+            transition_table_index: @transition_table.size,
+            n_transitions: state.transitions.size,
+            accepts: !!state.accepts,
+            token: token,
+            code_id: code_id,
+          }
+          state.transitions.each do |transition|
+            @transition_table << {
+              first: transition.code_point_range.first,
+              last: transition.code_point_range.last,
+              destination: states[transition.destination] + state_table_offset,
+            }
+          end
+        end
+      end
+    end
+
+  end
+end

lib/propane/lexer/dfa.rb

@@ -0,0 +1,89 @@
+class Propane
+  class Lexer
+
+    class DFA < FA
+
+      def initialize(patterns)
+        super()
+        start_nfa = Regex::NFA.new
+        patterns.each do |pattern|
+          start_nfa.start_state.add_transition(nil, pattern.nfa.start_state)
+        end
+        @nfa_state_sets = {}
+        @states = []
+        @to_process = Set.new
+        nil_transition_states = start_nfa.start_state.nil_transition_states
+        register_nfa_state_set(nil_transition_states)
+        while @to_process.size > 0
+          state_set = @to_process.first
+          @to_process.delete(state_set)
+          process_nfa_state_set(state_set)
+        end
+        @start_state = @states[0]
+      end
+
+      private
+
+      def register_nfa_state_set(nfa_state_set)
+        unless @nfa_state_sets.include?(nfa_state_set)
+          state_id = @states.size
+          @nfa_state_sets[nfa_state_set] = state_id
+          @states << State.new
+          @to_process << nfa_state_set
+        end
+      end
+
+      def process_nfa_state_set(nfa_state_set)
+        state_id = @nfa_state_sets[nfa_state_set]
+        state = @states[state_id]
+        if state_id > 0
+          nfa_state_set.each do |nfa_state|
+            if nfa_state.accepts
+              if state.accepts
+                if nfa_state.accepts.line_number < state.accepts.line_number
+                  state.accepts = nfa_state.accepts
+                end
+              else
+                state.accepts = nfa_state.accepts
+              end
+            end
+          end
+        end
+        transitions = transitions_for(nfa_state_set)
+        while transitions.size > 0
+          subrange = CodePointRange.first_subrange(transitions.map(&:code_point_range))
+          dest_nfa_states = transitions.reduce(Set.new) do |result, transition|
+            if transition.code_point_range.include?(subrange)
+              result << transition.destination
+            end
+            result
+          end
+          dest_nfa_states = dest_nfa_states.reduce(Set.new) do |result, dest_nfa_state|
+            result + dest_nfa_state.nil_transition_states
+          end
+          register_nfa_state_set(dest_nfa_states)
+          dest_state = @states[@nfa_state_sets[dest_nfa_states]]
+          state.add_transition(subrange, dest_state)
+          transitions.delete_if do |transition|
+            transition.code_point_range.last <= subrange.last
+          end
+          transitions.map! do |transition|
+            if transition.code_point_range.first <= subrange.last
+              Regex::NFA::State::Transition.new(CodePointRange.new(subrange.last + 1, transition.code_point_range.last), transition.destination)
+            else
+              transition
+            end
+          end
+        end
+      end
+
+      def transitions_for(nfa_state_set)
+        nfa_state_set.reduce([]) do |result, state|
+          result + state.cp_transitions
+        end
+      end
+
+    end
+
+  end
+end

lib/propane/parser.rb

@@ -0,0 +1,428 @@
+class Propane
+
+  class Parser
+
+    attr_reader :state_table
+    attr_reader :shift_table
+    attr_reader :reduce_table
+    attr_reader :rule_sets
+
+    def initialize(grammar, rule_sets, log, options)
+      @grammar = grammar
+      @rule_sets = rule_sets
+      @log = log
+      @item_sets = []
+      @item_sets_set = {}
+      @warnings = Set.new
+      @errors = Set.new
+      @options = options
+      start_items = grammar.rules[0...grammar.start_rules.length].map do |start_rule|
+        Item.new(start_rule, 0)
+      end
+      start_item_sets = start_items.map {|item| ItemSet.new([item])}
+      eval_item_sets = Set[*start_item_sets]
+
+      while eval_item_sets.size > 0
+        item_set =
+          if start_item_sets.size > 0
+            # Ensure we evaluate start_item_sets first in order
+            start_item_sets.slice!(0)
+          else
+            eval_item_sets.first
+          end
+        eval_item_sets.delete(item_set)
+        unless @item_sets_set.include?(item_set)
+          item_set.id = @item_sets.size
+          @item_sets << item_set
+          @item_sets_set[item_set] = item_set
+          item_set.next_symbols.each do |next_symbol|
+            unless next_symbol.name == "$EOF"
+              next_item_set = item_set.build_next_item_set(next_symbol)
+              eval_item_sets << next_item_set
+            end
+          end
+        end
+      end
+
+      @item_sets.each do |item_set|
+        process_item_set(item_set)
+      end
+
+      build_reduce_actions!
+      build_tables!
+      check_unused_rules!
+      write_log!
+      errormessage = ""
+      if @errors.size > 0
+        errormessage += @errors.join("\n")
+      end
+      if @warnings.size > 0 && @options[:warnings_as_errors]
+        if errormessage != ""
+          errormessage += "\n"
+        end
+        errormessage += "Fatal errors (-w):\n" + @warnings.join("\n")
+      end
+      if errormessage != ""
+        raise Error.new(errormessage)
+      end
+    end
+
+    private
+
+    # Record a warning for each user-defined rule that cannot be reached from
+    # any start rule.
+    #
+    # @return [void]
+    def check_unused_rules!
+      reachable = Set.new
+      pending = @grammar.rules[0...@grammar.start_rules.length].map(&:rule_set)
+      while rule_set = pending.pop
+        next if reachable.include?(rule_set)
+        reachable << rule_set
+        rule_set.rules.each do |rule|
+          rule.components.each do |component|
+            if component.is_a?(RuleSet) && !reachable.include?(component)
+              pending << component
+            end
+          end
+        end
+      end
+      @rule_sets.each do |name, rule_set|
+        # Skip rule sets generated internally (real start rules and optional
+        # component rules).
+        next if name.start_with?("$") || rule_set.optional?
+        unless reachable.include?(rule_set)
+          @warnings << "Unused rule #{name} (defined on line #{rule_set.rules.first.line_number})"
+        end
+      end
+    end
+
+    def build_tables!
+      @state_table = []
+      @shift_table = []
+      @reduce_table = []
+      @item_sets.each do |item_set|
+        unless item_set.reduce_rules.empty?
+          item_set.shift_entries.each do |shift_entry|
+            token = shift_entry[:symbol]
+            if item_set.reduce_actions
+              if rule = item_set.reduce_actions[token]
+                @warnings << "Shift/Reduce conflict (state #{item_set.id}) between token #{token.name} and rule #{rule.name} (defined on line #{rule.line_number})"
+              end
+            end
+          end
+        end
+        reduce_entries =
+          if rule = item_set.reduce_rule
+            [{token_id: @grammar.invalid_token_id, rule_id: rule.id, rule: rule,
+              rule_set_id: rule.rule_set.id, n_states: rule.components.size,
+              propagate_optional_target: rule.optional? && rule.components.size == 1}]
+          elsif reduce_actions = item_set.reduce_actions
+            reduce_actions.map do |token, rule|
+              {token: token, token_id: token.id, rule_id: rule.id, rule: rule,
+               rule_set_id: rule.rule_set.id, n_states: rule.components.size,
+               propagate_optional_target: rule.optional? && rule.components.size == 1}
+            end
+          else
+            []
+          end
+        @state_table << {
+          shift_index: @shift_table.size,
+          n_shifts: item_set.shift_entries.size,
+          reduce_index: @reduce_table.size,
+          n_reduces: reduce_entries.size,
+        }
+        @shift_table += item_set.shift_entries
+        @reduce_table += reduce_entries
+      end
+    end
+
+    def process_item_set(item_set)
+      item_set.next_symbols.each do |next_symbol|
+        unless next_symbol.name == "$EOF"
+          next_item_set = @item_sets_set[item_set.build_next_item_set(next_symbol)]
+          item_set.next_item_set[next_symbol] = next_item_set
+          next_item_set.in_sets << item_set
+        end
+      end
+    end
+
+    # Build the reduce actions for each ItemSet.
+    #
+    # @return [void]
+    def build_reduce_actions!
+      @item_sets.each do |item_set|
+        build_shift_entries(item_set)
+        build_reduce_actions_for_item_set(item_set)
+      end
+      item_sets_to_process = @item_sets.select do |item_set|
+        # We need lookahead reduce actions if:
+        # 1) There is more than one possible rule to reduce. In this case the
+        #    lookahead token can help choose which rule to reduce.
+        # 2) There is at least one shift action and one reduce action for
+        #    this item set. In this case the lookahead reduce actions are
+        #    needed to test for a Shift/Reduce conflict.
+        item_set.reduce_rules.size > 1 ||
+          (item_set.reduce_rules.size > 0 && item_set.shift_entries.size > 0)
+      end
+      if RbConfig::CONFIG["host_os"] =~ /linux/
+        item_sets_by_id = {}
+        item_sets_to_process.each do |item_set|
+          item_sets_by_id[item_set.object_id] = item_set
+        end
+        tokens_by_id = {}
+        @grammar.tokens.each do |token|
+          tokens_by_id[token.object_id] = token
+        end
+        rules_by_id = {}
+        @grammar.rules.each do |rule|
+          rules_by_id[rule.object_id] = rule
+        end
+        n_threads = Util.determine_n_threads
+        semaphore = Mutex.new
+        queue = Queue.new
+        threads = {}
+        n_threads.times do
+          piper, pipew = IO.pipe
+          thread = Thread.new do
+            loop do
+              item_set = nil
+              semaphore.synchronize do
+                item_set = item_sets_to_process.slice!(0)
+              end
+              break if item_set.nil?
+              fork do
+                piper.close
+                build_lookahead_reduce_actions_for_item_set(item_set, pipew)
+              end
+            end
+            queue.push(Thread.current)
+          end
+          threads[thread] = [piper, pipew]
+        end
+        until threads.empty?
+          thread = queue.pop
+          piper, pipew = threads[thread]
+          pipew.close
+          thread_txt = piper.read
+          thread_txt.each_line do |line|
+            if line.start_with?("RA,")
+              parts = line.split(",")
+              item_set_id, token_id, rule_id = parts[1..3].map(&:to_i)
+              item_set = item_sets_by_id[item_set_id]
+              unless item_set
+                raise "Internal error: could not find item set from thread"
+              end
+              token = tokens_by_id[token_id]
+              unless item_set
+                raise "Internal error: could not find token from thread"
+              end
+              rule = rules_by_id[rule_id]
+              unless item_set
+                raise "Internal error: could not find rule from thread"
+              end
+              item_set.reduce_actions ||= {}
+              item_set.reduce_actions[token] = rule
+            elsif line.start_with?("Error: ")
+              @errors << line.chomp
+            else
+              raise "Internal error: unhandled thread line #{line}"
+            end
+          end
+          thread.join
+          threads.delete(thread)
+        end
+      else
+        # Fall back to single threaded algorithm.
+        item_sets_to_process.each do |item_set|
+          item_set.reduce_actions = build_lookahead_reduce_actions_for_item_set(item_set)
+        end
+      end
+    end
+
+    # Build the shift entries for a single item set.
+    #
+    # @return [void]
+    def build_shift_entries(item_set)
+      item_set.shift_entries = item_set.next_symbols.map do |next_symbol|
+        state_id =
+          if next_symbol.name == "$EOF"
+            0
+          else
+            item_set.next_item_set[next_symbol].id
+          end
+        {
+          symbol: next_symbol,
+          state_id: state_id,
+        }
+      end
+    end
+
+    # Build the reduce actions for a single item set (parser state).
+    #
+    # @param item_set [ItemSet]
+    #   ItemSet (parser state)
+    #
+    # @return [void]
+    def build_reduce_actions_for_item_set(item_set)
+      # To build the reduce actions, we start by looking at any
+      # "complete" items, i.e., items where the parse position is at the
+      # end of a rule. These are the only rules that are candidates for
+      # reduction in the current ItemSet.
+      item_set.reduce_rules = Set.new(item_set.items.select(&:complete?).map(&:rule))
+
+      if item_set.reduce_rules.size == 1
+        item_set.reduce_rule = item_set.reduce_rules.first
+      end
+    end
+
+    # Build the reduce actions for a single item set (parser state).
+    #
+    # @param item_set [ItemSet]
+    #   ItemSet (parser state)
+    # @param fh [File]
+    #   Output file handle for multiprocessing mode.
+    #
+    # @return [Hash]
+    #   Mapping of lookahead Tokens to the Rules to reduce.
+    def build_lookahead_reduce_actions_for_item_set(item_set, fh = nil)
+      # We will be looking for all possible tokens that can follow instances of
+      # these rules. Rather than looking through the entire grammar for the
+      # possible following tokens, we will only look in the item sets leading
+      # up to this one. This restriction gives us a more precise lookahead set,
+      # and allows us to parse LALR grammars.
+      item_sets = Set[item_set] + item_set.leading_item_sets
+      item_set.reduce_rules.reduce({}) do |reduce_actions, reduce_rule|
+        lookahead_tokens_for_rule = build_lookahead_tokens_to_reduce(reduce_rule, item_sets)
+        lookahead_tokens_for_rule.each do |lookahead_token|
+          if existing_reduce_rule = reduce_actions[lookahead_token]
+            error = "Error: reduce/reduce conflict (state #{item_set.id}) between rule #{existing_reduce_rule.name}##{existing_reduce_rule.id} (defined on line #{existing_reduce_rule.line_number}) and rule #{reduce_rule.name}##{reduce_rule.id} (defined on line #{reduce_rule.line_number}) for lookahead token #{lookahead_token}"
+            @errors << error
+            fh.puts(error) if fh
+          end
+          reduce_actions[lookahead_token] = reduce_rule
+          fh.puts "RA,#{item_set.object_id},#{lookahead_token.object_id},#{reduce_rule.object_id}" if fh
+        end
+        reduce_actions
+      end
+    end
+
+    # Build the set of lookahead Tokens that should cause the given Rule to be
+    # reduced in the given context of ItemSets.
+    #
+    # @param rule [Rule]
+    #   Rule to reduce.
+    # @param item_sets [Set<ItemSet>]
+    #   ItemSets to consider for the context in which to reduce this Rule.
+    #
+    # @return [Set<Token>]
+    #   Possible lookahead Tokens for the given Rule within the context of the
+    #   given ItemSets.
+    def build_lookahead_tokens_to_reduce(rule, item_sets)
+      # We need to look for possible following tokens for this reduce rule. We
+      # do this by looking for tokens that follow the reduce rule, or the
+      # start token set for any other rule that follows the reduce rule.
+      # While doing this, the following situations could arise:
+      # 1. We may come across a following rule that could be empty. In this
+      #    case, in addition to the start token set for that rule, we must also
+      #    continue to the next following symbol after the potentially empty
+      #    rule and continue the search for potential following tokens.
+      # 2. We may reach the end of a rule that was not one of the original
+      #    reduce rules. In this case, we must also search for all potential
+      #    following tokens for this rule as well.
+      lookahead_tokens = Set.new
+      rule_sets_to_check_after = [rule.rule_set]
+      checked_rule_sets = Set.new
+      while !rule_sets_to_check_after.empty?
+        rule_set = rule_sets_to_check_after.slice!(0)
+        checked_rule_sets << rule_set
+        # For each RuleSet we're checking, we're going to look through all
+        # items in the item sets of interest and gather all possible following
+        # tokens to form the lookahead token set.
+        item_sets.each do |item_set|
+          item_set.items.each do |item|
+            if item.next_symbol == rule_set
+              (1..).each do |offset|
+                case symbol = item.next_symbol(offset)
+                when nil
+                  rule_set = item.rule.rule_set
+                  unless checked_rule_sets.include?(rule_set)
+                    rule_sets_to_check_after << rule_set
+                    checked_rule_sets << rule_set
+                  end
+                  break
+                when Token
+                  lookahead_tokens << symbol
+                  break
+                when RuleSet
+                  lookahead_tokens += symbol.start_token_set
+                  unless symbol.could_be_empty?
+                    break
+                  end
+                end
+              end
+            end
+          end
+        end
+      end
+      lookahead_tokens
+    end
+
+    def write_log!
+      @log.puts Util.banner("Parser Rules")
+      @grammar.rules.each do |rule|
+        @log.puts
+        @log.puts "Rule #{rule.id}:"
+        @log.puts "  #{rule}"
+      end
+
+      @log.puts
+      @log.puts Util.banner("Parser Tokens")
+      @log.puts
+      @grammar.tokens.each do |token|
+        @log.puts "Token #{token.id}: #{token.name}"
+      end
+
+      @log.puts
+      @log.puts Util.banner("Parser Rule Sets")
+      @rule_sets.each do |rule_set_name, rule_set|
+        @log.puts
+        @log.puts "Rule Set #{rule_set.id}: #{rule_set_name}"
+        @log.puts "  Start token set: #{rule_set.start_token_set.map(&:name).join(", ")}"
+      end
+
+      @log.puts
+      @log.puts Util.banner("Parser States")
+      @item_sets.each do |item_set|
+        @log.puts
+        @log.puts "State #{item_set.id}:"
+        @log.puts item_set.to_s.gsub(/^/, "  ")
+        incoming_ids = item_set.in_sets.map(&:id)
+        @log.puts
+        @log.puts "  Incoming states: #{incoming_ids.join(", ")}"
+        @log.puts "  Outgoing states:"
+        item_set.next_item_set.each do |next_symbol, next_item_set|
+          @log.puts "    #{next_symbol.name} => #{next_item_set.id}"
+        end
+        @log.puts
+        @log.puts "  Reduce actions:"
+        if item_set.reduce_rule
+          @log.puts "    * => rule #{item_set.reduce_rule.id}, rule set #{@rule_sets[item_set.reduce_rule.name].id} (#{item_set.reduce_rule.name})"
+        elsif item_set.reduce_actions
+          item_set.reduce_actions.each do |token, rule|
+            @log.puts "    lookahead #{token.name} => #{rule.name} (#{rule.id}), rule set ##{rule.rule_set.id}"
+          end
+        end
+      end
+      if @warnings.size > 0
+        @log.puts
+        @log.puts "Warnings:"
+        @warnings.each do |warning|
+          @log.puts "  #{warning}"
+        end
+      end
+    end
+
+  end
+
+end

lib/propane/parser/item.rb

@@ -0,0 +1,147 @@
+class Propane
+  class Parser
+
+    # Represent a parser "item", which is a position in a Rule that the parser
+    # could potentially be at.
+    class Item
+
+      # @return [Rule]
+      #   The Rule corresponding to this Item.
+      attr_reader :rule
+
+      # @return [Integer]
+      #   The parse position in this item.
+      attr_reader :position
+
+      # Construct an Item.
+      #
+      # @param rule [Rule]
+      #   The Rule corresponding to this Item.
+      # @param position [Integer]
+      #   The parse position in this Item.
+      def initialize(rule, position)
+        @rule = rule
+        @position = position
+        @_hash = [@rule, @position].hash
+      end
+
+      # Hash function.
+      #
+      # @return [Integer]
+      #   Hash code.
+      def hash
+        @_hash
+      end
+
+      # Compare Item objects.
+      #
+      # @param other [Item]
+      #   Item to compare to.
+      #
+      # @return [Boolean]
+      #   Whether the Items are equal.
+      def ==(other)
+        @rule == other.rule && @position == other.position
+      end
+
+      # Compare Item objects.
+      #
+      # @param other [Item]
+      #   Item to compare to.
+      #
+      # @return [Boolean]
+      #   Whether the Items are equal.
+      def eql?(other)
+        self == other
+      end
+
+      # Return the set of Items obtained by "closing" the current item.
+      #
+      # If the next symbol for the current item is another Rule name, then
+      # this method will return all Items for that Rule with a position of 0.
+      # Otherwise, an empty Array is returned.
+      #
+      # @return [Array<Item>]
+      #   Items obtained by "closing" the current item.
+      def closed_items
+        if @rule.components[@position].is_a?(RuleSet)
+          @rule.components[@position].rules.map do |rule|
+            Item.new(rule, 0)
+          end
+        else
+          []
+        end
+      end
+
+      # Return whether the item is "complete", meaning that the parse position
+      # marker is at the end of the rule.
+      #
+      # @return [Boolean]
+      #   Whether the item is "complete".
+      def complete?
+        @position == @rule.components.size
+      end
+
+      # Get the next symbol for the Item.
+      #
+      # That is, the symbol which is after the parse position marker in the
+      # current Item.
+      #
+      # @param offset [Integer]
+      #   Offset from current parse position to examine.
+      #
+      # @return [Token, RuleSet, nil]
+      #   Next symbol for the Item.
+      def next_symbol(offset = 0)
+        @rule.components[@position + offset]
+      end
+
+      # Get the previous symbol for the Item.
+      #
+      # That is, the symbol which precedes the parse position marker in the
+      # current Item.
+      #
+      # @return [Token, RuleSet, nil]
+      #   Previous symbol for this Item.
+      def previous_symbol
+        if @position > 0
+          @rule.components[@position - 1]
+        end
+      end
+
+      # Get whether this Item's next symbol is the given symbol.
+      #
+      # @param symbol [Token, RuleSet]
+      #   Symbol to query.
+      #
+      # @return [Boolean]
+      #   Whether this Item's next symbol is the given symbol.
+      def next_symbol?(symbol)
+        next_symbol == symbol
+      end
+
+      # Get the next item for this Item.
+      #
+      # That is, the Item formed by moving the parse position marker one place
+      # forward from its position in this Item.
+      #
+      # @return [Item]
+      #   The next item for this Item.
+      def next_item
+        Item.new(@rule, @position + 1)
+      end
+
+      # Represent the Item as a String.
+      #
+      # @return [String]
+      #   The Item represented as a String.
+      def to_s
+        parts = @rule.components.map(&:name)
+        parts[@position, 0] = "."
+        "#{@rule.name} -> #{parts.join(" ")}"
+      end
+
+    end
+
+  end
+end

lib/propane/parser/item_set.rb

@@ -0,0 +1,172 @@
+class Propane
+  class Parser
+
+    # Represent a parser "item set", which is a set of possible items that the
+    # parser could currently be parsing. This is equivalent to a parser state.
+    class ItemSet
+
+      # @return [Set<Item>]
+      #   Items in this ItemSet.
+      attr_reader :items
+
+      # @return [Integer]
+      #   ID of this ItemSet.
+      attr_accessor :id
+
+      # @return [Hash]
+      #   Maps a next symbol to its ItemSet.
+      attr_reader :next_item_set
+
+      # @return [Set<ItemSet>]
+      #   ItemSets leading to this item set.
+      attr_reader :in_sets
+
+      # @return [nil, Rule]
+      #   Rule to reduce if there is only one possibility.
+      attr_accessor :reduce_rule
+
+      # @return [Set<Rule>]
+      #   Set of rules that could be reduced in this parser state.
+      attr_accessor :reduce_rules
+
+      # @return [nil, Hash]
+      #   Reduce actions, mapping lookahead tokens to rules, if there is
+      #   more than one rule that could be reduced.
+      attr_accessor :reduce_actions
+
+      # @return [Array<Hash>]
+      #   Shift table entries.
+      attr_accessor :shift_entries
+
+      # Build an ItemSet.
+      #
+      # @param items [Array<Item>]
+      #   Items in this ItemSet.
+      def initialize(items)
+        @items = Set.new(items)
+        @next_item_set = {}
+        @in_sets = Set.new
+        close!
+      end
+
+      # Get the set of next symbols for all Items in this ItemSet.
+      #
+      # @return [Set<Token, RuleSet>]
+      #   Set of next symbols for all Items in this ItemSet.
+      def next_symbols
+        @_next_symbols ||= Set.new(@items.map(&:next_symbol).compact)
+      end
+
+      # Build a next ItemSet for the given next symbol.
+      #
+      # @param symbol [Token, RuleSet]
+      #   Next symbol to build the next ItemSet for.
+      #
+      # @return [ItemSet]
+      #   Next ItemSet for the given next symbol.
+      def build_next_item_set(symbol)
+        ItemSet.new(items_with_next(symbol).map(&:next_item))
+      end
+
+      # Hash function.
+      #
+      # @return [Integer]
+      #   Hash code.
+      def hash
+        @items.hash
+      end
+
+      # Compare ItemSet objects.
+      #
+      # @param other [ItemSet]
+      #   ItemSet to compare to.
+      #
+      # @return [Boolean]
+      #   Whether the ItemSets are equal.
+      def ==(other)
+        @items.eql?(other.items)
+      end
+
+      # Compare ItemSet objects.
+      #
+      # @param other [ItemSet]
+      #   ItemSet to compare to.
+      #
+      # @return [Boolean]
+      #   Whether the ItemSets are equal.
+      def eql?(other)
+        self == other
+      end
+
+      # Set of ItemSets that lead to this ItemSet.
+      #
+      # @return [Set<ItemSet>]
+      #   Set of all ItemSets that lead up to this ItemSet.
+      def leading_item_sets
+        @_leading_item_sets ||=
+          begin
+            result = Set.new
+            eval_sets = Set[self]
+            evaled = Set.new
+            while eval_sets.size > 0
+              eval_set = eval_sets.first
+              eval_sets.delete(eval_set)
+              evaled << eval_set
+              eval_set.in_sets.each do |in_set|
+                result << in_set
+                unless evaled.include?(in_set)
+                  eval_sets << in_set
+                end
+              end
+            end
+            result
+          end
+      end
+
+      # Represent the ItemSet as a String.
+      #
+      # @return [String]
+      #   The ItemSet represented as a String.
+      def to_s
+        @items.map(&:to_s).join("\n")
+      end
+
+      private
+
+      # Close the ItemSet.
+      #
+      # This is done by recursively adding the closed Items for each Item in
+      # the ItemSet.
+      #
+      # @return [void]
+      def close!
+        eval_items = @items.dup
+        while eval_items.size > 0
+          item = eval_items.first
+          eval_items.delete(item)
+          item.closed_items.each do |new_item|
+            unless @items.include?(new_item)
+              eval_items << new_item
+            end
+          end
+          @items += eval_items
+        end
+      end
+
+      # Get the Items with the given next symbol.
+      #
+      # @param symbol [Token, RuleSet]
+      #   Next symbol.
+      #
+      # @return [Array<Item>]
+      #   Items with the given next symbol.
+      def items_with_next(symbol)
+        @items.select do |item|
+          item.next_symbol?(symbol)
+        end
+      end
+
+    end
+
+  end
+end

lib/propane/pattern.rb

@@ -0,0 +1,65 @@
+class Propane
+
+  class Pattern
+
+    # @return [String, nil]
+    #   Code block to execute when the pattern is matched.
+    attr_reader :code
+
+    # @option options [Integer, nil] :code_id
+    #   Code block ID.
+    attr_accessor :code_id
+
+    # @return [String, nil]
+    #   Pattern.
+    attr_reader :pattern
+
+    # @return [Token, nil]
+    #   Token to be returned by this pattern.
+    attr_reader :token
+
+    # @return [Integer, nil]
+    #   Line number where the pattern was defined in the input grammar.
+    attr_reader :line_number
+
+    # @return [Regex::NFA, nil]
+    #   Regex NFA for matching the pattern.
+    attr_reader :nfa
+
+    # @return [Set]
+    #   Lexer modes for this pattern.
+    attr_accessor :modes
+
+    # @return [String, nil]
+    #   Parser value type name.
+    attr_accessor :ptypename
+
+    # Construct a Pattern.
+    #
+    # @param options [Hash]
+    #   Optional parameters.
+    # @option options [String, nil] :code
+    #   Code block to execute when the pattern is matched.
+    # @option options [String, nil] :pattern
+    #   Pattern.
+    # @option options [Token, nil] :token
+    #   Token to be returned by this pattern.
+    # @option options [Integer, nil] :line_number
+    #   Line number where the token was defined in the input grammar.
+    # @option options [String, nil] :modes
+    #   Lexer modes for this pattern.
+    def initialize(options)
+      @code = options[:code]
+      @pattern = options[:pattern]
+      @token = options[:token]
+      @line_number = options[:line_number]
+      @modes = options[:modes]
+      @ptypename = options[:ptypename]
+      regex = Regex.new(@pattern, @line_number)
+      regex.nfa.end_state.accepts = self
+      @nfa = regex.nfa
+    end
+
+  end
+
+end

lib/propane/regex.rb

@@ -0,0 +1,207 @@
+class Propane
+  class Regex
+
+    attr_reader :unit
+    attr_reader :nfa
+
+    def initialize(pattern, line_number)
+      @pattern = pattern.dup
+      @line_number = line_number
+      @unit = parse_alternates
+      @nfa = @unit.to_nfa
+      if @pattern != ""
+        raise Error.new(%[Line #{@line_number}: unexpected "#{@pattern}" in pattern])
+      end
+    end
+
+    private
+
+    def parse_alternates
+      au = AlternatesUnit.new
+      while @pattern != ""
+        c = @pattern[0]
+        return au if c == ")"
+        @pattern.slice!(0)
+        case c
+        when "["
+          au << parse_character_class
+        when "("
+          au << parse_group
+        when "*", "+", "?", "{"
+          if last_unit = au.last_unit
+            case c
+            when "*"
+              min_count, max_count = 0, nil
+            when "+"
+              min_count, max_count = 1, nil
+            when "?"
+              min_count, max_count = 0, 1
+            when "{"
+              min_count, max_count = parse_curly_count
+            end
+            mu = MultiplicityUnit.new(last_unit, min_count, max_count)
+            au.replace_last!(mu)
+          else
+            raise Error.new("Line #{@line_number}: #{c} follows nothing")
+          end
+        when "|"
+          au.new_alternate!
+        when "\\"
+          au << parse_backslash
+        when "."
+          au << period_character_class
+        else
+          au << CharacterRangeUnit.new(c)
+        end
+      end
+      au
+    end
+
+    def parse_group
+      au = parse_alternates
+      if @pattern[0] != ")"
+        raise Error.new("Line #{@line_number}: unterminated group in pattern")
+      end
+      @pattern.slice!(0)
+      au
+    end
+
+    def parse_character_class
+      ccu = CharacterClassUnit.new
+      index = 0
+      loop do
+        if @pattern == ""
+          raise Error.new("Line #{@line_number}: unterminated character class")
+        end
+        c = @pattern.slice!(0)
+        if c == "]"
+          break
+        elsif c == "^" && index == 0
+          ccu.negate = true
+        elsif c == "-" && (ccu.size == 0 || @pattern[0] == "]")
+          ccu << CharacterRangeUnit.new(c)
+        elsif c == "\\"
+          ccu << parse_backslash
+        elsif c == "-" && @pattern[0] != "]"
+          begin_cu = ccu.last_unit
+          unless begin_cu.is_a?(CharacterRangeUnit) && begin_cu.code_point_range.size == 1
+            raise Error.new("Line #{@line_number}: character range must be between single characters")
+          end
+          if @pattern[0] == "\\"
+            @pattern.slice!(0)
+            end_cu = parse_backslash
+            unless end_cu.is_a?(CharacterRangeUnit) && end_cu.code_point_range.size == 1
+              raise Error.new("Line #{@line_number}: character range must be between single characters")
+            end
+            max_code_point = end_cu.code_point
+          else
+            max_code_point = @pattern[0].ord
+            @pattern.slice!(0)
+          end
+          cru = CharacterRangeUnit.new(begin_cu.first, max_code_point)
+          ccu.replace_last!(cru)
+        else
+          ccu << CharacterRangeUnit.new(c)
+        end
+        index += 1
+      end
+      ccu
+    end
+
+    def parse_curly_count
+      if @pattern =~ /^(\d+)(?:(,)(\d*))?\}(.*)$/
+        min_count, comma, max_count, pattern = $1, $2, $3, $4
+        min_count = min_count.to_i
+        if comma.to_s == ""
+          max_count = min_count
+        elsif max_count.to_s != ""
+          max_count = max_count.to_i
+          if max_count < min_count
+            raise Error.new("Line #{@line_number}: maximum repetition count cannot be less than minimum repetition count")
+          end
+        else
+          max_count = nil
+        end
+        @pattern = pattern
+        [min_count, max_count]
+      else
+        raise Error.new("Line #{@line_number}: unexpected match count following {")
+      end
+    end
+
+    def parse_backslash
+      if @pattern == ""
+        raise Error.new("Line #{@line_number}: error: unfollowed \\")
+      else
+        c = @pattern.slice!(0)
+        case c
+        when "a"
+          CharacterRangeUnit.new("\a")
+        when "b"
+          CharacterRangeUnit.new("\b")
+        when "d"
+          CharacterRangeUnit.new("0", "9")
+        when "D"
+          ccu = CharacterClassUnit.new
+          ccu << CharacterRangeUnit.new("0", "9")
+          ccu.negate = true
+          ccu
+        when "f"
+          CharacterRangeUnit.new("\f")
+        when "n"
+          CharacterRangeUnit.new("\n")
+        when "r"
+          CharacterRangeUnit.new("\r")
+        when "s"
+          ccu = CharacterClassUnit.new
+          ccu << CharacterRangeUnit.new(" ")
+          ccu << CharacterRangeUnit.new("\t")
+          ccu << CharacterRangeUnit.new("\r")
+          ccu << CharacterRangeUnit.new("\n")
+          ccu << CharacterRangeUnit.new("\f")
+          ccu << CharacterRangeUnit.new("\v")
+          ccu
+        when "S"
+          ccu = CharacterClassUnit.new
+          ccu << CharacterRangeUnit.new(" ")
+          ccu << CharacterRangeUnit.new("\t")
+          ccu << CharacterRangeUnit.new("\r")
+          ccu << CharacterRangeUnit.new("\n")
+          ccu << CharacterRangeUnit.new("\f")
+          ccu << CharacterRangeUnit.new("\v")
+          ccu.negate = true
+          ccu
+        when "t"
+          CharacterRangeUnit.new("\t")
+        when "v"
+          CharacterRangeUnit.new("\v")
+        when "w"
+          ccu = CharacterClassUnit.new
+          ccu << CharacterRangeUnit.new("_")
+          ccu << CharacterRangeUnit.new("0", "9")
+          ccu << CharacterRangeUnit.new("a", "z")
+          ccu << CharacterRangeUnit.new("A", "Z")
+          ccu
+        when "W"
+          ccu = CharacterClassUnit.new
+          ccu << CharacterRangeUnit.new("_")
+          ccu << CharacterRangeUnit.new("0", "9")
+          ccu << CharacterRangeUnit.new("a", "z")
+          ccu << CharacterRangeUnit.new("A", "Z")
+          ccu.negate = true
+          ccu
+        else
+          CharacterRangeUnit.new(c)
+        end
+      end
+    end
+
+    def period_character_class
+      ccu = CharacterClassUnit.new
+      ccu << CharacterRangeUnit.new(0, "\n".ord - 1)
+      ccu << CharacterRangeUnit.new("\n".ord + 1, 0xFFFFFFFF)
+      ccu
+    end
+
+  end
+end

lib/propane/regex/nfa.rb

@@ -0,0 +1,26 @@
+class Propane
+  class Regex
+
+    class NFA < FA
+
+      attr_reader :end_state
+
+      def initialize
+        super()
+        @end_state = State.new
+      end
+
+      class << self
+
+        def empty
+          nfa = NFA.new
+          nfa.start_state.add_transition(nil, nfa.end_state)
+          nfa
+        end
+
+      end
+
+    end
+
+  end
+end

lib/propane/regex/unit.rb

@@ -0,0 +1,175 @@
+class Propane
+  class Regex
+
+    class Unit
+    end
+
+    class SequenceUnit < Unit
+      attr_accessor :units
+      def initialize
+        @units = []
+      end
+      def method_missing(*args)
+        @units.__send__(*args)
+      end
+      def to_nfa
+        if @units.empty?
+          NFA.empty
+        else
+          nfa = NFA.new
+          unit_nfas = @units.map do |unit|
+            unit.to_nfa
+          end
+          nfa.start_state.add_transition(nil, unit_nfas[0].start_state)
+          unit_nfas.reduce do |prev_nfa, next_nfa|
+            prev_nfa.end_state.add_transition(nil, next_nfa.start_state)
+            next_nfa
+          end.end_state.add_transition(nil, nfa.end_state)
+          nfa
+        end
+      end
+    end
+
+    class AlternatesUnit < Unit
+      attr_accessor :alternates
+      def initialize
+        @alternates = []
+        new_alternate!
+      end
+      def new_alternate!
+        @alternates << SequenceUnit.new
+      end
+      def <<(unit)
+        @alternates[-1] << unit
+      end
+      def last_unit
+        @alternates[-1][-1]
+      end
+      def replace_last!(new_unit)
+        @alternates[-1][-1] = new_unit
+      end
+      def to_nfa
+        if @alternates.size == 0
+          NFA.empty
+        elsif @alternates.size == 1
+          @alternates[0].to_nfa
+        else
+          nfa = NFA.new
+          alternate_nfas = @alternates.map do |alternate|
+            alternate.to_nfa
+          end
+          alternate_nfas.each do |alternate_nfa|
+            nfa.start_state.add_transition(nil, alternate_nfa.start_state)
+            alternate_nfa.end_state.add_transition(nil, nfa.end_state)
+          end
+          nfa
+        end
+      end
+    end
+
+    class CharacterRangeUnit < Unit
+      attr_reader :code_point_range
+      def initialize(c1, c2 = nil)
+        @code_point_range = CodePointRange.new(c1, c2)
+      end
+      def first
+        @code_point_range.first
+      end
+      def last
+        @code_point_range.last
+      end
+      def to_nfa
+        nfa = NFA.new
+        nfa.start_state.add_transition(@code_point_range, nfa.end_state)
+        nfa
+      end
+    end
+
+    class CharacterClassUnit < Unit
+      attr_accessor :units
+      attr_accessor :negate
+      def initialize
+        @units = []
+        @negate = false
+      end
+      def method_missing(*args, &block)
+        @units.__send__(*args, &block)
+      end
+      def <<(thing)
+        if thing.is_a?(CharacterClassUnit)
+          if thing.negate
+            CodePointRange.invert_ranges(thing.map(&:code_point_range)).each do |cpr|
+              CharacterRangeUnit.new(cpr.first, cpr.last)
+            end
+          else
+            thing.each do |ccu_unit|
+              @units << ccu_unit
+            end
+          end
+        else
+          @units << thing
+        end
+      end
+      def last_unit
+        @units[-1]
+      end
+      def replace_last!(new_unit)
+        @units[-1] = new_unit
+      end
+      def to_nfa
+        nfa = NFA.new
+        if @units.empty?
+          nfa.start_state.add_transition(nil, nfa.end_state)
+        else
+          code_point_ranges = @units.map(&:code_point_range)
+          if @negate
+            code_point_ranges = CodePointRange.invert_ranges(code_point_ranges)
+          end
+          code_point_ranges.each do |code_point_range|
+            nfa.start_state.add_transition(code_point_range, nfa.end_state)
+          end
+        end
+        nfa
+      end
+    end
+
+    class MultiplicityUnit < Unit
+      attr_accessor :unit
+      attr_accessor :min_count
+      attr_accessor :max_count
+      def initialize(unit, min_count, max_count)
+        @unit = unit
+        @min_count = min_count
+        @max_count = max_count
+      end
+      def to_nfa
+        nfa = NFA.new
+        last_state = nfa.start_state
+        unit_nfa = nil
+        @min_count.times do
+          unit_nfa = @unit.to_nfa
+          last_state.add_transition(nil, unit_nfa.start_state)
+          last_state = unit_nfa.end_state
+        end
+        last_state.add_transition(nil, nfa.end_state)
+        if @max_count.nil?
+          if @min_count == 0
+            unit_nfa = @unit.to_nfa
+            last_state.add_transition(nil, unit_nfa.start_state)
+          end
+          unit_nfa.end_state.add_transition(nil, unit_nfa.start_state)
+          unit_nfa.end_state.add_transition(nil, nfa.end_state)
+        else
+          (@max_count - @min_count).times do
+            unit_nfa = @unit.to_nfa
+            last_state.add_transition(nil, unit_nfa.start_state)
+            unit_nfa.end_state.add_transition(nil, nfa.end_state)
+            last_state = unit_nfa.end_state
+          end
+        end
+        nfa
+      end
+    end
+
+  end
+end

lib/propane/rule.rb

@@ -0,0 +1,114 @@
+class Propane
+
+  class Rule
+
+    # @return [Array<Token, RuleSet>]
+    #   Rule components.
+    attr_reader :components
+
+    # @return [Hash]
+    #   Field aliases.
+    attr_reader :aliases
+
+    # @return [String]
+    #   User code associated with the rule.
+    attr_reader :code
+
+    # @return [Integer]
+    #   Rule ID.
+    attr_accessor :id
+
+    # @return [String, nil]
+    #   Parser type name.
+    attr_accessor :ptypename
+
+    # @return [Integer]
+    #   Line number where the rule was defined in the input grammar.
+    attr_reader :line_number
+
+    # @return [String]
+    #   Rule name.
+    attr_reader :name
+
+    # @return [RuleSet]
+    #   The RuleSet that this Rule is a part of.
+    attr_accessor :rule_set
+
+    # @return [Array<Integer>]
+    #   Map this rule's components to their positions in the parent RuleSet's
+    #   node field pointer array. This is used for tree construction.
+    attr_accessor :rule_set_node_field_index_map
+
+    # Construct a Rule.
+    #
+    # @param name [String]
+    #   Rule name.
+    # @param components [Array<String>]
+    #   Rule components.
+    # @param code [String]
+    #   User code associated with the rule.
+    # @param ptypename [String, nil]
+    #   Parser type name.
+    # @param line_number [Integer]
+    #   Line number where the rule was defined in the input grammar.
+    def initialize(name, components, code, ptypename, line_number)
+      @name = name
+      @aliases = {}
+      @components = components.each_with_index.map do |component, i|
+        if component =~ /(\S+):(\S+)/
+          c, aliasname = $1, $2
+          if @aliases[aliasname]
+            raise Error.new("Error: duplicate field alias `#{aliasname}` for rule #{name} defined on line #{line_number}")
+          end
+          @aliases[aliasname] = i
+          c
+        else
+          component
+        end
+      end
+      @rule_set_node_field_index_map = components.map {0}
+      @code = code
+      @ptypename = ptypename
+      @line_number = line_number
+    end
+
+    # Return whether the Rule is empty.
+    #
+    # A Rule is empty if it has no components.
+    #
+    # @return [Boolean]
+    #   Whether the Rule is empty.
+    def empty?
+      @components.empty?
+    end
+
+    # Return whether this is an optional Rule.
+    #
+    # @return [Boolean]
+    #   Whether this is an optional Rule.
+    def optional?
+      @name.end_with?("?")
+    end
+
+    # Represent the Rule as a String.
+    #
+    # @return [String]
+    #   Rule represented as a String.
+    def to_s
+      "#{@name} -> #{@components.map(&:name).join(" ")}"
+    end
+
+    # Check whether the rule set node field index map is just a 1:1 mapping.
+    #
+    # @return [Boolean]
+    #   Boolean indicating whether the rule set node field index map is just a
+    #   1:1 mapping.
+    def flat_rule_set_node_field_index_map?
+      @rule_set_node_field_index_map.each_with_index.all? do |v, i|
+        v == i
+      end
+    end
+
+  end
+
+end

lib/propane/rule_set.rb

@@ -0,0 +1,169 @@
+class Propane
+
+  # A RuleSet collects all grammar rules of the same name.
+  class RuleSet
+
+    # @return [Array<Hash>]
+    #   tree fields.
+    attr_reader :tree_fields
+
+    # @return [Integer]
+    #   ID of the RuleSet.
+    attr_reader :id
+
+    # @return [String]
+    #   Name of the RuleSet.
+    attr_reader :name
+
+    # @return [String, nil]
+    #   Parser type name.
+    attr_accessor :ptypename
+
+    # @return [Array<Rule>]
+    #   Rules in the RuleSet.
+    attr_reader :rules
+
+    # @return [Boolean]
+    #   Whether the RuleSet could expand to an empty sequence.
+    attr_writer :could_be_empty
+
+    # Construct a RuleSet.
+    #
+    # @param name [String]
+    #   Name of the RuleSet.
+    # @param id [Integer]
+    #   ID of the RuleSet.
+    def initialize(name, id)
+      @id = id
+      @name = name
+      @rules = []
+      @could_be_empty = false
+    end
+
+    # Add a Rule to the RuleSet.
+    #
+    # @param rule [Rule]
+    #   Rule to add.
+    def <<(rule)
+      @rules << rule
+    end
+
+    # Return whether any Rule in the RuleSet is empty.
+    #
+    # @return [Boolean]
+    #   Whether any rule in the RuleSet is empty.
+    def could_be_empty?
+      @could_be_empty
+    end
+
+    # Return whether this is an optional RuleSet.
+    #
+    # @return [Boolean]
+    #   Whether this is an optional RuleSet.
+    def optional?
+      @name.end_with?("?")
+    end
+
+    # For optional rule sets, return the underlying component that is optional.
+    def option_target
+      @rules.each do |rule|
+        if rule.components.size > 0
+          return rule.components[0]
+        end
+      end
+      raise "Optional rule target not found"
+    end
+
+    # Build the start token set for the RuleSet.
+    #
+    # @return [Set<Token>]
+    #   Start token set for the RuleSet.
+    def start_token_set
+      if @_start_token_set.nil?
+        @_start_token_set = Set.new
+        @rules.each do |rule|
+          rule.components.each do |component|
+            if component.is_a?(Token)
+              @_start_token_set << component
+              break
+            else
+              @_start_token_set += component.start_token_set
+              unless component.could_be_empty?
+                break
+              end
+            end
+          end
+        end
+      end
+      @_start_token_set
+    end
+
+    # Finalize a RuleSet after adding all Rules to it.
+    def finalize(grammar)
+      if grammar.tree
+        build_tree_fields(grammar)
+      end
+    end
+
+    private
+
+    # Build the set of tree fields for this RuleSet.
+    #
+    # This is an Array of Hashes. Each entry in the Array corresponds to a
+    # field location in the tree node. The entry is a Hash. It could have one or
+    # two keys. It will always have the field name with a positional suffix as
+    # a key. It may also have the field name without the positional suffix if
+    # that field only exists in one position across all Rules in the RuleSet.
+    #
+    # @return [void]
+    def build_tree_fields(grammar)
+      field_tree_node_indexes = {}
+      field_indexes_across_all_rules = {}
+      # Stores the index into @tree_fields by field alias name.
+      field_aliases = {}
+      @tree_fields = []
+      @rules.each do |rule|
+        rule.components.each_with_index do |component, i|
+          if component.is_a?(RuleSet) && component.optional?
+            component = component.option_target
+          end
+          if component.is_a?(Token)
+            node_name = "Token"
+          else
+            node_name = component.name
+          end
+          struct_name = "#{grammar.tree_prefix}#{node_name}#{grammar.tree_suffix}"
+          field_name = "p#{node_name}#{i + 1}"
+          unless field_tree_node_indexes[field_name]
+            field_tree_node_indexes[field_name] = @tree_fields.size
+            @tree_fields << {field_name => struct_name}
+          end
+          rule.aliases.each do |alias_name, index|
+            if index == i
+              alias_tree_fields_index = field_tree_node_indexes[field_name]
+              if field_aliases[alias_name] && field_aliases[alias_name] != alias_tree_fields_index
+                raise Error.new("Error: conflicting tree node field positions for alias `#{alias_name}` in rule #{rule.name} defined on line #{rule.line_number}")
+              end
+              field_aliases[alias_name] = alias_tree_fields_index
+              @tree_fields[alias_tree_fields_index][alias_name] = @tree_fields[alias_tree_fields_index].first[1]
+            end
+          end
+          field_indexes_across_all_rules[node_name] ||= Set.new
+          field_indexes_across_all_rules[node_name] << field_tree_node_indexes[field_name]
+          rule.rule_set_node_field_index_map[i] = field_tree_node_indexes[field_name]
+        end
+      end
+      field_indexes_across_all_rules.each do |node_name, indexes_across_all_rules|
+        if indexes_across_all_rules.size == 1
+          # If this field was only seen in one position across all rules,
+          # then add an alias to the positional field name that does not
+          # include the position.
+          @tree_fields[indexes_across_all_rules.first]["p#{node_name}"] =
+            "#{grammar.tree_prefix}#{node_name}#{grammar.tree_suffix}"
+        end
+      end
+    end
+
+  end
+
+end

lib/propane/token.rb

@@ -0,0 +1,61 @@
+class Propane
+
+  class Token
+
+    class << self
+
+      # Name of the token to use in code (special characters replaced).
+      #
+      # @return [String]
+      #   Name of the token to use in code (special characters replaced).
+      def code_name(name)
+        name.sub(/^\$/, "__")
+      end
+
+    end
+
+    # @return [String, nil]
+    #   Token name.
+    attr_reader :name
+
+    # @return [String, nil]
+    #   Parser value type name.
+    attr_accessor :ptypename
+
+    # @return [Integer, nil]
+    #   Token ID.
+    attr_accessor :id
+
+    # @return [Integer, nil]
+    #   Line number where the token was defined in the input grammar.
+    attr_reader :line_number
+
+    # Construct a Token.
+    #
+    # @param name [String, nil]
+    #   Token name.
+    # @param ptypename [String, nil]
+    #   Parser value type for this token.
+    # @param line_number [Integer, nil]
+    #   Line number where the token was defined in the input grammar.
+    def initialize(name, ptypename, line_number)
+      @name = name
+      @ptypename = ptypename
+      @line_number = line_number
+    end
+
+    # Name of the token to use in code (special characters replaced).
+    #
+    # @return [String]
+    #   Name of the token to use in code (special characters replaced).
+    def code_name
+      self.class.code_name(@name)
+    end
+
+    def to_s
+      @name
+    end
+
+  end
+
+end

lib/propane/util.rb

@@ -0,0 +1,43 @@
+class Propane
+
+  # Utility methods.
+  module Util
+
+    class << self
+
+      def banner(message)
+        s = "*" * (message.size + 4)
+        "#{s}\n* #{message} *\n#{s}\n"
+      end
+
+      # Determine the number of threads to use.
+      #
+      # @return [Integer]
+      #   The number of threads to use.
+      def determine_n_threads
+        # Try to figure out how many threads are available on the host hardware.
+        begin
+          case RbConfig::CONFIG["host_os"]
+          when /linux/
+            return File.read("/proc/cpuinfo").scan(/^processor\s*:/).size
+          when /mswin|mingw|msys/
+            if `wmic cpu get NumberOfLogicalProcessors -value` =~ /NumberOfLogicalProcessors=(\d+)/
+              return $1.to_i
+            end
+          when /darwin/
+            if `sysctl -n hw.ncpu` =~ /(\d+)/
+              return $1.to_i
+            end
+          end
+        rescue
+        end
+
+        # If we can't figure it out, default to 4.
+        4
+      end
+
+    end
+
+  end
+
+end

lib/propane/version.rb

@@ -0,0 +1,3 @@
+class Propane
+  VERSION = "5.3.0"
+end

rb/assets/user_guide.html.erb

@@ -0,0 +1,68 @@
+<html>
+  <head>
+    <title>Propane User Guide<%= subpage_title %> - Version <%= Propane::VERSION %></title>
+    <style>
+      body {
+        background-color: #CCC;
+        margin: 0px;
+        padding: 0px;
+      }
+      #body-content {
+        margin-left: auto;
+        margin-right: auto;
+        margin-top: 1em;
+        margin-bottom: 1em;
+        border: 1px solid black;
+        background-color: #FFF;
+        width: 120ex;
+        padding: 2ex;
+      }
+      a {
+        text-decoration: none;
+      }
+      .separator {
+        height: 2em;
+      }
+      .img_block_center {
+        display: block;
+        margin-left: auto;
+        margin-right: auto;
+      }
+      .page_nav {
+        width: 100%;
+        table-layout: fixed;
+      }
+      .page_nav_toc {
+        text-align: center;
+      }
+      .page_nav_next {
+        text-align: right;
+      }
+      .code {
+        padding-left: 2ex;
+        width: 116ex;
+        overflow-x: auto;
+      }
+      .ruby_code .normal {}
+      .ruby_code .comment { color: #005; font-style: italic; }
+      .ruby_code .keyword { color: #A00; font-weight: bold; }
+      .ruby_code .method { color: #077; }
+      .ruby_code .class { color: #074; }
+      .ruby_code .module { color: #050; }
+      .ruby_code .punct { color: #447; font-weight: bold; }
+      .ruby_code .symbol { color: #099; }
+      .ruby_code .string { color: #090; }
+      .ruby_code .char { color: #F07; }
+      .ruby_code .ident { color: #004; }
+      .ruby_code .constant { color: #07F; }
+      .ruby_code .regex { color: #B66; }
+      .ruby_code .number { color: #D55; }
+      .ruby_code .attribute { color: #377; }
+      .ruby_code .global { color: #3B7; }
+      .ruby_code .expr { color: #227; }
+    </style>
+  </head>
+  <body>
+    <div id="body-content"><%= content %></div>
+  </body>
+</html>

rb/build_dist.rb

@@ -0,0 +1,129 @@
+#!/usr/bin/env ruby
+
+require "erb"
+require "fileutils"
+require "digest/md5"
+
+if File.read("lib/propane/version.rb") =~ /VERSION = "(.+)"/
+  VERSION = $1
+else
+  raise "Could not determine version."
+end
+PROG_NAME = "propane"
+START_FILE = "bin/#{PROG_NAME}"
+LIB_DIR = "lib"
+DIST = "dist"
+
+ASSETS_TEMPLATE = <<EOF
+class Propane
+  module Assets
+    class << self
+      def get(name)
+        case name
+<% Dir.glob("assets/*").each  do |asset_file| %>
+        when <%= File.basename(asset_file).inspect %>
+          <%= File.binread(asset_file).inspect %>
+<% end %>
+        end
+      end
+    end
+  end
+end
+EOF
+
+assets_module = ERB.new(ASSETS_TEMPLATE, trim_mode: "<>").result
+files_processed = {}
+combined_file = []
+
+combine_files = lambda do |file|
+  File.read(file, mode: "rb").each_line do |line|
+    if line =~ /^\s*require(?:_relative)?\s*"(.*)"$/
+      require_name = $1
+      if require_name =~ %r{^#{PROG_NAME}(?:/.*)?$}
+        path = "#{LIB_DIR}/#{require_name}.rb"
+        if File.exist?(path)
+          unless files_processed[path]
+            files_processed[path] = true
+            if require_name == "propane/assets"
+              combined_file << assets_module
+            else
+              combine_files[path]
+            end
+          end
+        else
+          raise "require path #{path.inspect} not found"
+        end
+      else
+        combined_file << line
+      end
+    else
+      combined_file << line
+    end
+  end
+end
+
+combine_files[START_FILE]
+
+# Strip Ruby comment lines and empty lines to save some space, but do not
+# remove lines that are in heredoc sections. This isn't terribly robust to be
+# used in the wild, but works for the heredoc instances for this project.
+stripped = []
+heredoc_end = nil
+combined_file.each do |line|
+  if line =~ /<<-?([A-Z]+)/
+    heredoc_end = $1
+  end
+  if heredoc_end and line =~ /^\s*#{heredoc_end}/
+    heredoc_end = nil
+  end
+  if line !~ /#\sspecs/
+    if heredoc_end or line !~ /^\s*(#[^!].*)?$/
+      stripped << line
+    end
+  end
+end
+
+license = File.read("LICENSE.txt").gsub(/^(.*?)$/) do |line|
+  if line.size > 0
+    "# #{line}"
+  else
+    "#"
+  end
+end
+
+require "zlib"
+require "base64"
+compressed_script = Zlib::Deflate.deflate(stripped.join)
+encoded_compressed_script = Base64.encode64(compressed_script).gsub("\n", "")
+hash = Digest::MD5.hexdigest(encoded_compressed_script)
+
+FileUtils.rm_rf(DIST)
+FileUtils.mkdir_p(DIST)
+File.open("#{DIST}/#{PROG_NAME}", "wb", 0755) do |fh|
+  fh.write(<<EOF)
+#!/usr/bin/env ruby
+
+#{license}
+
+script = File.join(File.dirname(__FILE__), ".propane-#{VERSION}-#{hash}.rb")
+unless File.exist?(script)
+  if File.read(__FILE__, mode: "rb") =~ /^#==>(.*)/
+    require "zlib"
+    require "base64"
+    encoded_compressed = $1
+    unescaped_compressed = Base64.decode64(encoded_compressed)
+    inflated = Zlib::Inflate.inflate(unescaped_compressed)
+    File.open(script, "wb") do |fh|
+      fh.write(inflated)
+    end
+  else
+    raise "Could not decompress."
+  end
+end
+load script
+if __FILE__ == $0
+  Propane::CLI.new.run(ARGV)
+end
+#==>#{encoded_compressed_script}
+EOF
+end

rb/gen_user_guide.rb

@@ -0,0 +1,227 @@
+#!/usr/bin/env ruby
+
+require "erb"
+require "fileutils"
+require "redcarpet"
+require "syntax"
+require "syntax/convertors/html"
+require "propane/version"
+
+def load_file(file_name)
+  contents = File.read(file_name)
+  contents.gsub(/\$\{include (.*?)\}/) do |match|
+    include_file_name = $1
+    File.read(include_file_name)
+  end
+end
+
+class Generator
+  class Section
+    attr_reader :number
+    attr_reader :title
+    attr_reader :page
+    attr_reader :contents
+    attr_reader :anchor
+    def initialize(number, title, page, anchor)
+      @number = number
+      @title = title.gsub(/[`]/, "")
+      @page = page
+      @anchor = anchor
+      @contents = ""
+    end
+    def append(contents)
+      @contents += contents
+    end
+  end
+
+  class Page
+    attr_reader :name
+    attr_reader :title
+    attr_accessor :contents
+    def initialize(name, title, contents = "")
+      @name = name
+      @title = title
+      @contents = contents
+    end
+  end
+
+  def initialize(input, output_file, multi_page)
+    current_page =
+      if multi_page
+        nil
+      else
+        File.basename(output_file).sub(/\.html$/, "")
+      end
+    @sections = []
+    @current_section_number = [0]
+    @lines = input.lines
+    while @lines.size > 0
+      line = @lines.slice!(0)
+      if line =~ /^```(.*)$/
+        @sections.last.append(gather_code_section($1))
+      elsif line.chomp == "${remove}"
+        remove_section
+      elsif line =~ /^(#+)(>)?\s*(.*)$/
+        level_text, new_page_text, title_text = $1, $2, $3
+        level = $1.size
+        new_page = !new_page_text.nil?
+        section_number = get_next_section_number(level)
+        anchor = make_anchor(section_number, title_text)
+        if multi_page and (new_page or current_page.nil?)
+          current_page = anchor
+        end
+        @sections << Section.new(section_number, title_text, current_page, anchor)
+        @sections.last.append("#{level_text} #{section_number} #{title_text}")
+      elsif @sections.size > 0
+        @sections.last.append(line)
+      end
+    end
+
+    renderer = Redcarpet::Render::HTML.new
+    @markdown_renderer = Redcarpet::Markdown.new(renderer)
+    changelog = @markdown_renderer.render(File.read("CHANGELOG.md"))
+
+    @pages = [Page.new("toc", "Table of Contents", render_toc)]
+    @sections.each do |section|
+      unless @pages.last.name == section.page
+        @pages << Page.new(section.page, "#{section.number} #{section.title}")
+      end
+      @pages.last.contents += render_section(section)
+    end
+
+    @pages.each do |page|
+      page.contents.gsub!("${changelog}", changelog)
+      page.contents.gsub!(%r{\$\{#(.+?)\}}) do |match|
+        section_name = $1
+        href = get_link_to_section(section_name)
+        %[<a href="#{href}">#{section_name}</a>]
+      end
+    end
+
+    template = File.read("rb/assets/user_guide.html.erb")
+    erb = ERB.new(template, trim_mode: "<>")
+
+    if multi_page
+      @pages.each_with_index do |page, page_index|
+        subpage_title = " - #{page.title}"
+        page_nav_bar = render_page_nav_bar(page_index)
+        content = page_nav_bar + separator + page.contents + separator + page_nav_bar
+        html_result = erb.result(binding.clone)
+        File.open(File.join(output_file, "#{page.name}.html"), "w") do |fh|
+          fh.write(html_result)
+        end
+      end
+    else
+      subpage_title = ""
+      content = @pages.reduce("") do |result, page|
+        result + page.contents
+      end
+      html_result = erb.result(binding.clone)
+      File.open(output_file, "w") do |fh|
+        fh.write(html_result)
+      end
+    end
+  end
+
+  def separator
+    %[<div class="separator"></div>]
+  end
+
+  def render_page_nav_bar(page_index)
+    page_nav_prev =
+      if page_index > 1
+        %[<a href="#{@pages[page_index - 1].name}.html">&laquo; Prev<br/>#{@pages[page_index - 1].title}</a>]
+      else
+        ""
+      end
+    page_nav_toc =
+      if page_index > 0
+        %[<a href="toc.html">Table of Contents</a>]
+      else
+        ""
+      end
+    page_nav_next =
+      if page_index < @pages.size - 1
+        %[<a href="#{@pages[page_index + 1].name}.html">Next &raquo<br/>#{@pages[page_index + 1].title}</a>]
+      else
+        ""
+      end
+    %[<table class="page_nav"><tr>] + \
+    %[<td class="page_nav_prev">#{page_nav_prev}</td>] + \
+    %[<td class="page_nav_toc">#{page_nav_toc}</td>] + \
+    %[<td class="page_nav_next">#{page_nav_next}</td>] + \
+    %[</tr></table>]
+  end
+
+  def render_toc
+    toc_content = %[<h1>Table of Contents</h1>\n]
+    @sections.each do |section|
+      indent = section.number.split(".").size - 1
+      toc_content += %[<span style="padding-left: #{4 * indent}ex;">]
+      toc_content += %[<a href="#{section.page}.html##{section.anchor}">#{section.number} #{section.title}</a><br/>\n]
+      toc_content += %[</span>]
+    end
+    toc_content
+  end
+
+  def render_section(section)
+    %[<a name="#{section.anchor}" />] + \
+      @markdown_renderer.render(section.contents)
+  end
+
+  def make_anchor(section_number, section_title)
+    "s" + ("#{section_number} #{section_title}").gsub(/[^a-zA-Z0-9]/, "_")
+  end
+
+  def get_link_to_section(section_name)
+    section = @sections.find do |section|
+      section.title == section_name
+    end
+    raise "Could not find section #{section_name}" unless section
+    "#{section.page}.html##{section.anchor}"
+  end
+
+  def gather_code_section(syntax)
+    code = ""
+    loop do
+      line = @lines.slice!(0)
+      if line =~ /^```/
+        break
+      end
+      code += line
+    end
+    if syntax != ""
+      convertor = Syntax::Convertors::HTML.for_syntax(syntax)
+      %[<div class="code #{syntax}_code">\n#{convertor.convert(code)}\n</div>\n]
+    else
+      %[<div class="code">\n<pre>#{code}</pre>\n</div>\n]
+    end
+  end
+
+  def remove_section
+    loop do
+      line = @lines.slice!(0)
+      if line.chomp == "${/remove}"
+        break
+      end
+    end
+  end
+
+  def get_next_section_number(level)
+    if @current_section_number.size == level - 1
+      @current_section_number << 1
+    elsif @current_section_number.size >= level
+      @current_section_number[level - 1] += 1
+      @current_section_number.slice!(level, @current_section_number.size)
+    else
+      raise "Section level change from #{@current_section_number.size} to #{level}"
+    end
+    @current_section_number.join(".")
+  end
+end
+
+input = load_file("doc/user_guide.md")
+FileUtils.rm_rf("gen/user_guide")
+FileUtils.mkdir_p("gen/user_guide")
+Generator.new(input, "gen/user_guide/user_guide.html", false)
+Generator.new(input, "gen/user_guide", true)

spec/json_parser.c.propane

@@ -0,0 +1,183 @@
+<<header
+#include "json_types.h"
+#include "testutils.h"
+>>
+<<
+#include "math.h"
+#include <stdbool.h>
+static str_t string_value;
+>>
+
+ptype JSONValue *;
+
+drop /\s+/;
+token lbrace /\{/;
+token rbrace /\}/;
+token lbracket /\[/;
+token rbracket /\]/;
+token comma /,/;
+token colon /:/;
+token number /-?(0|[1-9][0-9]*)(\.[0-9]+)?([eE][-+]?[0-9]+)?/ <<
+    double n = 0.0;
+    bool negative = false;
+    size_t i = 0u;
+    if (match_text[i] == '-')
+    {
+        negative = true;
+        i++;
+    }
+    while ('0' <= match_text[i] && match_text[i] <= '9')
+    {
+        n *= 10.0;
+        n += (match_text[i] - '0');
+        i++;
+    }
+    if (match_text[i] == '.')
+    {
+        i++;
+        double mult = 0.1;
+        while ('0' <= match_text[i] && match_text[i] <= '9')
+        {
+            n += mult * (match_text[i] - '0');
+            mult /= 10.0;
+            i++;
+        }
+    }
+    if (match_text[i] == 'e' || match_text[i] == 'E')
+    {
+        bool exp_negative = false;
+        i++;
+        if (match_text[i] == '-')
+        {
+            exp_negative = true;
+            i++;
+        }
+        else if (match_text[i] == '+')
+        {
+            i++;
+        }
+        long exp = 0.0;
+        while ('0' <= match_text[i] && match_text[i] <= '9')
+        {
+            exp *= 10;
+            exp += (match_text[i] - '0');
+            i++;
+        }
+        if (exp_negative)
+        {
+            exp = -exp;
+        }
+        n = pow(n, exp);
+    }
+    if (negative)
+    {
+        n = -n;
+    }
+    $$ = JSONValue_new(JSON_NUMBER);
+    $$->number = n;
+>>
+token true <<
+  $$ = JSONValue_new(JSON_TRUE);
+>>
+token false <<
+  $$ = JSONValue_new(JSON_FALSE);
+>>
+token null <<
+  $$ = JSONValue_new(JSON_NULL);
+>>
+/"/ <<
+  $mode(string);
+  str_init(&string_value, "");
+>>
+string: token string /"/ <<
+  $$ = JSONValue_new(JSON_STRING);
+  $$->string = string_value;
+  $mode(default);
+>>
+string: /\\"/ <<
+  str_append(&string_value, "\"");
+>>
+string: /\\\\/ <<
+  str_append(&string_value, "\\");
+>>
+string: /\\\// <<
+  str_append(&string_value, "/");
+>>
+string: /\\b/ <<
+  str_append(&string_value, "\b");
+>>
+string: /\\f/ <<
+  str_append(&string_value, "\f");
+>>
+string: /\\n/ <<
+  str_append(&string_value, "\n");
+>>
+string: /\\r/ <<
+  str_append(&string_value, "\r");
+>>
+string: /\\t/ <<
+  str_append(&string_value, "\t");
+>>
+string: /\\u[0-9a-fA-F]{4}/ <<
+  /* Not actually going to encode the code point for this example... */
+  char s[] = {'{', (char)match_text[2], (char)match_text[3], (char)match_text[4], (char)match_text[5], '}', 0};
+  str_append(&string_value, s);
+>>
+string: /[^\\]/ <<
+  char s[] = {(char)match_text[0], 0};
+  str_append(&string_value, s);
+>>
+Start -> Value <<
+  $$ = $1;
+>>
+Value -> string <<
+  $$ = $1;
+>>
+Value -> number <<
+  $$ = $1;
+>>
+Value -> Object <<
+  $$ = $1;
+>>
+Value -> Array <<
+  $$ = $1;
+>>
+Value -> true <<
+  $$ = $1;
+>>
+Value -> false <<
+  $$ = $1;
+>>
+Value -> null <<
+  $$ = $1;
+>>
+Object -> lbrace rbrace <<
+  $$ = JSONObject_new();
+>>
+Object -> lbrace KeyValues rbrace <<
+  $$ = $2;
+>>
+KeyValues -> KeyValue <<
+  $$ = $1;
+>>
+KeyValues -> KeyValues comma KeyValue <<
+  JSONObject_append($1, $3->object.entries[0].name, $3->object.entries[0].value);
+  $$ = $1;
+>>
+KeyValue -> string colon Value <<
+  $$ = JSONObject_new();
+  JSONObject_append($$, str_cstr(&$1->string), $3);
+>>
+Array -> lbracket rbracket <<
+  $$ = JSONArray_new();
+>>
+Array -> lbracket Values rbracket <<
+  $$ = $2;
+>>
+Values -> Value <<
+  $$ = $1;
+>>
+Values -> Values comma Value <<
+  JSONArray_append($1, $3);
+  $$ = $1;
+>>

spec/json_parser.d.propane

@@ -0,0 +1,179 @@
+<<
+import std.math;
+import json_types;
+private string string_value;
+>>
+
+ptype JSONValue;
+ptype array = JSONValue[];
+ptype dict = JSONValue[string];
+ptype string = string;
+
+drop /\s+/;
+token lbrace /\{/;
+token rbrace /\}/;
+token lbracket /\[/;
+token rbracket /\]/;
+token comma /,/;
+token colon /:/;
+token number /-?(0|[1-9][0-9]*)(\.[0-9]+)?([eE][-+]?[0-9]+)?/ <<
+    double n;
+    bool negative;
+    size_t i = 0u;
+    if (match_text[i] == '-')
+    {
+        negative = true;
+        i++;
+    }
+    while ('0' <= match_text[i] && match_text[i] <= '9')
+    {
+        n *= 10.0;
+        n += (match_text[i] - '0');
+        i++;
+    }
+    if (match_text[i] == '.')
+    {
+        i++;
+        double mult = 0.1;
+        while ('0' <= match_text[i] && match_text[i] <= '9')
+        {
+            n += mult * (match_text[i] - '0');
+            mult /= 10.0;
+            i++;
+        }
+    }
+    if (match_text[i] == 'e' || match_text[i] == 'E')
+    {
+        bool exp_negative;
+        i++;
+        if (match_text[i] == '-')
+        {
+            exp_negative = true;
+            i++;
+        }
+        else if (match_text[i] == '+')
+        {
+            i++;
+        }
+        long exp;
+        while ('0' <= match_text[i] && match_text[i] <= '9')
+        {
+            exp *= 10;
+            exp += (match_text[i] - '0');
+            i++;
+        }
+        if (exp_negative)
+        {
+            exp = -exp;
+        }
+        n = pow(n, exp);
+    }
+    if (negative)
+    {
+        n = -n;
+    }
+    $$ = new JSONNumber(n);
+>>
+token true <<
+  $$ = new JSONTrue();
+>>
+token false <<
+  $$ = new JSONFalse();
+>>
+token null <<
+  $$ = new JSONNull();
+>>
+/"/ <<
+  $mode(string);
+  string_value = "";
+>>
+string: token string (string) /"/ <<
+  $$ = string_value;
+  $mode(default);
+>>
+string: /\\"/ <<
+  string_value ~= "\"";
+>>
+string: /\\\\/ <<
+  string_value ~= "\\";
+>>
+string: /\\\// <<
+  string_value ~= "/";
+>>
+string: /\\b/ <<
+  string_value ~= "\b";
+>>
+string: /\\f/ <<
+  string_value ~= "\f";
+>>
+string: /\\n/ <<
+  string_value ~= "\n";
+>>
+string: /\\r/ <<
+  string_value ~= "\r";
+>>
+string: /\\t/ <<
+  string_value ~= "\t";
+>>
+string: /\\u[0-9a-fA-F]{4}/ <<
+  /* Not actually going to encode the code point for this example... */
+  string_value ~= "{" ~ match_text[2..6] ~ "}";
+>>
+string: /[^\\]/ <<
+  string_value ~= match_text;
+>>
+Start -> Value <<
+  $$ = $1;
+>>
+Value -> string <<
+  $$ = new JSONString($1);
+>>
+Value -> number <<
+  $$ = $1;
+>>
+Value -> Object <<
+  $$ = $1;
+>>
+Value -> Array <<
+  $$ = $1;
+>>
+Value -> true <<
+  $$ = $1;
+>>
+Value -> false <<
+  $$ = $1;
+>>
+Value -> null <<
+  $$ = $1;
+>>
+Object -> lbrace rbrace <<
+  $$ = new JSONObject();
+>>
+Object -> lbrace KeyValues rbrace <<
+  $$ = new JSONObject($2);
+>>
+KeyValues (dict) -> KeyValue <<
+  $$ = $1;
+>>
+KeyValues -> KeyValues comma KeyValue <<
+  foreach (key, value; $3)
+  {
+    $1[key] = value;
+  }
+  $$ = $1;
+>>
+KeyValue (dict) -> string colon Value <<
+  $$ = [$1: $3];
+>>
+Array -> lbracket rbracket <<
+  $$ = new JSONArray();
+>>
+Array -> lbracket Values rbracket <<
+  $$ = new JSONArray($2);
+>>
+Values (array) -> Value <<
+  $$ = [$1];
+>>
+Values -> Values comma Value <<
+  $$ = $1 ~ [$3];
+>>

spec/json_parser.rust.propane

@@ -0,0 +1,176 @@
+<<
+pub const JSON_OBJECT: usize = 0;
+pub const JSON_ARRAY: usize = 1;
+pub const JSON_NUMBER: usize = 2;
+pub const JSON_STRING: usize = 3;
+pub const JSON_TRUE: usize = 4;
+pub const JSON_FALSE: usize = 5;
+pub const JSON_NULL: usize = 6;
+
+#[derive(Clone, Default)]
+pub enum JSONValue {
+    #[default]
+    Null,
+    Object(Vec<(String, JSONValue)>),
+    Array(Vec<JSONValue>),
+    Number(f64),
+    StringVal(String),
+    True,
+    False,
+}
+
+impl JSONValue {
+    pub fn id(&self) -> usize {
+        match self {
+            JSONValue::Object(_) => JSON_OBJECT,
+            JSONValue::Array(_) => JSON_ARRAY,
+            JSONValue::Number(_) => JSON_NUMBER,
+            JSONValue::StringVal(_) => JSON_STRING,
+            JSONValue::True => JSON_TRUE,
+            JSONValue::False => JSON_FALSE,
+            JSONValue::Null => JSON_NULL,
+        }
+    }
+    pub fn number(&self) -> f64 {
+        if let JSONValue::Number(n) = self { *n } else { 0.0 }
+    }
+    pub fn string(&self) -> &str {
+        if let JSONValue::StringVal(s) = self { s.as_str() } else { "" }
+    }
+    pub fn object_len(&self) -> usize {
+        if let JSONValue::Object(e) = self { e.len() } else { 0 }
+    }
+    pub fn array_len(&self) -> usize {
+        if let JSONValue::Array(e) = self { e.len() } else { 0 }
+    }
+}
+>>
+
+context_user_fields <<
+    pub string_value: String,
+>>
+
+ptype JSONValue;
+
+drop /\s+/;
+token lbrace /\{/;
+token rbrace /\}/;
+token lbracket /\[/;
+token rbracket /\]/;
+token comma /,/;
+token colon /:/;
+token number /-?(0|[1-9][0-9]*)(\.[0-9]+)?([eE][-+]?[0-9]+)?/ <<
+    let n: f64 = std::str::from_utf8(match_text).unwrap().parse().unwrap();
+    $$ = JSONValue::Number(n);
+>>
+token true <<
+  $$ = JSONValue::True;
+>>
+token false <<
+  $$ = JSONValue::False;
+>>
+token null <<
+  $$ = JSONValue::Null;
+>>
+/"/ <<
+  $mode(string);
+  ${context.string_value} = String::new();
+>>
+string: token string /"/ <<
+  $$ = JSONValue::StringVal(std::mem::take(&mut ${context.string_value}));
+  $mode(default);
+>>
+string: /\\"/ <<
+  ${context.string_value}.push('"');
+>>
+string: /\\\\/ <<
+  ${context.string_value}.push('\\');
+>>
+string: /\\\// <<
+  ${context.string_value}.push('/');
+>>
+string: /\\b/ <<
+  ${context.string_value}.push('\u{0008}');
+>>
+string: /\\f/ <<
+  ${context.string_value}.push('\u{000C}');
+>>
+string: /\\n/ <<
+  ${context.string_value}.push('\n');
+>>
+string: /\\r/ <<
+  ${context.string_value}.push('\r');
+>>
+string: /\\t/ <<
+  ${context.string_value}.push('\t');
+>>
+string: /\\u[0-9a-fA-F]{4}/ <<
+  /* Not actually going to encode the code point for this example... */
+  let s: String = ['{', match_text[2] as char, match_text[3] as char, match_text[4] as char, match_text[5] as char, '}'].iter().collect();
+  ${context.string_value}.push_str(&s);
+>>
+string: /[^\\]/ <<
+  ${context.string_value}.push(match_text[0] as char);
+>>
+Start -> Value <<
+  $$ = $1;
+>>
+Value -> string <<
+  $$ = $1;
+>>
+Value -> number <<
+  $$ = $1;
+>>
+Value -> Object <<
+  $$ = $1;
+>>
+Value -> Array <<
+  $$ = $1;
+>>
+Value -> true <<
+  $$ = $1;
+>>
+Value -> false <<
+  $$ = $1;
+>>
+Value -> null <<
+  $$ = $1;
+>>
+Object -> lbrace rbrace <<
+  $$ = JSONValue::Object(Vec::new());
+>>
+Object -> lbrace KeyValues rbrace <<
+  $$ = $2;
+>>
+KeyValues -> KeyValue <<
+  $$ = $1;
+>>
+KeyValues -> KeyValues comma KeyValue <<
+  let mut obj = $1;
+  if let JSONValue::Object(kve) = $3 {
+      if let JSONValue::Object(entries) = &mut obj {
+          entries.extend(kve);
+      }
+  }
+  $$ = obj;
+>>
+KeyValue -> string colon Value <<
+  let name = if let JSONValue::StringVal(s) = $1 { s } else { String::new() };
+  $$ = JSONValue::Object(vec![(name, $3)]);
+>>
+Array -> lbracket rbracket <<
+  $$ = JSONValue::Array(Vec::new());
+>>
+Array -> lbracket Values rbracket <<
+  $$ = $2;
+>>
+Values -> Value <<
+  $$ = $1;
+>>
+Values -> Values comma Value <<
+  let mut arr = $1;
+  if let JSONValue::Array(elems) = &mut arr {
+      elems.push($3);
+  }
+  $$ = arr;
+>>

spec/json_types.c

@@ -0,0 +1,64 @@
+#include "json_types.h"
+#include <string.h>
+#include <stdlib.h>
+#include "testutils.h"
+
+JSONValue * JSONValue_new(size_t id)
+{
+    JSONValue * jv = (JSONValue *)calloc(1, sizeof(JSONValue));
+    jv->id = id;
+    return jv;
+}
+
+JSONValue * JSONObject_new(void)
+{
+    JSONValue * jv = JSONValue_new(JSON_OBJECT);
+    jv->object.size = 0u;
+    return jv;
+}
+
+void JSONObject_append(JSONValue * object, char const * name, JSONValue * value)
+{
+    size_t const size = object->object.size;
+    for (size_t i = 0u; i < size; i++)
+    {
+        if (strcmp(name, object->object.entries[i].name) == 0)
+        {
+            object->object.entries[i].value = value;
+            return;
+        }
+    }
+    size_t const new_size = size + 1;
+    JSONObjectEntry * new_entries = (JSONObjectEntry *)malloc(sizeof(object->object.entries[0]) * new_size);
+    if (size > 0)
+    {
+        memcpy(new_entries, object->object.entries, size * sizeof(object->object.entries[0]));
+        free(object->object.entries);
+    }
+    object->object.entries = new_entries;
+    object->object.entries[size].name = name;
+    object->object.entries[size].value = value;
+    object->object.size = new_size;
+}
+
+JSONValue * JSONArray_new(void)
+{
+    JSONValue * jv = JSONValue_new(JSON_ARRAY);
+    jv->array.size = 0u;
+    return jv;
+}
+
+void JSONArray_append(JSONValue * array, JSONValue * value)
+{
+    size_t const size = array->array.size;
+    size_t const new_size = size + 1;
+    JSONValue ** new_entries = (JSONValue **)malloc(sizeof(JSONValue *) * new_size);
+    if (array->array.size > 0)
+    {
+        memcpy(new_entries, array->array.entries, sizeof(JSONValue *) * size);
+        free(array->array.entries);
+    }
+    array->array.entries = new_entries;
+    array->array.entries[size] = value;
+    array->array.size = new_size;
+}

spec/json_types.d

@@ -0,0 +1,63 @@
+class JSONValue
+{
+}
+
+class JSONObject : JSONValue
+{
+    JSONValue[string] value;
+
+    this()
+    {
+    }
+
+    this(JSONValue[string] value)
+    {
+        this.value = value;
+    }
+}
+
+class JSONArray : JSONValue
+{
+    JSONValue[] value;
+
+    this()
+    {
+    }
+
+    this(JSONValue[] value)
+    {
+        this.value = value;
+    }
+}
+
+class JSONNumber : JSONValue
+{
+    double value;
+
+    this(double value)
+    {
+        this.value = value;
+    }
+}
+
+class JSONString : JSONValue
+{
+    string value;
+
+    this(string value)
+    {
+        this.value = value;
+    }
+}
+
+class JSONTrue : JSONValue
+{
+}
+
+class JSONFalse : JSONValue
+{
+}
+
+class JSONNull : JSONValue
+{
+}

spec/json_types.h

@@ -0,0 +1,48 @@
+#pragma once
+
+#include <stddef.h>
+#include "testutils.h"
+
+#define JSON_OBJECT 0u
+#define JSON_ARRAY 1u
+#define JSON_NUMBER 2u
+#define JSON_STRING 3u
+#define JSON_TRUE 4u
+#define JSON_FALSE 5u
+#define JSON_NULL 6u
+
+typedef struct JSONObjectEntry_s
+{
+    char const * name;
+    struct JSONValue_s * value;
+} JSONObjectEntry;
+
+typedef struct JSONValue_s
+{
+    size_t id;
+    union
+    {
+        struct
+        {
+            size_t size;
+            JSONObjectEntry * entries;
+        } object;
+        struct
+        {
+            size_t size;
+            struct JSONValue_s ** entries;
+        } array;
+        double number;
+        str_t string;
+    };
+} JSONValue;
+
+JSONValue * JSONValue_new(size_t id);
+
+JSONValue * JSONObject_new(void);
+
+void JSONObject_append(JSONValue * object, char const * name, JSONValue * value);
+
+JSONValue * JSONArray_new(void);
+
+void JSONArray_append(JSONValue * array, JSONValue * value);

spec/macros.c.propane

@@ -0,0 +1,25 @@
+<<
+#include <stdlib.h>
+size_t mylexfn(p_context_t * context, p_token_info_t * out_token_info);
+void record(int v);
+>>
+
+ptype int;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lbrace /\{/;
+token rbrace /\}/;
+token plus /\+/;
+token macro;
+token macroname /@[a-zA-Z_]\w*/;
+token num /\d+/ << char b[100]; memcpy(b, match_text, match_length); b[match_length] = '\0'; $$ = atoi(b); >>
+
+Start -> Statements;
+Statements -> ;
+Statements -> Statement Statements;
+Statement -> Add;
+Statement -> MacroStart;
+Add -> num plus num << $$ = $1 + $3; record($$); >>
+MacroStart -> macro macroname lbrace;

spec/macros.d.propane

@@ -0,0 +1,31 @@
+<<
+import test_macros;
+>>
+
+ptype int;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lbrace /\{/;
+token rbrace /\}/;
+token plus /\+/;
+token macro;
+token macroname /@[a-zA-Z_]\w*/;
+token num /\d+/ <<
+    int n = 0;
+    foreach (c; match_text)
+    {
+        n *= 10;
+        n += (c - '0');
+    }
+    $$ = n;
+>>
+
+Start -> Statements;
+Statements -> ;
+Statements -> Statement Statements;
+Statement -> Add;
+Statement -> MacroStart;
+Add -> num plus num << $$ = $1 + $3; record($$); >>
+MacroStart -> macro macroname lbrace;

spec/macros.rust.propane

@@ -0,0 +1,80 @@
+<<
+fn mylexfn(context: &mut p_context_t, out_token_info: &mut p_token_info_t) -> usize {
+    loop {
+        if context.expanding {
+            let ei = context.expand_i;
+            context.expand_i += 1;
+            if context.expand_i >= context.token_infos.len() {
+                context.expanding = false;
+            }
+            *out_token_info = context.token_infos[ei].clone();
+            return P_SUCCESS;
+        }
+
+        let lex_result = p_lex(context, out_token_info);
+        if lex_result != P_SUCCESS {
+            return lex_result;
+        }
+
+        if out_token_info.token == TOKEN_macro {
+            context.defining = true;
+        } else if out_token_info.token == TOKEN_macroname {
+            if !context.defining {
+                context.expanding = true;
+                context.expand_i = 0;
+                continue;
+            }
+        } else if out_token_info.token == TOKEN_lbrace {
+            if context.defining {
+                /* Capture the macro body tokens (up to the closing '}'). */
+                let mut infos: Vec<p_token_info_t> = Vec::new();
+                loop {
+                    let mut ti = p_token_info_t::default();
+                    assert_eq!(P_SUCCESS, p_lex(context, &mut ti));
+                    if ti.token == TOKEN_rbrace {
+                        break;
+                    }
+                    infos.push(ti);
+                }
+                context.token_infos = infos;
+                context.defining = false;
+            }
+        } else {
+            context.defining = false;
+        }
+        return lex_result;
+    }
+}
+>>
+
+context_user_fields <<
+    pub defining: bool,
+    pub expanding: bool,
+    pub expand_i: usize,
+    pub token_infos: Vec<p_token_info_t>,
+    pub nums: Vec<i64>,
+>>
+
+ptype i64;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lbrace /\{/;
+token rbrace /\}/;
+token plus /\+/;
+token macro;
+token macroname /@[a-zA-Z_]\w*/;
+token num /\d+/ <<
+  let mut v: i64 = 0;
+  for c in match_text { v = v * 10 + (*c - b'0') as i64; }
+  $$ = v;
+>>
+
+Start -> Statements;
+Statements -> ;
+Statements -> Statement Statements;
+Statement -> Add;
+Statement -> MacroStart;
+Add -> num plus num << $$ = $1 + $3; ${context.nums}.push($$); >>
+MacroStart -> macro macroname lbrace;

spec/parse_inner_nested.c.propane

@@ -0,0 +1,19 @@
+<<
+#include <stdlib.h>
+#include <string.h>
+size_t mylexfn(p_context_t * context, p_token_info_t * out_token_info);
+>>
+
+ptype int;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lparen /\(/;
+token rparen /\)/;
+token plus /\+/;
+token num /\d+/ << char b[32]; memcpy(b, match_text, match_length); b[match_length] = '\0'; $$ = atoi(b); >>
+
+Start -> Expr << $$ = $1; >>
+Expr -> num << $$ = $1; >>
+Expr -> Expr plus num << $$ = $1 + $3; >>

spec/parse_inner_nested.d.propane

@@ -0,0 +1,17 @@
+<<
+import test_parse_inner_nested;
+>>
+
+ptype int;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lparen /\(/;
+token rparen /\)/;
+token plus /\+/;
+token num /\d+/ << int n = 0; foreach (ch; match_text) { n *= 10; n += (ch - '0'); } $$ = n; >>
+
+Start -> Expr << $$ = $1; >>
+Expr -> num << $$ = $1; >>
+Expr -> Expr plus num << $$ = $1 + $3; >>

spec/parse_inner_nested.rust.propane

@@ -0,0 +1,41 @@
+<<
+fn mylexfn(context: &mut p_context_t, out_token_info: &mut p_token_info_t) -> usize {
+    let result = p_lex(context, out_token_info);
+    if result != P_SUCCESS {
+        return result;
+    }
+    if out_token_info.token == TOKEN_lparen {
+        /* Reentrant nested parse of the parenthesized sub-expression. */
+        let inner_result = p_parse_inner_Start(context, &[TOKEN_rparen]);
+        if inner_result != P_SUCCESS {
+            return inner_result;
+        }
+        let value = p_result_Start(context);
+        /* p_parse_inner rewound the input so ')' was not consumed; consume it. */
+        let mut rparen_info = p_token_info_t::default();
+        assert_eq!(P_SUCCESS, p_lex(context, &mut rparen_info));
+        assert_eq!(TOKEN_rparen, rparen_info.token);
+        out_token_info.token = TOKEN_num;
+        out_token_info.pvalue = p_value(value);
+    }
+    P_SUCCESS
+}
+>>
+
+ptype i64;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lparen /\(/;
+token rparen /\)/;
+token plus /\+/;
+token num /\d+/ <<
+  let mut v: i64 = 0;
+  for c in match_text { v = v * 10 + (*c - b'0') as i64; }
+  $$ = v;
+>>
+
+Start -> Expr << $$ = $1; >>
+Expr -> num << $$ = $1; >>
+Expr -> Expr plus num << $$ = $1 + $3; >>

spec/parse_inner_nested_tree.c.propane

@@ -0,0 +1,17 @@
+<<
+size_t mylexfn(p_context_t * context, p_token_info_t * out_token_info);
+>>
+
+tree;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lparen /\(/;
+token rparen /\)/;
+token plus /\+/;
+token num /\d+/;
+
+Start -> Expr;
+Expr -> num;
+Expr -> Expr plus num;

spec/parse_inner_nested_tree.d.propane

@@ -0,0 +1,17 @@
+<<
+import test_parse_inner_nested_tree;
+>>
+
+tree;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lparen /\(/;
+token rparen /\)/;
+token plus /\+/;
+token num /\d+/;
+
+Start -> Expr;
+Expr -> num;
+Expr -> Expr plus num;

spec/parse_inner_nested_tree.rust.propane

@@ -0,0 +1,44 @@
+<<
+fn mylexfn(context: &mut p_context_t, out_token_info: &mut p_token_info_t) -> usize {
+    let result = p_lex(context, out_token_info);
+    if result != P_SUCCESS {
+        return result;
+    }
+    if out_token_info.token == TOKEN_lparen {
+        let start_position = out_token_info.position;
+        let inner_result = p_parse_inner_Start(context, &[TOKEN_rparen]);
+        if inner_result != P_SUCCESS {
+            return inner_result;
+        }
+        /* Read the inner subtree's span before re-borrowing context to lex. */
+        let inner = p_result_Start(context);
+        assert!(inner.valid());
+        let inner_start_col = inner.position().col;
+        let inner_end_col = inner.end_position().col;
+        let mut rparen_info = p_token_info_t::default();
+        assert_eq!(P_SUCCESS, p_lex(context, &mut rparen_info));
+        assert_eq!(TOKEN_rparen, rparen_info.token);
+        assert_eq!(start_position.col + 1, inner_start_col);
+        assert_eq!(rparen_info.position.col - 1, inner_end_col);
+        /* Synthesize a num token spanning the whole "( ... )" group. */
+        out_token_info.token = TOKEN_num;
+        out_token_info.position = start_position;
+        out_token_info.end_position = rparen_info.end_position;
+    }
+    P_SUCCESS
+}
+>>
+
+tree;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token lparen /\(/;
+token rparen /\)/;
+token plus /\+/;
+token num /\d+/;
+
+Start -> Expr;
+Expr -> num;
+Expr -> Expr plus num;

spec/propane/code_point_range_spec.rb

@@ -0,0 +1,87 @@
+class Propane
+  describe CodePointRange do
+
+    describe "#<=>" do
+      it "sorts ranges" do
+        arr = [
+          CodePointRange.new(100,102),
+          CodePointRange.new(65, 68),
+          CodePointRange.new(65, 65),
+          CodePointRange.new(100, 100),
+          CodePointRange.new(68, 70),
+        ]
+        arr.sort!
+        expect(arr[0]).to eq CodePointRange.new(65, 65)
+        expect(arr[1]).to eq CodePointRange.new(65, 68)
+        expect(arr[2]).to eq CodePointRange.new(68, 70)
+        expect(arr[3]).to eq CodePointRange.new(100, 100)
+        expect(arr[4]).to eq CodePointRange.new(100, 102)
+      end
+    end
+
+    describe "#include?" do
+      it "returns whether the code point is included in the range" do
+        expect(CodePointRange.new(100).include?(100)).to be_truthy
+        expect(CodePointRange.new(100, 100).include?(99)).to be_falsey
+        expect(CodePointRange.new(100, 100).include?(101)).to be_falsey
+        expect(CodePointRange.new(100, 120).include?(99)).to be_falsey
+        expect(CodePointRange.new(100, 120).include?(100)).to be_truthy
+        expect(CodePointRange.new(100, 120).include?(110)).to be_truthy
+        expect(CodePointRange.new(100, 120).include?(120)).to be_truthy
+        expect(CodePointRange.new(100, 120).include?(121)).to be_falsey
+      end
+
+      it "returns whether the range is included in the range" do
+        expect(CodePointRange.new(100).include?(CodePointRange.new(100))).to be_truthy
+        expect(CodePointRange.new(100, 100).include?(CodePointRange.new(99))).to be_falsey
+        expect(CodePointRange.new(100, 100).include?(CodePointRange.new(99, 100))).to be_falsey
+        expect(CodePointRange.new(100, 120).include?(CodePointRange.new(90, 110))).to be_falsey
+        expect(CodePointRange.new(100, 120).include?(CodePointRange.new(110, 130))).to be_falsey
+        expect(CodePointRange.new(100, 120).include?(CodePointRange.new(100, 120))).to be_truthy
+        expect(CodePointRange.new(100, 120).include?(CodePointRange.new(100, 110))).to be_truthy
+        expect(CodePointRange.new(100, 120).include?(CodePointRange.new(110, 120))).to be_truthy
+        expect(CodePointRange.new(100, 120).include?(CodePointRange.new(102, 118))).to be_truthy
+      end
+    end
+
+    describe ".invert_ranges" do
+      it "inverts ranges" do
+        expect(CodePointRange.invert_ranges(
+          [CodePointRange.new(60, 90),
+           CodePointRange.new(80, 85),
+           CodePointRange.new(80, 100),
+           CodePointRange.new(101),
+           CodePointRange.new(200, 300)])).to eq [
+             CodePointRange.new(0, 59),
+             CodePointRange.new(102, 199),
+             CodePointRange.new(301, 0x7FFFFFFF)]
+        expect(CodePointRange.invert_ranges(
+          [CodePointRange.new(0, 500),
+           CodePointRange.new(7000, 0x7FFFFFFF)])).to eq [
+             CodePointRange.new(501, 6999)]
+      end
+    end
+
+    describe ".first_subrange" do
+      it "returns the first subrange to split" do
+        expect(CodePointRange.first_subrange(
+          [CodePointRange.new(65, 90),
+           CodePointRange.new(66, 66),
+           CodePointRange.new(80, 90)])).to eq CodePointRange.new(65)
+        expect(CodePointRange.first_subrange(
+          [CodePointRange.new(65, 90)])).to eq CodePointRange.new(65, 90)
+        expect(CodePointRange.first_subrange(
+          [CodePointRange.new(65, 90),
+           CodePointRange.new(80, 90)])).to eq CodePointRange.new(65, 79)
+        expect(CodePointRange.first_subrange(
+          [CodePointRange.new(65, 90),
+           CodePointRange.new(65, 100),
+           CodePointRange.new(65, 95)])).to eq CodePointRange.new(65, 90)
+        expect(CodePointRange.first_subrange(
+          [CodePointRange.new(100, 120),
+           CodePointRange.new(70, 90)])).to eq CodePointRange.new(70, 90)
+      end
+    end
+
+  end
+end

spec/propane/grammar_spec.rb

@@ -0,0 +1,266 @@
+class Propane
+  describe Grammar do
+    it "parses a user grammar" do
+      input = <<EOF
+# Comment line
+
+module a.b;
+ptype   XYZ  *  ;
+
+token while;
+
+token id
+  /[a-zA-Z_][a-zA-Z_0-9]*/;
+
+token token_with_code <<
+Code for the token
+>>
+
+tokenid token_with_no_pattern;
+
+drop /\\s+/;
+
+A -> B <<
+  a = 42;
+>>
+B -> C while id;
+B -> <<
+  b = 0;
+>>
+EOF
+      grammar = Grammar.new(input, "test.propane")
+      expect(grammar.modulename).to eq "a.b"
+      expect(grammar.ptype).to eq "XYZ  *"
+      expect(grammar.ptypes).to eq("default" => "XYZ  *")
+      expect(grammar.prefix).to eq "p_"
+
+      o = grammar.tokens.find {|token| token.name == "while"}
+      expect(o).to_not be_nil
+      expect(o.line_number).to eq 6
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.pattern).to eq "while"
+      expect(o.line_number).to eq 6
+      expect(o.code).to be_nil
+
+      o = grammar.tokens.find {|token| token.name == "id"}
+      expect(o).to_not be_nil
+      expect(o.line_number).to eq 9
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.pattern).to eq "[a-zA-Z_][a-zA-Z_0-9]*"
+      expect(o.line_number).to eq 9
+      expect(o.code).to be_nil
+
+      o = grammar.tokens.find {|token| token.name == "token_with_code"}
+      expect(o).to_not be_nil
+      expect(o.line_number).to eq 11
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.pattern).to eq "token_with_code"
+      expect(o.line_number).to eq 11
+      expect(o.code).to eq %[#line 12 "test.propane"\nCode for the token\n#linereset\n]
+
+      o = grammar.tokens.find {|token| token.name == "token_with_no_pattern"}
+      expect(o).to_not be_nil
+      expect(o.line_number).to eq 15
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to be_nil
+
+      o = grammar.patterns.find {|pattern| pattern.pattern == "\\s+"}
+      expect(o).to_not be_nil
+      expect(o.line_number).to eq 17
+      expect(o.token).to be_nil
+      expect(o.code).to be_nil
+
+      expect(grammar.rules.size).to eq 3
+
+      o = grammar.rules[0]
+      expect(o.name).to eq "A"
+      expect(o.components).to eq %w[B]
+      expect(o.line_number).to eq 19
+      expect(o.code).to eq %[#line 20 "test.propane"\n  a = 42;\n#linereset\n]
+
+      o = grammar.rules[1]
+      expect(o.name).to eq "B"
+      expect(o.components).to eq %w[C while id]
+      expect(o.line_number).to eq 22
+      expect(o.code).to be_nil
+
+      o = grammar.rules[2]
+      expect(o.name).to eq "B"
+      expect(o.components).to eq []
+      expect(o.line_number).to eq 23
+      expect(o.code).to eq %[#line 24 "test.propane"\n  b = 0;\n#linereset\n]
+    end
+
+    it "parses code segments with semicolons" do
+      input = <<EOF
+token code1 <<
+  a = b;
+  return c;
+>>
+
+token code2 <<
+  writeln("Hello there");
+>>
+
+tokenid token_with_no_pattern;
+
+prefix myparser_;
+EOF
+      grammar = Grammar.new(input, "test.propane")
+      expect(grammar.prefix).to eq "myparser_"
+
+      o = grammar.tokens.find {|token| token.name == "code1"}
+      expect(o).to_not be_nil
+      expect(o.line_number).to eq 1
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.code).to eq %[#line 2 "test.propane"\n  a = b;\n  return c;\n#linereset\n]
+
+      o = grammar.tokens.find {|token| token.name == "code2"}
+      expect(o).to_not be_nil
+      expect(o.line_number).to eq 6
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.code).to eq %[#line 7 "test.propane"\n  writeln("Hello there");\n#linereset\n]
+    end
+
+    it "does not emit #line directives with noline statement" do
+      input = <<EOF
+noline;
+
+token code1 <<
+  a = b;
+  return c;
+>>
+
+token code2 <<
+  writeln("Hello there");
+>>
+
+tokenid token_with_no_pattern;
+
+prefix myparser_;
+EOF
+      grammar = Grammar.new(input, "test.propane")
+      expect(grammar.prefix).to eq "myparser_"
+
+      o = grammar.tokens.find {|token| token.name == "code1"}
+      expect(o).to_not be_nil
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.code).to eq %[\n  a = b;\n  return c;]
+
+      o = grammar.tokens.find {|token| token.name == "code2"}
+      expect(o).to_not be_nil
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.code).to eq %[\n  writeln("Hello there");]
+    end
+
+    it "supports mode labels" do
+      input = <<EOF
+token a;
+m1: token b;
+/foo/ <<
+>>
+m2: /bar/ <<
+>>
+drop /q/;
+m3: drop /r/;
+EOF
+      grammar = Grammar.new(input, "test.propane")
+
+      o = grammar.tokens.find {|token| token.name == "a"}
+      expect(o).to_not be_nil
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.modes).to be_empty
+
+      o = grammar.tokens.find {|token| token.name == "b"}
+      expect(o).to_not be_nil
+
+      o = grammar.patterns.find {|pattern| pattern.token == o}
+      expect(o).to_not be_nil
+      expect(o.modes).to eq Set["m1"]
+
+      o = grammar.patterns.find {|pattern| pattern.pattern == "foo"}
+      expect(o).to_not be_nil
+      expect(o.modes).to be_empty
+
+      o = grammar.patterns.find {|pattern| pattern.pattern == "bar"}
+      expect(o).to_not be_nil
+      expect(o.modes).to eq Set["m2"]
+
+      o = grammar.patterns.find {|pattern| pattern.pattern == "q"}
+      expect(o).to_not be_nil
+      expect(o.modes).to be_empty
+
+      o = grammar.patterns.find {|pattern| pattern.pattern == "r"}
+      expect(o).to_not be_nil
+      expect(o.modes).to eq Set["m3"]
+    end
+
+    it "allows assigning ptypes to tokens and rules" do
+      input = <<EOF
+ptype Subnode *;
+ptype string = char *;
+ptype integer = int;
+ptype node = Node *;
+
+token abc(string);
+token bar;
+tokenid int(integer);
+
+/xyz/ (string) <<
+>>
+
+/z28/ <<
+>>
+
+Start (node) -> R;
+R -> abc int;
+EOF
+      grammar = Grammar.new(input, "test.propane")
+
+      o = grammar.tokens.find {|token| token.name == "abc"}
+      expect(o).to_not be_nil
+      expect(o.ptypename).to eq "string"
+
+      o = grammar.tokens.find {|token| token.name == "bar"}
+      expect(o).to_not be_nil
+      expect(o.ptypename).to be_nil
+
+      o = grammar.tokens.find {|token| token.name == "int"}
+      expect(o).to_not be_nil
+      expect(o.ptypename).to eq "integer"
+
+      o = grammar.rules.find {|rule| rule.name == "Start"}
+      expect(o).to_not be_nil
+      expect(o.ptypename).to eq "node"
+
+      o = grammar.rules.find {|rule| rule.name == "R"}
+      expect(o).to_not be_nil
+      expect(o.ptypename).to be_nil
+
+      o = grammar.patterns.find {|pattern| pattern.pattern == "xyz"}
+      expect(o).to_not be_nil
+      expect(o.ptypename).to eq "string"
+
+      o = grammar.patterns.find {|pattern| pattern.pattern == "z28"}
+      expect(o).to_not be_nil
+      expect(o.ptypename).to be_nil
+    end
+  end
+end

spec/propane/lexer/dfa_spec.rb

@@ -0,0 +1,199 @@
+class TestLexer
+  def initialize(token_dfa)
+    @token_dfa = token_dfa
+  end
+
+  def lex(input)
+    input_chars = input.chars
+    output = []
+    while lexed_token = lex_token(input_chars)
+      output << lexed_token
+      input_chars.slice!(0, lexed_token[1].size)
+    end
+    unless input_chars.empty?
+      raise "Unmatched input #{input_chars.join(" ")}"
+    end
+    output
+  end
+
+  def lex_token(input_chars)
+    return nil if input_chars.empty?
+    s = ""
+    current_state = @token_dfa.start_state
+    last_accepts = nil
+    last_s = nil
+    input_chars.each_with_index do |input_char, index|
+      if next_state = transition(current_state, input_char)
+        s += input_char
+        current_state = next_state
+        if current_state.accepts
+          last_accepts = current_state.accepts
+          last_s = s
+        end
+      else
+        break
+      end
+    end
+    if last_accepts
+      name = last_accepts.token ? last_accepts.token.name : nil
+      [name, last_s]
+    end
+  end
+
+  def transition(state, input_char)
+    state.transitions.each do |transition|
+      if transition.code_point_range.include?(input_char.ord)
+        return transition.destination
+      end
+    end
+    nil
+  end
+end
+
+def run(grammar, input)
+  grammar = Propane::Grammar.new(grammar, "test.propane")
+  token_dfa = Propane::Lexer::DFA.new(grammar.patterns)
+  test_lexer = TestLexer.new(token_dfa)
+  test_lexer.lex(input)
+end
+
+describe Propane::Lexer::DFA do
+  it "lexes a simple token" do
+    expect(run(<<EOF, "foo")).to eq [["foo", "foo"]]
+token foo;
+EOF
+  end
+
+  it "lexes two tokens" do
+    expected = [
+      ["foo", "foo"],
+      ["bar", "bar"],
+    ]
+    expect(run(<<EOF, "foobar")).to eq expected
+token foo;
+token bar;
+EOF
+  end
+
+  it "lexes the longer of multiple options" do
+    expected = [
+      ["identifier", "foobar"],
+    ]
+    expect(run(<<EOF, "foobar")).to eq expected
+token foo;
+token bar;
+token identifier /[a-z]+/;
+EOF
+    expected = [
+      ["plusplus", "++"],
+      ["plus", "+"],
+    ]
+    expect(run(<<EOF, "+++")).to eq expected
+token plus /\\+/;
+token plusplus /\\+\\+/;
+EOF
+  end
+
+  it "lexes whitespace" do
+    expected = [
+      ["foo", "foo"],
+      ["WS", " \t"],
+      ["bar", "bar"],
+    ]
+    expect(run(<<EOF, "foo \tbar")).to eq expected
+token foo;
+token bar;
+token WS /\\s+/;
+EOF
+  end
+
+  it "allows dropping a matched pattern" do
+    expected = [
+      ["foo", "foo"],
+      [nil, " \t"],
+      ["bar", "bar"],
+    ]
+    expect(run(<<EOF, "foo \tbar")).to eq expected
+token foo;
+token bar;
+drop /\\s+/;
+EOF
+  end
+
+  it "matches a semicolon" do
+    expected = [
+      ["semicolon", ";"],
+    ]
+    expect(run(<<EOF, ";")).to eq expected
+token semicolon /;/;
+EOF
+  end
+
+  it "matches a negated character class" do
+    expected = [
+      ["pattern", "/abc/"],
+    ]
+    expect(run(<<EOF, "/abc/")).to eq expected
+token pattern /\\/[^\\s]*\\//;
+EOF
+  end
+
+  it "matches special character classes " do
+    expected = [
+      ["a", "abc123_FOO"],
+    ]
+    expect(run(<<EOF, "abc123_FOO")).to eq expected
+token a /\\w+/;
+EOF
+    expected = [
+      ["b", "FROG*%$#"],
+    ]
+    expect(run(<<EOF, "FROG*%$#")).to eq expected
+token b /FROG\\D{1,4}/;
+EOF
+    expected = [
+      ["c", "$883366"],
+    ]
+    expect(run(<<EOF, "$883366")).to eq expected
+token c /$\\d+/;
+EOF
+    expected = [
+      ["d", "^&$@"],
+    ]
+    expect(run(<<EOF, "^&$@")).to eq expected
+token d /^\\W+/;
+EOF
+    expected = [
+      ["a", "abc123_FOO"],
+      [nil, " "],
+      ["b", "FROG*%$#"],
+      [nil, " "],
+      ["c", "$883366"],
+      [nil, " "],
+      ["d", "^&$@"],
+    ]
+    expect(run(<<EOF, "abc123_FOO FROG*%$# $883366 ^&$@")).to eq expected
+token a /\\w+/;
+token b /FROG\\D{1,4}/;
+token c /$\\d+/;
+token d /^\\W+/;
+drop /\\s+/;
+EOF
+  end
+
+  it "matches a negated character class with a nested inner negated character class" do
+    expected = [
+      ["t", "$&*"],
+    ]
+    expect(run(<<EOF, "$&*")).to eq expected
+token t /[^%\\W]+/;
+EOF
+  end
+
+  it "\\s matches a newline" do
+    expected = [["s", "\n"]]
+    expect(run(<<EOF, "\n")).to eq expected
+token s /\\s/;
+EOF
+  end
+end

spec/propane/parser/item_spec.rb

@@ -0,0 +1,19 @@
+class Propane
+  class Parser
+
+    describe Item do
+
+      it "operates properly with a set" do
+        rule = Object.new
+        item1 = Item.new(rule, 2)
+        item2 = Item.new(rule, 2)
+        expect(item1).to eq item2
+        expect(item1.eql?(item2)).to be_truthy
+        set = Set.new([item1, item2])
+        expect(set.size).to eq 1
+      end
+
+    end
+
+  end
+end

spec/propane/regex_spec.rb

@@ -0,0 +1,350 @@
+class Propane
+  RSpec.describe Regex do
+
+    it "parses an empty expression" do
+      regex = Regex.new("", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0].size).to eq 0
+    end
+
+    it "parses a single character unit expression" do
+      regex = Regex.new("a", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses a group with a single character unit expression" do
+      regex = Regex.new("(a)", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::AlternatesUnit
+      alt_unit = seq_unit[0]
+      expect(alt_unit.alternates.size).to eq 1
+      expect(alt_unit.alternates[0]).to be_a Regex::SequenceUnit
+      expect(alt_unit.alternates[0][0]).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses a *" do
+      regex = Regex.new("a*", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::MultiplicityUnit
+      m_unit = seq_unit[0]
+      expect(m_unit.min_count).to eq 0
+      expect(m_unit.max_count).to be_nil
+      expect(m_unit.unit).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses a +" do
+      regex = Regex.new("a+", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::MultiplicityUnit
+      m_unit = seq_unit[0]
+      expect(m_unit.min_count).to eq 1
+      expect(m_unit.max_count).to be_nil
+      expect(m_unit.unit).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses a ?" do
+      regex = Regex.new("a?", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::MultiplicityUnit
+      m_unit = seq_unit[0]
+      expect(m_unit.min_count).to eq 0
+      expect(m_unit.max_count).to eq 1
+      expect(m_unit.unit).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses a multiplicity count" do
+      regex = Regex.new("a{5}", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::MultiplicityUnit
+      m_unit = seq_unit[0]
+      expect(m_unit.min_count).to eq 5
+      expect(m_unit.max_count).to eq 5
+      expect(m_unit.unit).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses a minimum-only multiplicity count" do
+      regex = Regex.new("a{5,}", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::MultiplicityUnit
+      m_unit = seq_unit[0]
+      expect(m_unit.min_count).to eq 5
+      expect(m_unit.max_count).to be_nil
+      expect(m_unit.unit).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses a minimum and maximum multiplicity count" do
+      regex = Regex.new("a{5,8}", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::MultiplicityUnit
+      m_unit = seq_unit[0]
+      expect(m_unit.min_count).to eq 5
+      expect(m_unit.max_count).to eq 8
+      expect(m_unit.unit).to be_a Regex::CharacterRangeUnit
+      expect(m_unit.unit.first).to eq "a".ord
+    end
+
+    it "parses an escaped *" do
+      regex = Regex.new("a\\*", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 2
+      expect(seq_unit[0]).to be_a Regex::CharacterRangeUnit
+      expect(seq_unit[0].first).to eq "a".ord
+      expect(seq_unit[1]).to be_a Regex::CharacterRangeUnit
+      expect(seq_unit[1].first).to eq "*".ord
+    end
+
+    it "parses an escaped +" do
+      regex = Regex.new("a\\+", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 2
+      expect(seq_unit[0]).to be_a Regex::CharacterRangeUnit
+      expect(seq_unit[0].first).to eq "a".ord
+      expect(seq_unit[1]).to be_a Regex::CharacterRangeUnit
+      expect(seq_unit[1].first).to eq "+".ord
+    end
+
+    it "parses an escaped \\" do
+      regex = Regex.new("\\\\d", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 2
+      expect(seq_unit[0]).to be_a Regex::CharacterRangeUnit
+      expect(seq_unit[0].first).to eq "\\".ord
+      expect(seq_unit[1]).to be_a Regex::CharacterRangeUnit
+      expect(seq_unit[1].first).to eq "d".ord
+    end
+
+    it "parses a character class" do
+      regex = Regex.new("[a-z_]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.negate).to be_falsey
+      expect(ccu.size).to eq 2
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq "a".ord
+      expect(ccu[0].last).to eq "z".ord
+      expect(ccu[1]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[1].first).to eq "_".ord
+    end
+
+    it "parses a negated character class" do
+      regex = Regex.new("[^xyz]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.negate).to be_truthy
+      expect(ccu.size).to eq 3
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq "x".ord
+    end
+
+    it "parses a negated character class with inner character classes" do
+      regex = Regex.new("[^x\\sz]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.negate).to be_truthy
+      expect(ccu.size).to eq 8
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq "x".ord
+      expect(ccu[1].first).to eq " ".ord
+      expect(ccu[7].first).to eq "z".ord
+    end
+
+    it "parses - as a plain character at beginning of a character class" do
+      regex = Regex.new("[-9]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.size).to eq 2
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq "-".ord
+    end
+
+    it "parses - as a plain character at end of a character class" do
+      regex = Regex.new("[0-]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.size).to eq 2
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq "0".ord
+      expect(ccu[1]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[1].first).to eq "-".ord
+    end
+
+    it "parses - as a plain character at beginning of a negated character class" do
+      regex = Regex.new("[^-9]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.negate).to be_truthy
+      expect(ccu.size).to eq 2
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq "-".ord
+    end
+
+    it "parses . as a plain character in a character class" do
+      regex = Regex.new("[.]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.negate).to be_falsey
+      expect(ccu.size).to eq 1
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq ".".ord
+    end
+
+    it "parses - as a plain character when escaped in middle of character class" do
+      regex = Regex.new("[0\\-9]", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      seq_unit = regex.unit.alternates[0]
+      expect(seq_unit.size).to eq 1
+      expect(seq_unit[0]).to be_a Regex::CharacterClassUnit
+      ccu = seq_unit[0]
+      expect(ccu.negate).to be_falsey
+      expect(ccu.size).to eq 3
+      expect(ccu[0]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[0].first).to eq "0".ord
+      expect(ccu[1]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[1].first).to eq "-".ord
+      expect(ccu[2]).to be_a Regex::CharacterRangeUnit
+      expect(ccu[2].first).to eq "9".ord
+    end
+
+    it "parses alternates" do
+      regex = Regex.new("ab|c", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 2
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[1]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[0].size).to eq 2
+      expect(regex.unit.alternates[1].size).to eq 1
+    end
+
+    it "parses a ." do
+      regex = Regex.new("a.b", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 1
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[0][0]).to be_a Regex::CharacterRangeUnit
+      expect(regex.unit.alternates[0][1]).to be_a Regex::CharacterClassUnit
+      expect(regex.unit.alternates[0][1].units.size).to eq 2
+      expect(regex.unit.alternates[0][2]).to be_a Regex::CharacterRangeUnit
+    end
+
+    it "parses something complex" do
+      regex = Regex.new("(a|)*|[^^]|\\|v|[x-y]+", 1)
+      expect(regex.unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates.size).to eq 4
+      expect(regex.unit.alternates[0]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[0].size).to eq 1
+      expect(regex.unit.alternates[0][0]).to be_a Regex::MultiplicityUnit
+      expect(regex.unit.alternates[0][0].min_count).to eq 0
+      expect(regex.unit.alternates[0][0].max_count).to be_nil
+      expect(regex.unit.alternates[0][0].unit).to be_a Regex::AlternatesUnit
+      expect(regex.unit.alternates[0][0].unit.alternates.size).to eq 2
+      expect(regex.unit.alternates[0][0].unit.alternates[0]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[0][0].unit.alternates[0].size).to eq 1
+      expect(regex.unit.alternates[0][0].unit.alternates[0][0]).to be_a Regex::CharacterRangeUnit
+      expect(regex.unit.alternates[0][0].unit.alternates[1]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[0][0].unit.alternates[1].size).to eq 0
+      expect(regex.unit.alternates[1]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[1].size).to eq 1
+      expect(regex.unit.alternates[1][0]).to be_a Regex::CharacterClassUnit
+      expect(regex.unit.alternates[1][0].negate).to be_truthy
+      expect(regex.unit.alternates[1][0].size).to eq 1
+      expect(regex.unit.alternates[1][0][0]).to be_a Regex::CharacterRangeUnit
+      expect(regex.unit.alternates[2]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[2].size).to eq 2
+      expect(regex.unit.alternates[2][0]).to be_a Regex::CharacterRangeUnit
+      expect(regex.unit.alternates[2][0].first).to eq "|".ord
+      expect(regex.unit.alternates[2][1]).to be_a Regex::CharacterRangeUnit
+      expect(regex.unit.alternates[2][1].first).to eq "v".ord
+      expect(regex.unit.alternates[3]).to be_a Regex::SequenceUnit
+      expect(regex.unit.alternates[3].size).to eq 1
+      expect(regex.unit.alternates[3][0]).to be_a Regex::MultiplicityUnit
+      expect(regex.unit.alternates[3][0].min_count).to eq 1
+      expect(regex.unit.alternates[3][0].max_count).to be_nil
+      expect(regex.unit.alternates[3][0].unit).to be_a Regex::CharacterClassUnit
+      expect(regex.unit.alternates[3][0].unit.size).to eq 1
+      expect(regex.unit.alternates[3][0].unit[0]).to be_a Regex::CharacterRangeUnit
+      expect(regex.unit.alternates[3][0].unit[0].first).to eq "x".ord
+      expect(regex.unit.alternates[3][0].unit[0].last).to eq "y".ord
+    end
+
+  end
+end

spec/rewind.c.propane

@@ -0,0 +1,23 @@
+<<
+#include <stdlib.h>
+#include <string.h>
+size_t mylexfn(p_context_t * context, p_token_info_t * out_token_info);
+void record(int value);
+>>
+
+ptype int;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token repeat /repeat/;
+token lbrace /\{/;
+token rbrace /\}/;
+token plus /\+/;
+token num /\d+/ << char b[32]; memcpy(b, match_text, match_length); b[match_length] = '\0'; $$ = atoi(b); >>
+
+Start -> Statements;
+Statements -> ;
+Statements -> Statement Statements;
+Statement -> Add;
+Add -> num plus num << record($1 + $3); >>

spec/rewind.d.propane

@@ -0,0 +1,20 @@
+<<
+import test_rewind;
+>>
+
+ptype int;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token repeat /repeat/;
+token lbrace /\{/;
+token rbrace /\}/;
+token plus /\+/;
+token num /\d+/ << int n = 0; foreach (ch; match_text) { n *= 10; n += (ch - '0'); } $$ = n; >>
+
+Start -> Statements;
+Statements -> ;
+Statements -> Statement Statements;
+Statement -> Add;
+Add -> num plus num << record($1 + $3); >>

spec/rewind.rust.propane

@@ -0,0 +1,67 @@
+<<
+fn mylexfn(context: &mut p_context_t, out_token_info: &mut p_token_info_t) -> usize {
+    loop {
+        let result = p_lex(context, out_token_info);
+        if result != P_SUCCESS {
+            return result;
+        }
+        if out_token_info.token == TOKEN_repeat {
+            let mut count_info = p_token_info_t::default();
+            assert_eq!(P_SUCCESS, p_lex(context, &mut count_info));
+            assert_eq!(TOKEN_num, count_info.token);
+            let mut brace_info = p_token_info_t::default();
+            assert_eq!(P_SUCCESS, p_lex(context, &mut brace_info));
+            assert_eq!(TOKEN_lbrace, brace_info.token);
+            context.remaining = p_value_get(&count_info.pvalue);
+            context.body_index = p_input_index(context);
+            context.body_position = p_position(context);
+            continue;
+        }
+        if out_token_info.token == TOKEN_rbrace {
+            if context.remaining > 1 {
+                context.remaining -= 1;
+                let bi = context.body_index;
+                let bp = context.body_position;
+                p_set_input_index(context, bi);
+                p_set_position(context, bp);
+                continue;
+            }
+            context.remaining = 0;
+            continue;
+        }
+        if out_token_info.token == TOKEN_num {
+            context.num_cols.push(out_token_info.position.col);
+        }
+        return result;
+    }
+}
+>>
+
+context_user_fields <<
+    pub nums: Vec<i64>,
+    pub num_cols: Vec<u32>,
+    pub remaining: i64,
+    pub body_index: usize,
+    pub body_position: p_position_t,
+>>
+
+ptype i64;
+
+lex_fn mylexfn;
+
+drop /\s+/;
+token repeat /repeat/;
+token lbrace /\{/;
+token rbrace /\}/;
+token plus /\+/;
+token num /\d+/ <<
+  let mut v: i64 = 0;
+  for c in match_text { v = v * 10 + (*c - b'0') as i64; }
+  $$ = v;
+>>
+
+Start -> Statements;
+Statements -> ;
+Statements -> Statement Statements;
+Statement -> Add;
+Add -> num plus num << ${context.nums}.push($1 + $3); >>

spec/spec_helper.rb

@@ -0,0 +1,36 @@
+unless ENV["dist_specs"]
+  require "bundler/setup"
+  require "simplecov"
+
+  class MyFormatter
+    def format(*args)
+    end
+  end
+  SimpleCov.start do
+    skip "/spec/"
+    skip "/.bundle/"
+    if ENV["partial_specs"]
+      command_name "RSpec-partial"
+    else
+      command_name "RSpec"
+    end
+    project_name "Propane"
+    # Keep this process's results separate from the propane subprocess results
+    # so that nothing has to merge on the fly; the spec Rake task collates all
+    # of the parts once the suite is done.
+    coverage_dir "coverage/parts/rspec"
+    merge_timeout 3600
+    formatter(MyFormatter)
+  end
+
+  RSpec.configure do |config|
+    # Enable flags like --only-failures and --next-failure
+    config.example_status_persistence_file_path = ".rspec_status"
+
+    config.expect_with :rspec do |c|
+      c.syntax = :expect
+    end
+  end
+end
+
+require "propane"

spec/test_basic_math_grammar.c

@@ -0,0 +1,33 @@
+#include "testparser.h"
+#include "testutils.h"
+#include <string.h>
+
+int main()
+{
+    char const * input = "1 + 2 * 3 + 4";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(11, p_result(context));
+    p_context_delete(context);
+
+    input = "1 * 2 ** 4 * 3";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(48, p_result(context));
+    p_context_delete(context);
+
+    input = "(1 + 2) * 3 + 4";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(13, p_result(context));
+    p_context_delete(context);
+
+    input = "(2 * 2) ** 3 + 4 + 5";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(73, p_result(context));
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_basic_math_grammar.d

@@ -0,0 +1,32 @@
+import testparser;
+import std.stdio;
+import testutils;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = "1 + 2 * 3 + 4";
+    p_context_t * context;
+    context = p_context_new(input);
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(11, p_result(context));
+
+    input = "1 * 2 ** 4 * 3";
+    context = p_context_new(input);
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(48, p_result(context));
+
+    input = "(1 + 2) * 3 + 4";
+    context = p_context_new(input);
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(13, p_result(context));
+
+    input = "(2 * 2) ** 3 + 4 + 5";
+    context = p_context_new(input);
+    assert_eq(P_SUCCESS, p_parse(context));
+    assert_eq(73, p_result(context));
+}

spec/test_basic_math_grammar.rs

@@ -0,0 +1,16 @@
+use testparser::*;
+
+fn main() {
+    let cases: [(&[u8], u64); 4] = [
+        (b"1 + 2 * 3 + 4", 11),
+        (b"1 * 2 ** 4 * 3", 48),
+        (b"(1 + 2) * 3 + 4", 13),
+        (b"(2 * 2) ** 3 + 4 + 5", 73),
+    ];
+    for (input, expected) in cases {
+        let mut context = p_context_new(input);
+        assert_eq!(P_SUCCESS, p_parse(&mut context));
+        assert_eq!(expected, p_result(&context));
+        p_context_delete(context);
+    }
+}

spec/test_context_field_handle.c

@@ -0,0 +1,39 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+#include "testutils.h"
+
+int main()
+{
+    char input[128];
+    size_t i;
+    p_context_t * context;
+    Token token;
+
+    /* Enough tokens that the tree node arena is reallocated during the parse. */
+    memset(input, 0, sizeof(input));
+    for (i = 0u; i < 40u; i++)
+    {
+        input[i] = 'a';
+    }
+
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert_eq(P_SUCCESS, p_parse(context));
+
+    /* The handle was stored in a context user field during the parse, before
+     * the remaining nodes were created. It still refers to the same node. */
+    assert_eq(1u, context->have_first);
+    assert(p_node_valid(context->first_item));
+    token = p_Item_pToken1(context->first_item);
+    assert(p_node_valid(token));
+    assert_eq(TOKEN_a, p_Token_token(token));
+    assert_eq(7u, p_Token_pvalue(token));
+
+    /* The stored handle refers to the first Item, which starts at column 1. */
+    assert_eq(1u, p_node_position(context->first_item).row);
+    assert_eq(1u, p_node_position(context->first_item).col);
+
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_context_field_handle.cpp

@@ -0,0 +1,36 @@
+#include "testparser.h"
+#include <cassert>
+#include <cstring>
+#include "testutils.h"
+
+int main()
+{
+    char input[128];
+
+    /* Enough tokens that the tree node arena is reallocated during the parse. */
+    memset(input, 0, sizeof(input));
+    for (size_t i = 0u; i < 40u; i++)
+    {
+        input[i] = 'a';
+    }
+
+    p_context_t * context = p_context_new((uint8_t const *)input, strlen(input));
+    assert_eq(P_SUCCESS, p_parse(context));
+
+    /* The handle was stored in a context user field during the parse, before
+     * the remaining nodes were created. It still refers to the same node. */
+    assert_eq(1u, context->have_first);
+    assert(context->first_item.valid());
+    Token token = context->first_item.pToken1();
+    assert(token.valid());
+    assert_eq(TOKEN_a, token.token());
+    assert_eq(7u, token.pvalue());
+
+    /* The stored handle refers to the first Item, which starts at column 1. */
+    assert_eq(1u, context->first_item.position().row);
+    assert_eq(1u, context->first_item.position().col);
+
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_context_field_handle.d

@@ -0,0 +1,35 @@
+import testparser;
+import testutils;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    /* Enough tokens that the tree node array is reallocated during the parse. */
+    string input;
+    foreach (i; 0 .. 40)
+    {
+        input ~= "a";
+    }
+
+    p_context_t * context = p_context_new(input);
+    assert_eq(P_SUCCESS, p_parse(context));
+
+    /* The handle was stored in a context user field during the parse, before
+     * the remaining nodes were created. It still refers to the same node. */
+    assert_eq(1, context.have_first);
+    assert(context.first_item.valid);
+    Token token = context.first_item.pToken1;
+    assert(token.valid);
+    assert_eq(TOKEN_a, token.token);
+    assert_eq(7, token.pvalue);
+
+    /* The stored handle refers to the first Item, which starts at column 1. */
+    assert_eq(1u, context.first_item.position.row);
+    assert_eq(1u, context.first_item.position.col);
+
+    p_context_delete(context);
+}

spec/test_custom_lex_fn.c

@@ -0,0 +1,15 @@
+#include "testparser.h"
+#include "testutils.h"
+#include <string.h>
+
+int main()
+{
+    char const * input = "cbacba";
+    p_context_t * context = p_context_new((uint8_t const *)input, strlen(input));
+    assert_eq(P_SUCCESS, p_parse(context));
+    size_t result = p_result(context);
+    assert_eq(0x932187932187, result);
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_custom_lex_fn.d

@@ -0,0 +1,18 @@
+import testparser;
+import std.stdio;
+import testutils;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = "cbacba";
+    p_context_t * context = p_context_new(input);
+    assert_eq(P_SUCCESS, p_parse(context));
+    size_t result = p_result(context);
+    assert_eq(0x932187932187, result);
+    p_context_delete(context);
+}

spec/test_custom_lex_fn.rs

@@ -0,0 +1,8 @@
+use testparser::*;
+
+fn main() {
+    let mut c = p_context_new(b"cbacba");
+    assert_eq!(P_SUCCESS, p_parse(&mut c));
+    assert_eq!(0x932187932187, p_result(&c));
+    p_context_delete(c);
+}

spec/test_drop_code_block.c

@@ -0,0 +1,15 @@
+#include "testparser.h"
+#include <assert.h>
+#include <stdio.h>
+#include <string.h>
+
+int main()
+{
+    char const * input = "    # comment 1\n#    comment 2\na\n";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_drop_code_block.d

@@ -0,0 +1,16 @@
+import testparser;
+import std.stdio;
+import testutils;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = "    # comment 1\n#    comment 2\na\n";
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+}

spec/test_drop_code_block.rs

@@ -0,0 +1,7 @@
+use testparser::*;
+
+fn main() {
+    let mut c = p_context_new(b"    # comment 1\n#    comment 2\na\n");
+    assert_eq!(P_SUCCESS, p_parse(&mut c));
+    p_context_delete(c);
+}

spec/test_error_positions.c

@@ -0,0 +1,47 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+
+int main()
+{
+    char const * input = "a 42";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    p_context_delete(context);
+
+    input = "a\n123\na  a";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_UNEXPECTED_TOKEN);
+    assert(p_position(context).row == 3);
+    assert(p_position(context).col == 4);
+    assert(p_token(context) == TOKEN_a);
+    p_context_delete(context);
+
+    input = "12";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_UNEXPECTED_TOKEN);
+    assert(p_position(context).row == 1);
+    assert(p_position(context).col == 1);
+    assert(p_token(context) == TOKEN_num);
+    p_context_delete(context);
+
+    input = "a 12\n\nab";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_UNEXPECTED_INPUT);
+    assert(p_position(context).row == 3);
+    assert(p_position(context).col == 2);
+    p_context_delete(context);
+
+    input = "a 12\n\na\n\n77\na   \xAA";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_DECODE_ERROR);
+    assert(p_position(context).row == 6);
+    assert(p_position(context).col == 5);
+
+    assert(strcmp(p_token_names[TOKEN_a], "a") == 0);
+    assert(strcmp(p_token_names[TOKEN_num], "num") == 0);
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_error_positions.d

@@ -0,0 +1,40 @@
+import testparser;
+import std.stdio;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = "a 42";
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+
+    input = "a\n123\na  a";
+    context = p_context_new(input);
+    assert(p_parse(context) == P_UNEXPECTED_TOKEN);
+    assert(p_position(context) == p_position_t(3, 4));
+    assert(p_token(context) == TOKEN_a);
+
+    input = "12";
+    context = p_context_new(input);
+    assert(p_parse(context) == P_UNEXPECTED_TOKEN);
+    assert(p_position(context) == p_position_t(1, 1));
+    assert(p_token(context) == TOKEN_num);
+
+    input = "a 12\n\nab";
+    context = p_context_new(input);
+    assert(p_parse(context) == P_UNEXPECTED_INPUT);
+    assert(p_position(context) == p_position_t(3, 2));
+
+    input = "a 12\n\na\n\n77\na   \xAA";
+    context = p_context_new(input);
+    assert(p_parse(context) == P_DECODE_ERROR);
+    assert(p_position(context) == p_position_t(6, 5));
+
+    assert(p_token_names[TOKEN_a] == "a");
+    assert(p_token_names[TOKEN_num] == "num");
+}

spec/test_error_positions.rs

@@ -0,0 +1,35 @@
+use testparser::*;
+
+fn main() {
+    let mut c = p_context_new(b"a 42");
+    assert_eq!(P_SUCCESS, p_parse(&mut c));
+    p_context_delete(c);
+
+    let mut c = p_context_new(b"a\n123\na  a");
+    assert_eq!(P_UNEXPECTED_TOKEN, p_parse(&mut c));
+    assert_eq!(3, p_position(&c).row);
+    assert_eq!(4, p_position(&c).col);
+    assert_eq!(TOKEN_a, p_token(&c));
+    p_context_delete(c);
+
+    let mut c = p_context_new(b"12");
+    assert_eq!(P_UNEXPECTED_TOKEN, p_parse(&mut c));
+    assert_eq!(1, p_position(&c).row);
+    assert_eq!(1, p_position(&c).col);
+    assert_eq!(TOKEN_num, p_token(&c));
+    p_context_delete(c);
+
+    let mut c = p_context_new(b"a 12\n\nab");
+    assert_eq!(P_UNEXPECTED_INPUT, p_parse(&mut c));
+    assert_eq!(3, p_position(&c).row);
+    assert_eq!(2, p_position(&c).col);
+    p_context_delete(c);
+
+    let mut c = p_context_new(b"a 12\n\na\n\n77\na   \xAA");
+    assert_eq!(P_DECODE_ERROR, p_parse(&mut c));
+    assert_eq!(6, p_position(&c).row);
+    assert_eq!(5, p_position(&c).col);
+    assert_eq!("a", p_token_names[TOKEN_a as usize]);
+    assert_eq!("num", p_token_names[TOKEN_num as usize]);
+    p_context_delete(c);
+}

spec/test_field_aliases.c

@@ -0,0 +1,14 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+#include "testutils.h"
+
+int main()
+{
+    char const * input = "foo1\nbar2";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    p_context_delete(context);
+    return 0;
+}

spec/test_field_aliases.d

@@ -0,0 +1,15 @@
+import testparser;
+import std.stdio;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = "foo1\nbar2";
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+}

spec/test_field_aliases.rs

@@ -0,0 +1,7 @@
+use testparser::*;
+
+fn main() {
+    let mut c = p_context_new(b"foo1\nbar2");
+    assert_eq!(P_SUCCESS, p_parse(&mut c));
+    p_context_delete(c);
+}

spec/test_input_index.c

@@ -0,0 +1,60 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+#include "testutils.h"
+
+int main()
+{
+    /* Grammar (simple):
+     *   drop /\\s+/;
+     *   token a; token b;
+     *   Start -> a b;
+     *
+     * Verifies that p_input_index() reports the parser/lexer's current byte
+     * offset into the input text. */
+
+    /* Fresh context: input_index starts at 0. */
+    {
+        char const * input = "ab";
+        p_context_t * context = p_context_new((uint8_t const *)input, strlen(input));
+        assert_eq(0u, p_input_index(context));
+        p_context_delete(context);
+    }
+
+    /* After each successful lex the byte offset advances past the token. */
+    {
+        char const * input = "a b";
+        p_context_t * context = p_context_new((uint8_t const *)input, strlen(input));
+        p_token_info_t token_info;
+        assert(p_lex(context, &token_info) == P_SUCCESS);
+        assert_eq((size_t)TOKEN_a, (size_t)token_info.token);
+        assert_eq(1u, p_input_index(context));
+        assert(p_lex(context, &token_info) == P_SUCCESS);
+        assert_eq((size_t)TOKEN_b, (size_t)token_info.token);
+        /* The dropped space between `a` and `b` advances input_index too. */
+        assert_eq(3u, p_input_index(context));
+        p_context_delete(context);
+    }
+
+    /* After a full successful parse, input_index has reached the end. */
+    {
+        char const * input = "ab";
+        p_context_t * context = p_context_new((uint8_t const *)input, strlen(input));
+        assert(p_parse_Start(context) == P_SUCCESS);
+        assert_eq(2u, p_input_index(context));
+        p_context_delete(context);
+    }
+
+    /* When parse_inner completes via a follow token, the follow token is not
+     * consumed, so input_index points at the start of the follow token. */
+    {
+        char const * input = "abb";
+        p_context_t * context = p_context_new((uint8_t const *)input, strlen(input));
+        p_token_t follow_tokens[] = { TOKEN_b };
+        assert(p_parse_inner_Start(context, follow_tokens, 1u) == P_SUCCESS);
+        assert_eq(2u, p_input_index(context));
+        p_context_delete(context);
+    }
+
+    return 0;
+}

spec/test_input_index.d

@@ -0,0 +1,51 @@
+import testparser;
+import std.stdio;
+import testutils;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    /* See test_input_index.c for details on the grammar and cases. */
+
+    /* Fresh context: input_index starts at 0. */
+    {
+        string input = "ab";
+        p_context_t * context = p_context_new(input);
+        assert(p_input_index(context) == 0);
+    }
+
+    /* After each successful lex the byte offset advances past the token. */
+    {
+        string input = "a b";
+        p_context_t * context = p_context_new(input);
+        p_token_info_t token_info;
+        assert(p_lex(context, &token_info) == P_SUCCESS);
+        assert(token_info.token == TOKEN_a);
+        assert(p_input_index(context) == 1);
+        assert(p_lex(context, &token_info) == P_SUCCESS);
+        assert(token_info.token == TOKEN_b);
+        assert(p_input_index(context) == 3);
+    }
+
+    /* After a full successful parse, input_index has reached the end. */
+    {
+        string input = "ab";
+        p_context_t * context = p_context_new(input);
+        assert(p_parse_Start(context) == P_SUCCESS);
+        assert(p_input_index(context) == 2);
+    }
+
+    /* When parse_inner completes via a follow token, the follow token is not
+     * consumed, so input_index points at the start of the follow token. */
+    {
+        string input = "abb";
+        p_context_t * context = p_context_new(input);
+        p_token_t[] follow_tokens = [TOKEN_b];
+        assert(p_parse_inner_Start(context, follow_tokens) == P_SUCCESS);
+        assert(p_input_index(context) == 2);
+    }
+}