Add Rust target

Store tree nodes in congruent, compact arena array.
Define handle types to refer to tree nodes rather than pointers to
structure instances.
Free tree with context.

.github/workflows/run-tests.yml

@@ -0,0 +1,38 @@
+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: Install dependencies
+      run: bundle install
+
+    - name: Run tests
+      run: rake all

.rspec

@@ -1,3 +1,2 @@
---format documentation
 --color
 --require spec_helper

CHANGELOG.md

@@ -1,3 +1,270 @@
+## v5.0.0
+
+### New Features
+
+- Add Rust target language output.
+
+### API Changes
+
+- 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 return handle values.
+- 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.
+
+## 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

@@ -1,7 +1,9 @@
 source "https://rubygems.org"
 
+gem "base64"
 gem "rake"
 gem "rspec"
 gem "rdoc"
 gem "redcarpet"
 gem "syntax"
+gem "simplecov"

Gemfile.lock

@@ -1,38 +1,54 @@
 GEM
   remote: https://rubygems.org/
   specs:
-    diff-lcs (1.5.0)
-    psych (5.0.1)
+    base64 (0.3.0)
+    date (3.5.1)
+    diff-lcs (1.6.2)
+    docile (1.4.1)
+    erb (6.0.4)
+    psych (5.4.0)
+      date
       stringio
-    rake (13.0.6)
-    rdoc (6.5.0)
+    rake (13.4.2)
+    rdoc (7.2.0)
+      erb
       psych (>= 4.0.0)
-    redcarpet (3.5.1)
-    rspec (3.11.0)
-      rspec-core (~> 3.11.0)
-      rspec-expectations (~> 3.11.0)
-      rspec-mocks (~> 3.11.0)
-    rspec-core (3.11.0)
-      rspec-support (~> 3.11.0)
-    rspec-expectations (3.11.0)
+      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 (~> 3.11.0)
-    rspec-mocks (3.11.1)
+      rspec-support (~> 3.13.0)
+    rspec-mocks (3.13.8)
       diff-lcs (>= 1.2.0, < 2.0)
-      rspec-support (~> 3.11.0)
-    rspec-support (3.11.0)
-    stringio (3.0.4)
+      rspec-support (~> 3.13.0)
+    rspec-support (3.13.7)
+    simplecov (0.22.0)
+      docile (~> 1.1)
+      simplecov-html (~> 0.11)
+      simplecov_json_formatter (~> 0.1)
+    simplecov-html (0.13.2)
+    simplecov_json_formatter (0.1.4)
+    stringio (3.2.0)
     syntax (1.2.2)
+    tsort (0.2.0)
 
 PLATFORMS
   ruby
 
 DEPENDENCIES
+  base64
   rake
   rdoc
   redcarpet
   rspec
+  simplecov
   syntax
 
 BUNDLED WITH
-   2.3.7
+  4.0.14

LICENSE.txt

@@ -1,6 +1,6 @@
 The MIT License (MIT)
 
-Copyright (c) 2010-2023 Josh Holtrop
+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

README.md

@@ -1,21 +1,106 @@
 # The Propane Parser Generator
 
-Propane is an LR Parser Generator (LPG) which:
+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 parser to parse input in linear time
+  * 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
 
-TODO
+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
 
-TODO: Write usage instructions here
+### 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 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)
+  {
+    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
 

Rakefile

@@ -1,18 +0,0 @@
-require "rspec/core/rake_task"
-
-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
-    task.rspec_opts = %W[-e "#{args.example_pattern}" -f documentation]
-  end
-end
-
-task :default => :spec
-
-desc "Build user guide"
-task :user_guide do
-  system("ruby", "-Ilib", "rb/gen_user_guide.rb")
-end

Rakefile.rb

@@ -0,0 +1,46 @@
+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
+    task.rspec_opts = %W[-e "#{args.example_pattern}" -f documentation]
+  end
+end
+task :spec do |task, args|
+  unless ENV["dist_specs"]
+    original_stdout = $stdout
+    sio = StringIO.new
+    $stdout = sio
+    SimpleCov.collate Dir["coverage/.resultset.json"]
+    $stdout = original_stdout
+    sio.string.lines.each do |line|
+      $stdout.write(line) unless line =~ /Coverage report generated for/
+    end
+  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 => [:spec, :dspec, :user_guide]

UPGRADING.md

@@ -0,0 +1,61 @@
+## v5.0.0
+
+The generated API for tree generation mode (`tree;`) has been changed
+significantly for this version.
+The lexer/parser value APIs for non-tree grammars are 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).
+
+## 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 +1,248 @@
+/**
+ * @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;
+
+/**
+ * Lexer and parser context.
+ *
+ * The user must allocate an instance of this structure and pass it to any
+ * public API function.
+ */
+typedef struct
+{
+    /* 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 %>
+} <%= @grammar.prefix %>context_t;
+
+<% 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);

extra/vim/ftdetect/propane.vim

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

extra/vim/syntax/propane.vim

@@ -0,0 +1,51 @@
+" Vim syntax file for Propane
+" Language: propane
+" Maintainer: Josh Holtrop
+" URL: https://github.com/holtrop/propane
+
+if exists("b:current_syntax")
+  finish
+endif
+
+if !exists("b:propane_subtype")
+  if search('\<import\s\+\%(std\|core\)\.', 'nw') > 0
+    let b:propane_subtype = "d"
+  else
+    let b:propane_subtype = "cpp"
+  endif
+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

lib/propane.rb

@@ -1,6 +1,7 @@
 require "erb"
 require "set"
 require "stringio"
+require_relative "propane/assets"
 require_relative "propane/cli"
 require_relative "propane/code_point_range"
 require_relative "propane/fa"
@@ -30,10 +31,10 @@ class Propane
 
   class << self
 
-    def run(input_file, output_file, log_file)
+    def run(input_file, output_file, log_file, options)
       begin
-        grammar = Grammar.new(File.read(input_file))
-        generator = Generator.new(grammar, output_file, log_file)
+        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

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

@@ -4,15 +4,21 @@ class Propane
     USAGE = <<EOF
 Usage: #{$0} [options] <input-file> <output-file>
 Options:
-  --log LOG   Write log file
-  --version   Show program version and exit
-  -h, --help  Show this usage and exit
+  -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 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
@@ -24,11 +30,13 @@ EOF
               log_file = args[i]
             end
           when "--version"
-            puts "propane v#{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
@@ -45,7 +53,7 @@ EOF
           $stderr.puts "Error: cannot read #{params[0]}"
           return 2
         end
-        Propane.run(*params, log_file)
+        Propane.run(*params, log_file, options)
       end
 
     end

lib/propane/generator.rb

@@ -2,7 +2,7 @@ class Propane
 
   class Generator
 
-    def initialize(grammar, output_file, log_file)
+    def initialize(grammar, output_file, log_file, options)
       @grammar = grammar
       @output_file = output_file
       if log_file
@@ -10,15 +10,52 @@ class Propane
       else
         @log = StringIO.new
       end
-      @classname = @grammar.classname || File.basename(output_file).sub(%r{[^a-zA-Z0-9].*}, "").capitalize
+      @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
-      erb = ERB.new(File.read(File.join(File.dirname(File.expand_path(__FILE__)), "../../assets/parser.d.erb")), trim_mode: "<>")
-      result = erb.result(binding.clone)
-      File.open(@output_file, "wb") do |fh|
-        fh.write(result)
+      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]+$}, ".#{extension}")
+        else
+          output_file = @output_file
+        end
+        erb = ERB.new(template, trim_mode: "<>")
+        result = erb.result(binding.clone).lines.each_with_index.map do |line, i|
+          if @language == "rust"
+            # Rust has no #line directive support, so strip the directives that
+            # the grammar embeds in user code blocks.
+            line = line.sub(/^#line \d+ "[^"]*"/, "")
+            line == "#linereset\n" ? "" : line
+          elsif line == "#linereset\n"
+            %[#line #{i + 2} "#{output_file}"\n]
+          else
+            line
+          end
+        end.join
+        File.open(output_file, "wb") do |fh|
+          fh.write(result)
+        end
       end
       @log.close
     end
@@ -29,8 +66,8 @@ class Propane
       # Assign default pattern mode to patterns without a mode assigned.
       found_default = false
       @grammar.patterns.each do |pattern|
-        if pattern.mode.nil?
-          pattern.mode = "default"
+        if pattern.modes.empty?
+          pattern.modes << "default"
           found_default = true
         end
         pattern.ptypename ||= "default"
@@ -38,6 +75,7 @@ class Propane
       unless found_default
         raise Error.new("No patterns found for default mode")
       end
+      check_ptypes!
       # Add EOF token.
       @grammar.tokens << Token.new("$EOF", nil, nil)
       tokens_by_name = {}
@@ -52,12 +90,18 @@ class Propane
         end
         tokens_by_name[token.name] = token
       end
-      # Check for user start rule.
-      unless @grammar.rules.find {|rule| rule.name == "Start"}
-        raise Error.new("Start rule not found")
+      # 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 rule.
-      @grammar.rules.unshift(Rule.new("$Start", ["Start", "$EOF"], nil, nil, nil))
+      # 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|
@@ -106,10 +150,56 @@ class Propane
         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)
+      @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.
@@ -183,19 +273,129 @@ class Propane
     #   Expanded user code block.
     def expand_code(code, parser, rule, pattern)
       code = code.gsub(/\$token\(([$\w]+)\)/) do |match|
-        "Token(TOKEN_#{Token.code_name($1)})"
+        "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|
-          "_pvalue.v_#{rule.ptypename}"
+          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 = $1.to_i
-          "statevalues[$-1-n_states+#{index}].pvalue.v_#{rule.components[index - 1].ptypename}"
+          "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|
-          "result.pvalue.v_#{pattern.ptypename}"
+          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
@@ -203,23 +403,488 @@ class Propane
           unless mode_id
             raise Error.new("Lexer mode '#{mode_name}' not found")
           end
-          "m_mode = #{mode_id}u"
+          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[$-1-n_states+#{index}].pvalue.v_#{component.ptypename}"
+        when "rust"
+          "statevalues[statevalues.len() - 1 - n_states + #{index}].pvalue.get_v_#{component.ptypename}()"
+        end
+      end
+    end
+
+    # Construct a tree node handle expression for the target language.
+    #
+    # A handle is a small value pairing the parser context with a node ID
+    # (an index into the context's node arena). All handle types share this
+    # layout; the distinct types exist for documentation and, in C, to drive
+    # the tree walk macro's type threading.
+    #
+    # @param typename [String]
+    #   Handle type name.
+    # @param id_expr [String]
+    #   Expression yielding the node ID.
+    #
+    # @return [String]
+    #   Handle constructor expression.
+    def tree_handle(typename, id_expr)
+      if @cpp
+        "(#{typename}{context, #{id_expr}})"
+      elsif @language == "c"
+        "((#{typename}){context, #{id_expr}})"
+      elsif @language == "rust"
+        "(#{typename} { context, id: #{id_expr} })"
+      else
+        "#{typename}(context, #{id_expr})"
+      end
+    end
+
+    # Get the list of non-optional, non-internal rule sets that get a tree node
+    # handle type generated for them.
+    #
+    # @return [Array<Propane::RuleSet>]
+    #   Rule sets with generated tree node handle types.
+    def tree_node_rule_sets
+      @parser.rule_sets.reject do |name, rule_set|
+        name.start_with?("$") || 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 C/C++ tree node handle type section for the header.
+    #
+    # @return [String]
+    #   Header handle section.
+    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_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 << 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 type section for the header.
+    def cpp_tree_types_header
+      p = @grammar.prefix
+      out = []
+      out << "/** Tree node handle types. @{ */"
+      tree_handle_types.each {|t| out << "struct #{t};"}
+      out << ""
+      # Token handle (all methods inline; no handle-typed returns).
+      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 << "    #{p}node_data_t * data() const { return &__context->#{p}tree_nodes[__id]; }"
+      out << "    #{p}position_t position() const { return __context->#{p}tree_nodes[__id].position; }"
+      out << "    #{p}position_t end_position() const { return __context->#{p}tree_nodes[__id].end_position; }"
+      out << "    uint16_t n_fields() const { return __id ? __context->#{p}tree_nodes[__id].n_fields : (uint16_t)0u; }"
+      out << "    #{p}token_t token() const { return __context->#{p}tree_nodes[__id].token; }"
+      out << "    #{p}value_t pvalue() const { return __context->#{p}tree_nodes[__id].pvalue; }"
+      out << "};"
+      out << ""
+      # Rule set handles: navigation methods declared, defined out-of-line below.
+      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 << "    #{p}node_data_t * data() const { return &__context->#{p}tree_nodes[__id]; }"
+        out << "    #{p}position_t position() const { return __context->#{p}tree_nodes[__id].position; }"
+        out << "    #{p}position_t end_position() const { return __context->#{p}tree_nodes[__id].end_position; }"
+        out << "    uint16_t n_fields() const { return __id ? __context->#{p}tree_nodes[__id].n_fields : (uint16_t)0u; }"
+        each_tree_field(rule_set) do |rt, field_name, child_type, slot|
+          out << "    #{child_type} #{field_name}() const;"
+        end
+        out << "};"
+        out << ""
+      end
+      # 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
+
+    # Generate the Rust tree node record and handle types.
+    #
+    # Mirrors the C tree node record plus the C++ handle structs: each rule set
+    # and the Token node get a handle type ({context, id}) with accessor methods.
+    #
+    # @return [String]
+    #   Rust tree node type definitions.
+    def rust_tree_types
+      p = @grammar.prefix
+      out = []
+      out << "/** Tree node record. */"
+      out << "#[derive(Clone, Default)]"
+      out << "pub struct #{p}node_data_t {"
+      out << "    pub position: #{p}position_t,"
+      out << "    pub end_position: #{p}position_t,"
+      out << "    pub child_offset: #{p}node_id_t,"
+      out << "    pub n_fields: u16,"
+      out << "    pub is_token: bool,"
+      out << "    pub token: #{p}token_t,"
+      out << "    pub pvalue: #{p}value_t,"
+      unless @grammar.token_user_fields.to_s.strip.empty?
+        out << @grammar.token_user_fields
+      end
+      out << "}"
+      out << ""
+      out << "/** Tree node handle types. */"
+      tree_handle_types.each do |t|
+        out << "#[derive(Clone, Copy)]"
+        out << "pub struct #{t}<'a> { context: &'a #{p}context_t, id: #{p}node_id_t }"
+      end
+      out << ""
+      # Common accessors for every handle type.
+      tree_handle_types.each do |t|
+        out << "impl<'a> #{t}<'a> {"
+        out << "    /** Return whether this handle refers to a valid (non-null) node. */"
+        out << "    pub fn valid(&self) -> bool { self.id != 0 }"
+        out << "    /** Return the node ID (for identity comparison). */"
+        out << "    pub fn node_id(&self) -> #{p}node_id_t { self.id }"
+        out << "    /** Access the underlying node record. */"
+        out << "    pub fn data(&self) -> &'a #{p}node_data_t { &self.context.#{p}tree_nodes[self.id as usize] }"
+        out << "    /** Text position of the first code point spanned by this node. */"
+        out << "    pub fn position(&self) -> #{p}position_t { self.context.#{p}tree_nodes[self.id as usize].position }"
+        out << "    /** Text position of the last code point spanned by this node. */"
+        out << "    pub fn end_position(&self) -> #{p}position_t { self.context.#{p}tree_nodes[self.id as usize].end_position }"
+        out << "    /** Number of child fields in this node. */"
+        out << "    pub fn n_fields(&self) -> u16 { if self.id != 0 { self.context.#{p}tree_nodes[self.id as usize].n_fields } else { 0 } }"
+        if t == h_type("Token")
+          out << "    /** Token ID for this token node. */"
+          out << "    pub fn token(&self) -> #{p}token_t { self.context.#{p}tree_nodes[self.id as usize].token }"
+          out << "    /** Parser value associated with this token node. */"
+          out << "    pub fn pvalue(&self) -> #{p}value_t { self.context.#{p}tree_nodes[self.id as usize].pvalue.clone() }"
+        end
+        out << "}"
+      end
+      out << ""
+      # Navigation accessors for rule set handles.
+      tree_node_rule_sets.each do |rule_set|
+        rtype = h_type(rule_set.name)
+        out << "impl<'a> #{rtype}<'a> {"
+        each_tree_field(rule_set) do |rt, field_name, child_type, slot|
+          out << "    /** Access the #{field_name} child node. */"
+          out << "    pub fn #{rust_ident(field_name)}(&self) -> #{child_type}<'a> {"
+          out << "        if self.id == 0 {"
+          out << "            return #{child_type} { context: self.context, id: 0 };"
+          out << "        }"
+          out << "        #{child_type} { context: self.context, id: self.context.#{p}tree_children[self.context.#{p}tree_nodes[self.id as usize].child_offset as usize + #{slot}] }"
+          out << "    }"
+        end
+        out << "}"
+      end
+      out.join("\n")
+    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
+    def start_rule_type(start_rule_index = 0)
       start_rule = @grammar.rules.find do |rule|
-        rule.name == "Start"
+        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

@@ -5,31 +5,63 @@ class Propane
     # 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 :classname
+    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_reader :rules
+    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)
+    def initialize(input, filename)
+      @filename = filename
       @patterns = []
+      @start_rules = []
       @tokens = []
       @rules = []
-      @code_blocks = []
+      @code_blocks = {}
       @line_number = 1
       @next_line_number = @line_number
-      @mode = nil
+      @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!
@@ -41,16 +73,26 @@ class Propane
     def parse_statement!
       if parse_white_space!
       elsif parse_comment_line!
-      elsif @mode.nil? && parse_mode_label!
+      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_class_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) + "..."
@@ -60,8 +102,8 @@ class Propane
     end
 
     def parse_mode_label!
-      if md = consume!(/(#{IDENTIFIER_REGEX})\s*:/)
-        @mode = md[1]
+      if md = consume!(/(#{IDENTIFIER_REGEX}(?:\s*,\s*#{IDENTIFIER_REGEX})*)\s*:/)
+        @modeline = md[1]
       end
     end
 
@@ -73,23 +115,75 @@ class Propane
       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 `;'")
-        @mode = nil
+        @modeline = nil
         true
       end
     end
 
-    def parse_class_statement!
-      if consume!(/class\s+/)
-        md = consume!(/([\w.]+)\s*/, "expected class name")
-        @classname = md[1]
-        consume!(/;/, "expected `;'")
-        @mode = nil
-        true
+    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
 
@@ -97,6 +191,9 @@ class Propane
       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")
@@ -109,18 +206,21 @@ class Propane
         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 pattern or `;' or 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, mode: @mode, ptypename: ptypename)
+        pattern = Pattern.new(pattern: pattern, token: token, line_number: @line_number, code: code, modes: get_modes_from_modeline, ptypename: ptypename)
         @patterns << pattern
-        @mode = nil
+        @modeline = nil
         true
       end
     end
@@ -130,12 +230,15 @@ class Propane
         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
-        @mode = nil
+        @modeline = nil
         true
       end
     end
@@ -147,9 +250,11 @@ class Propane
           raise Error.new("Line #{@line_number}: expected pattern to follow `drop'")
         end
         consume!(/\s+/)
-        consume!(/;/, "expected `;'")
-        @patterns << Pattern.new(pattern: pattern, line_number: @line_number, drop: true, mode: @mode)
-        @mode = nil
+        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
@@ -157,13 +262,16 @@ class Propane
     def parse_rule_statement!
       if md = consume!(/(#{IDENTIFIER_REGEX})\s*(?:\((#{IDENTIFIER_REGEX})\))?\s*->\s*/)
         rule_name, ptypename = *md[1, 2]
-        md = consume!(/((?:#{IDENTIFIER_REGEX}\s*)*)\s*/, "expected rule component list")
+        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 pattern or `;' or code block")
+          consume!(/;/, "expected `;' or code block")
         end
         @rules << Rule.new(rule_name, components, code, ptypename, @line_number)
-        @mode = nil
+        @modeline = nil
         true
       end
     end
@@ -172,21 +280,60 @@ class Propane
       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, mode: @mode, ptypename: ptypename)
-        @mode = nil
+        @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 code = parse_code_block!
-        @code_blocks << code
-        @mode = nil
+      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
@@ -204,6 +351,8 @@ class Propane
             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
@@ -211,8 +360,16 @@ class Propane
     end
 
     def parse_code_block!
-      if md = consume!(/<<\n(.*?)^>>\n/m)
-        md[1]
+      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
 
@@ -244,6 +401,14 @@ class Propane
       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

@@ -1,55 +1,13 @@
 class Propane
   class Lexer
 
+    attr_reader :state_table
+    attr_reader :transition_table
+    attr_reader :mode_table
+
     def initialize(grammar)
       @grammar = grammar
-    end
-
-    def build_tables
-      @modes = @grammar.patterns.group_by do |pattern|
-        pattern.mode
-      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
-      transition_table = []
-      state_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|
-          drop = state.accepts && state.accepts.drop?
-          token =
-            if state.accepts && state.accepts.token
-              state.accepts.token.id
-            end
-          code_id =
-            if state.accepts && state.accepts.code_id
-              state.accepts.code_id
-            end
-          state_table << {
-            transition_table_index: transition_table.size,
-            n_transitions: state.transitions.size,
-            drop: drop,
-            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
-      [transition_table, state_table, mode_table]
+      build_tables!
     end
 
     # Get ID for a mode.
@@ -65,5 +23,52 @@ class Propane
       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/parser.rb

@@ -2,26 +2,43 @@ class Propane
 
   class Parser
 
-    def initialize(grammar, rule_sets, log)
+    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 = {}
-      start_item = Item.new(grammar.rules.first, 0)
-      eval_item_sets = Set[ItemSet.new([start_item])]
+      @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 = eval_item_sets.first
+        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.following_symbols.each do |following_symbol|
-            unless following_symbol.name == "$EOF"
-              following_set = item_set.build_following_item_set(following_symbol)
-              eval_item_sets << following_set
+          item_set.next_symbols.each do |next_symbol|
+            unless next_symbol.name == "$EOF"
+              next_item_set = item_set.build_next_item_set(next_symbol)
+              eval_item_sets << next_item_set
             end
           end
         end
@@ -32,59 +49,71 @@ class Propane
       end
 
       build_reduce_actions!
+      build_tables!
       write_log!
-    end
-
-    def build_tables
-      shift_table = []
-      state_table = []
-      reduce_table = []
-      @item_sets.each do |item_set|
-        shift_entries = item_set.following_symbols.map do |following_symbol|
-          state_id =
-            if following_symbol.name == "$EOF"
-              0
-            else
-              item_set.following_item_set[following_symbol].id
-            end
-          {
-            token_id: following_symbol.id,
-            state_id: state_id,
-          }
-        end
-        reduce_entries =
-          case ra = item_set.reduce_actions
-          when Rule
-            [{token_id: @grammar.tokens.size, rule_id: ra.id,
-              rule_set_id: ra.rule_set.id, n_states: ra.components.size}]
-          when Hash
-            ra.map do |token, rule|
-              {token_id: token.id, rule_id: rule.id,
-               rule_set_id: rule.rule_set.id, n_states: rule.components.size}
-            end
-          else
-            []
-          end
-        state_table << {
-          shift_index: shift_table.size,
-          n_shifts: shift_entries.size,
-          reduce_index: reduce_table.size,
-          n_reduces: reduce_entries.size,
-        }
-        shift_table += shift_entries
-        reduce_table += reduce_entries
+      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
-      [state_table, shift_table, reduce_table]
     end
 
     private
 
+    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.following_symbols.each do |following_symbol|
-        unless following_symbol.name == "$EOF"
-          following_set = @item_sets_set[item_set.build_following_item_set(following_symbol)]
-          item_set.following_item_set[following_symbol] = following_set
-          following_set.in_sets << item_set
+      item_set.next_symbols.each do |next_symbol|
+        unless next_symbol.name == "$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
@@ -94,7 +123,109 @@ class Propane
     # @return [void]
     def build_reduce_actions!
       @item_sets.each do |item_set|
-        item_set.reduce_actions = build_reduce_actions_for_item_set(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
 
@@ -103,40 +234,45 @@ class Propane
     # @param item_set [ItemSet]
     #   ItemSet (parser state)
     #
-    # @return [nil, Rule, Hash]
-    #   If no reduce actions are possible for the given item set, nil.
-    #   If only one reduce action is possible for the given item set, the Rule
-    #   to reduce.
-    #   Otherwise, a mapping of lookahead Tokens to the Rules to reduce.
+    # @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.
-      reduce_rules = Set.new(item_set.items.select(&:complete?).map(&:rule))
+      item_set.reduce_rules = Set.new(item_set.items.select(&:complete?).map(&:rule))
 
-      # If there are no rules to reduce for this ItemSet, we're done here.
-      return nil if reduce_rules.size == 0
-
-      # If there is exactly one rule to reduce for this ItemSet, then do not
-      # figure out the lookaheads; just reduce it.
-      return reduce_rules.first if reduce_rules.size == 1
-
-      # Otherwise, we have more than one possible rule to reduce.
+      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 = item_set.leading_item_sets
-      reduce_rules.reduce({}) do |reduce_actions, reduce_rule|
+      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]
-            raise Error.new("Error: reduce/reduce conflict between rule #{existing_reduce_rule.id} (#{existing_reduce_rule.name}) and rule #{reduce_rule.id} (#{reduce_rule.name})")
+            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
@@ -176,13 +312,14 @@ class Propane
         # tokens to form the lookahead token set.
         item_sets.each do |item_set|
           item_set.items.each do |item|
-            if item.following_symbol == rule_set
+            if item.next_symbol == rule_set
               (1..).each do |offset|
-                case symbol = item.following_symbol(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
@@ -235,20 +372,26 @@ class Propane
         @log.puts
         @log.puts "  Incoming states: #{incoming_ids.join(", ")}"
         @log.puts "  Outgoing states:"
-        item_set.following_item_set.each do |following_symbol, following_item_set|
-          @log.puts "    #{following_symbol.name} => #{following_item_set.id}"
+        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:"
-        case item_set.reduce_actions
-        when Rule
-          @log.puts "    * => #{item_set.reduce_actions.id} (#{item_set.reduce_actions.name})"
-        when Hash
+        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

lib/propane/parser/item.rb

@@ -22,6 +22,7 @@ class Propane
       def initialize(rule, position)
         @rule = rule
         @position = position
+        @_hash = [@rule, @position].hash
       end
 
       # Hash function.
@@ -29,7 +30,7 @@ class Propane
       # @return [Integer]
       #   Hash code.
       def hash
-        [@rule, @position].hash
+        @_hash
       end
 
       # Compare Item objects.
@@ -56,7 +57,7 @@ class Propane
 
       # Return the set of Items obtained by "closing" the current item.
       #
-      # If the following symbol for the current item is another Rule name, then
+      # 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.
       #
@@ -81,17 +82,17 @@ class Propane
         @position == @rule.components.size
       end
 
-      # Get the following symbol for the Item.
+      # Get the next symbol for the Item.
       #
-      # That is, the symbol which follows the parse position marker in the
+      # 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]
-      #   Following symbol for the Item.
-      def following_symbol(offset = 0)
+      #   Next symbol for the Item.
+      def next_symbol(offset = 0)
         @rule.components[@position + offset]
       end
 
@@ -108,25 +109,25 @@ class Propane
         end
       end
 
-      # Get whether this Item is followed by the provided symbol.
+      # Get whether this Item's next symbol is the given symbol.
       #
       # @param symbol [Token, RuleSet]
       #   Symbol to query.
       #
       # @return [Boolean]
-      #   Whether this Item is followed by the provided symbol.
-      def followed_by?(symbol)
-        following_symbol == symbol
+      #   Whether this Item's next symbol is the given symbol.
+      def next_symbol?(symbol)
+        next_symbol == symbol
       end
 
-      # Get the following item for this Item.
+      # 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 following item for this Item.
-      def following_item
+      #   The next item for this Item.
+      def next_item
         Item.new(@rule, @position + 1)
       end
 

lib/propane/parser/item_set.rb

@@ -2,7 +2,7 @@ class Propane
   class Parser
 
     # Represent a parser "item set", which is a set of possible items that the
-    # parser could currently be parsing.
+    # parser could currently be parsing. This is equivalent to a parser state.
     class ItemSet
 
       # @return [Set<Item>]
@@ -14,45 +14,58 @@ class Propane
       attr_accessor :id
 
       # @return [Hash]
-      #   Maps a following symbol to its ItemSet.
-      attr_reader :following_item_set
+      #   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, Hash]
-      #   Reduce actions, mapping lookahead tokens to rules.
+      # @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)
-        @following_item_set = {}
+        @next_item_set = {}
         @in_sets = Set.new
         close!
       end
 
-      # Get the set of following symbols for all Items in this ItemSet.
+      # Get the set of next symbols for all Items in this ItemSet.
       #
       # @return [Set<Token, RuleSet>]
-      #   Set of following symbols for all Items in this ItemSet.
-      def following_symbols
-        Set.new(@items.map(&:following_symbol).compact)
+      #   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 following ItemSet for the given following symbol.
+      # Build a next ItemSet for the given next symbol.
       #
       # @param symbol [Token, RuleSet]
-      #   Following symbol to build the following ItemSet for.
+      #   Next symbol to build the next ItemSet for.
       #
       # @return [ItemSet]
-      #   Following ItemSet for the given following symbol.
-      def build_following_item_set(symbol)
-        ItemSet.new(items_followed_by(symbol).map(&:following_item))
+      #   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.
@@ -87,14 +100,27 @@ class Propane
 
       # Set of ItemSets that lead to this ItemSet.
       #
-      # This set includes this ItemSet.
-      #
       # @return [Set<ItemSet>]
       #   Set of all ItemSets that lead up to this ItemSet.
       def leading_item_sets
-        @in_sets.reduce(Set[self]) do |result, item_set|
-          result + item_set.leading_item_sets
-        end
+        @_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.
@@ -127,16 +153,16 @@ class Propane
         end
       end
 
-      # Get the Items followed by the given following symbol.
+      # Get the Items with the given next symbol.
       #
       # @param symbol [Token, RuleSet]
-      #   Following symbol.
+      #   Next symbol.
       #
       # @return [Array<Item>]
-      #   Items followed by the given following symbol.
-      def items_followed_by(symbol)
+      #   Items with the given next symbol.
+      def items_with_next(symbol)
         @items.select do |item|
-          item.followed_by?(symbol)
+          item.next_symbol?(symbol)
         end
       end
 

lib/propane/pattern.rb

@@ -26,9 +26,9 @@ class Propane
     #   Regex NFA for matching the pattern.
     attr_reader :nfa
 
-    # @return [String, nil]
-    #   Lexer mode for this pattern.
-    attr_accessor :mode
+    # @return [Set]
+    #   Lexer modes for this pattern.
+    attr_accessor :modes
 
     # @return [String, nil]
     #   Parser value type name.
@@ -40,37 +40,26 @@ class Propane
     #   Optional parameters.
     # @option options [String, nil] :code
     #   Code block to execute when the pattern is matched.
-    # @option options [Boolean] :drop
-    #   Whether this is a drop pattern.
     # @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] :mode
-    #   Lexer mode for this pattern.
+    # @option options [String, nil] :modes
+    #   Lexer modes for this pattern.
     def initialize(options)
       @code = options[:code]
-      @drop = options[:drop]
       @pattern = options[:pattern]
       @token = options[:token]
       @line_number = options[:line_number]
-      @mode = options[:mode]
+      @modes = options[:modes]
       @ptypename = options[:ptypename]
-      regex = Regex.new(@pattern)
+      regex = Regex.new(@pattern, @line_number)
       regex.nfa.end_state.accepts = self
       @nfa = regex.nfa
     end
 
-    # Whether the pattern is a drop pattern.
-    #
-    # @return [Boolean]
-    #   Whether the pattern is a drop pattern.
-    def drop?
-      @drop
-    end
-
   end
 
 end

lib/propane/regex.rb

@@ -4,12 +4,13 @@ class Propane
     attr_reader :unit
     attr_reader :nfa
 
-    def initialize(pattern)
+    def initialize(pattern, line_number)
       @pattern = pattern.dup
+      @line_number = line_number
       @unit = parse_alternates
       @nfa = @unit.to_nfa
       if @pattern != ""
-        raise Error.new(%[Unexpected "#{@pattern}" in pattern])
+        raise Error.new(%[Line #{@line_number}: unexpected "#{@pattern}" in pattern])
       end
     end
 
@@ -41,7 +42,7 @@ class Propane
             mu = MultiplicityUnit.new(last_unit, min_count, max_count)
             au.replace_last!(mu)
           else
-            raise Error.new("#{c} follows nothing")
+            raise Error.new("Line #{@line_number}: #{c} follows nothing")
           end
         when "|"
           au.new_alternate!
@@ -59,7 +60,7 @@ class Propane
     def parse_group
       au = parse_alternates
       if @pattern[0] != ")"
-        raise Error.new("Unterminated group in pattern")
+        raise Error.new("Line #{@line_number}: unterminated group in pattern")
       end
       @pattern.slice!(0)
       au
@@ -70,7 +71,7 @@ class Propane
       index = 0
       loop do
         if @pattern == ""
-          raise Error.new("Unterminated character class")
+          raise Error.new("Line #{@line_number}: unterminated character class")
         end
         c = @pattern.slice!(0)
         if c == "]"
@@ -84,13 +85,13 @@ class Propane
         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("Character range must be between single characters")
+            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("Character range must be between single characters")
+              raise Error.new("Line #{@line_number}: character range must be between single characters")
             end
             max_code_point = end_cu.code_point
           else
@@ -116,7 +117,7 @@ class Propane
         elsif max_count.to_s != ""
           max_count = max_count.to_i
           if max_count < min_count
-            raise Error.new("Maximum repetition count cannot be less than minimum repetition count")
+            raise Error.new("Line #{@line_number}: maximum repetition count cannot be less than minimum repetition count")
           end
         else
           max_count = nil
@@ -124,18 +125,33 @@ class Propane
         @pattern = pattern
         [min_count, max_count]
       else
-        raise Error.new("Unexpected match count at #{@pattern}")
+        raise Error.new("Line #{@line_number}: unexpected match count following {")
       end
     end
 
     def parse_backslash
       if @pattern == ""
-        raise Error.new("Error: unfollowed \\")
+        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(" ")
@@ -145,6 +161,35 @@ class Propane
           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

lib/propane/regex/unit.rb

@@ -92,16 +92,19 @@ class Propane
         @units = []
         @negate = false
       end
-      def initialize
-        @units = []
-      end
-      def method_missing(*args)
-        @units.__send__(*args)
+      def method_missing(*args, &block)
+        @units.__send__(*args, &block)
       end
       def <<(thing)
         if thing.is_a?(CharacterClassUnit)
-          thing.each do |ccu_unit|
-            @units << ccu_unit
+          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

lib/propane/rule.rb

@@ -6,6 +6,10 @@ class Propane
     #   Rule components.
     attr_reader :components
 
+    # @return [Hash]
+    #   Field aliases.
+    attr_reader :aliases
+
     # @return [String]
     #   User code associated with the rule.
     attr_reader :code
@@ -30,6 +34,11 @@ class Propane
     #   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]
@@ -44,7 +53,20 @@ class Propane
     #   Line number where the rule was defined in the input grammar.
     def initialize(name, components, code, ptypename, line_number)
       @name = name
-      @components = components
+      @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
@@ -60,6 +82,14 @@ class Propane
       @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]
@@ -68,6 +98,17 @@ class Propane
       "#{@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

@@ -1,10 +1,15 @@
 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_accessor :id
+    attr_reader :id
 
     # @return [String]
     #   Name of the RuleSet.
@@ -51,6 +56,24 @@ class Propane
       @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>]
@@ -75,6 +98,72 @@ class Propane
       @_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/util.rb

@@ -10,6 +10,32 @@ class Propane
         "#{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

lib/propane/version.rb

@@ -1,3 +1,3 @@
 class Propane
-  VERSION = "0.1.0"
+  VERSION = "4.8.1"
 end

propane.sh

@@ -1,2 +0,0 @@
-#!/bin/sh
-exec bundle exec ruby -Ilib bin/propane "$@"

rb/build_dist.rb

@@ -1,5 +1,6 @@
 #!/usr/bin/env ruby
 
+require "erb"
 require "fileutils"
 require "digest/md5"
 
@@ -13,6 +14,24 @@ 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 = []
 
@@ -25,7 +44,11 @@ combine_files = lambda do |file|
         if File.exist?(path)
           unless files_processed[path]
             files_processed[path] = true
-            combine_files[path]
+            if require_name == "propane/assets"
+              combined_file << assets_module
+            else
+              combine_files[path]
+            end
           end
         else
           raise "require path #{path.inspect} not found"

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[i] == '-')
+    {
+        negative = true;
+        i++;
+    }
+    while ('0' <= match[i] && match[i] <= '9')
+    {
+        n *= 10.0;
+        n += (match[i] - '0');
+        i++;
+    }
+    if (match[i] == '.')
+    {
+        i++;
+        double mult = 0.1;
+        while ('0' <= match[i] && match[i] <= '9')
+        {
+            n += mult * (match[i] - '0');
+            mult /= 10.0;
+            i++;
+        }
+    }
+    if (match[i] == 'e' || match[i] == 'E')
+    {
+        bool exp_negative = false;
+        i++;
+        if (match[i] == '-')
+        {
+            exp_negative = true;
+            i++;
+        }
+        else if (match[i] == '+')
+        {
+            i++;
+        }
+        long exp = 0.0;
+        while ('0' <= match[i] && match[i] <= '9')
+        {
+            exp *= 10;
+            exp += (match[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);
+  $$->string = string_value;
+  $mode(default);
+>>
+string: /\\"/ <<
+  str_append(&string_value, "\"");
+>>
+string: /\\\\/ <<
+  str_append(&string_value, "\\");
+>>
+string: /\\\// <<
+  str_append(&string_value, "/");
+>>
+string: /\\b/ <<
+  str_append(&string_value, "\b");
+>>
+string: /\\f/ <<
+  str_append(&string_value, "\f");
+>>
+string: /\\n/ <<
+  str_append(&string_value, "\n");
+>>
+string: /\\r/ <<
+  str_append(&string_value, "\r");
+>>
+string: /\\t/ <<
+  str_append(&string_value, "\t");
+>>
+string: /\\u[0-9a-fA-F]{4}/ <<
+  /* Not actually going to encode the code point for this example... */
+  char s[] = {'{', (char)match[2], (char)match[3], (char)match[4], (char)match[5], '}', 0};
+  str_append(&string_value, s);
+>>
+string: /[^\\]/ <<
+  char s[] = {(char)match[0], 0};
+  str_append(&string_value, s);
+>>
+Start -> Value <<
+  $$ = $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.rust.propane

@@ -0,0 +1,176 @@
+<<header
+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_).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 /"/ <<
+  $$ = JSONValue::StringVal(std::mem::take(&mut ${context.string_value}));
+  $mode(default);
+>>
+string: /\\"/ <<
+  ${context.string_value}.push('"');
+>>
+string: /\\\\/ <<
+  ${context.string_value}.push('\\');
+>>
+string: /\\\// <<
+  ${context.string_value}.push('/');
+>>
+string: /\\b/ <<
+  ${context.string_value}.push('\u{0008}');
+>>
+string: /\\f/ <<
+  ${context.string_value}.push('\u{000C}');
+>>
+string: /\\n/ <<
+  ${context.string_value}.push('\n');
+>>
+string: /\\r/ <<
+  ${context.string_value}.push('\r');
+>>
+string: /\\t/ <<
+  ${context.string_value}.push('\t');
+>>
+string: /\\u[0-9a-fA-F]{4}/ <<
+  /* Not actually going to encode the code point for this example... */
+  let s: String = ['{', match_[2] as char, match_[3] as char, match_[4] as char, match_[5] as char, '}'].iter().collect();
+  ${context.string_value}.push_str(&s);
+>>
+string: /[^\\]/ <<
+  ${context.string_value}.push(match_[0] as char);
+>>
+Start -> Value <<
+  $$ = $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.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, 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)
+    {
+        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_ { 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, 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) { 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_ { 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/grammar_spec.rb

@@ -5,7 +5,6 @@ class Propane
 # Comment line
 
 module a.b;
-class Foobar;
 ptype   XYZ  *  ;
 
 token while;
@@ -29,52 +28,52 @@ B -> <<
   b = 0;
 >>
 EOF
-      grammar = Grammar.new(input)
-      expect(grammar.classname).to eq "Foobar"
+      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 7
+      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 7
+      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 10
+      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 10
+      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 12
+      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 12
-      expect(o.code).to eq "Code for the token\n"
+      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 16
+      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 18
+      expect(o.line_number).to eq 17
       expect(o.token).to be_nil
       expect(o.code).to be_nil
 
@@ -83,20 +82,20 @@ EOF
       o = grammar.rules[0]
       expect(o.name).to eq "A"
       expect(o.components).to eq %w[B]
-      expect(o.line_number).to eq 20
-      expect(o.code).to eq "  a = 42;\n"
+      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 23
+      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 24
-      expect(o.code).to eq "  b = 0;\n"
+      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
@@ -111,8 +110,11 @@ token code2 <<
 >>
 
 tokenid token_with_no_pattern;
+
+prefix myparser_;
 EOF
-      grammar = Grammar.new(input)
+      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
@@ -120,7 +122,7 @@ EOF
 
       o = grammar.patterns.find {|pattern| pattern.token == o}
       expect(o).to_not be_nil
-      expect(o.code).to eq "  a = b;\n  return c;\n"
+      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
@@ -128,7 +130,42 @@ EOF
 
       o = grammar.patterns.find {|pattern| pattern.token == o}
       expect(o).to_not be_nil
-      expect(o.code).to eq %[  writeln("Hello there");\n]
+      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
@@ -142,37 +179,37 @@ m2: /bar/ <<
 drop /q/;
 m3: drop /r/;
 EOF
-      grammar = Grammar.new(input)
+      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.mode).to 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.mode).to eq "m1"
+      expect(o.modes).to eq Set["m1"]
 
       o = grammar.patterns.find {|pattern| pattern.pattern == "foo"}
       expect(o).to_not be_nil
-      expect(o.mode).to be_nil
+      expect(o.modes).to be_empty
 
       o = grammar.patterns.find {|pattern| pattern.pattern == "bar"}
       expect(o).to_not be_nil
-      expect(o.mode).to eq "m2"
+      expect(o.modes).to eq Set["m2"]
 
       o = grammar.patterns.find {|pattern| pattern.pattern == "q"}
       expect(o).to_not be_nil
-      expect(o.mode).to be_nil
+      expect(o.modes).to be_empty
 
       o = grammar.patterns.find {|pattern| pattern.pattern == "r"}
       expect(o).to_not be_nil
-      expect(o.mode).to eq "m3"
+      expect(o.modes).to eq Set["m3"]
     end
 
     it "allows assigning ptypes to tokens and rules" do
@@ -195,7 +232,7 @@ tokenid int(integer);
 Start (node) -> R;
 R -> abc int;
 EOF
-      grammar = Grammar.new(input)
+      grammar = Grammar.new(input, "test.propane")
 
       o = grammar.tokens.find {|token| token.name == "abc"}
       expect(o).to_not be_nil

spec/propane/lexer/dfa_spec.rb

@@ -51,7 +51,7 @@ class TestLexer
 end
 
 def run(grammar, input)
-  grammar = Propane::Grammar.new(grammar)
+  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)
@@ -126,6 +126,74 @@ EOF
     ]
     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/regex_spec.rb

@@ -2,14 +2,14 @@ class Propane
   RSpec.describe Regex do
 
     it "parses an empty expression" do
-      regex = Regex.new("")
+      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")
+      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
@@ -19,7 +19,7 @@ class Propane
     end
 
     it "parses a group with a single character unit expression" do
-      regex = Regex.new("(a)")
+      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
@@ -33,7 +33,7 @@ class Propane
     end
 
     it "parses a *" do
-      regex = Regex.new("a*")
+      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
@@ -47,7 +47,7 @@ class Propane
     end
 
     it "parses a +" do
-      regex = Regex.new("a+")
+      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
@@ -61,7 +61,7 @@ class Propane
     end
 
     it "parses a ?" do
-      regex = Regex.new("a?")
+      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
@@ -75,7 +75,7 @@ class Propane
     end
 
     it "parses a multiplicity count" do
-      regex = Regex.new("a{5}")
+      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
@@ -89,7 +89,7 @@ class Propane
     end
 
     it "parses a minimum-only multiplicity count" do
-      regex = Regex.new("a{5,}")
+      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
@@ -103,7 +103,7 @@ class Propane
     end
 
     it "parses a minimum and maximum multiplicity count" do
-      regex = Regex.new("a{5,8}")
+      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
@@ -118,7 +118,7 @@ class Propane
     end
 
     it "parses an escaped *" do
-      regex = Regex.new("a\\*")
+      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
@@ -131,7 +131,7 @@ class Propane
     end
 
     it "parses an escaped +" do
-      regex = Regex.new("a\\+")
+      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
@@ -144,7 +144,7 @@ class Propane
     end
 
     it "parses an escaped \\" do
-      regex = Regex.new("\\\\d")
+      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
@@ -157,7 +157,7 @@ class Propane
     end
 
     it "parses a character class" do
-      regex = Regex.new("[a-z_]")
+      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
@@ -175,7 +175,7 @@ class Propane
     end
 
     it "parses a negated character class" do
-      regex = Regex.new("[^xyz]")
+      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
@@ -189,8 +189,25 @@ class Propane
       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]")
+      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
@@ -204,7 +221,7 @@ class Propane
     end
 
     it "parses - as a plain character at end of a character class" do
-      regex = Regex.new("[0-]")
+      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
@@ -220,7 +237,7 @@ class Propane
     end
 
     it "parses - as a plain character at beginning of a negated character class" do
-      regex = Regex.new("[^-9]")
+      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
@@ -235,7 +252,7 @@ class Propane
     end
 
     it "parses . as a plain character in a character class" do
-      regex = Regex.new("[.]")
+      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
@@ -250,7 +267,7 @@ class Propane
     end
 
     it "parses - as a plain character when escaped in middle of character class" do
-      regex = Regex.new("[0\\-9]")
+      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
@@ -269,7 +286,7 @@ class Propane
     end
 
     it "parses alternates" do
-      regex = Regex.new("ab|c")
+      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
@@ -279,7 +296,7 @@ class Propane
     end
 
     it "parses a ." do
-      regex = Regex.new("a.b")
+      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
@@ -290,7 +307,7 @@ class Propane
     end
 
     it "parses something complex" do
-      regex = Regex.new("(a|)*|[^^]|\\|v|[x-y]+")
+      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

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, 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) { 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_ { 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

@@ -1,11 +1,32 @@
-require "bundler/setup"
-require "propane"
+unless ENV["dist_specs"]
+  require "bundler/setup"
+  require "simplecov"
 
-RSpec.configure do |config|
-  # Enable flags like --only-failures and --next-failure
-  config.example_status_persistence_file_path = ".rspec_status"
+  class MyFormatter
+    def format(*args)
+    end
+  end
+  SimpleCov.start do
+    add_filter "/spec/"
+    add_filter "/.bundle/"
+    if ENV["partial_specs"]
+      command_name "RSpec-partial"
+    else
+      command_name "RSpec"
+    end
+    project_name "Propane"
+    merge_timeout 3600
+    formatter(MyFormatter)
+  end
 
-  config.expect_with :rspec do |c|
-    c.syntax = :expect
+  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_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_d_lexer.d

@@ -1,52 +0,0 @@
-import testparser;
-import std.stdio;
-
-int main()
-{
-    return 0;
-}
-
-unittest
-{
-    alias Result = Testparser.Decoder.Result;
-    Result result;
-
-    result = Testparser.Decoder.decode_code_point("5");
-    assert(result == Result.success('5', 1u));
-
-    result = Testparser.Decoder.decode_code_point("");
-    assert(result == Result.eof());
-
-    result = Testparser.Decoder.decode_code_point("\xC2\xA9");
-    assert(result == Result.success(0xA9u, 2u));
-
-    result = Testparser.Decoder.decode_code_point("\xf0\x9f\xa7\xa1");
-    assert(result == Result.success(0x1F9E1, 4u));
-
-    result = Testparser.Decoder.decode_code_point("\xf0\x9f\x27");
-    assert(result == Result.decode_error());
-
-    result = Testparser.Decoder.decode_code_point("\xf0\x9f\xa7\xFF");
-    assert(result == Result.decode_error());
-
-    result = Testparser.Decoder.decode_code_point("\xfe");
-    assert(result == Result.decode_error());
-}
-
-unittest
-{
-    alias Result = Testparser.Lexer.Result;
-    string input = "5 + 4 * \n677 + 567";
-    Testparser.Lexer lexer = new Testparser.Lexer(input);
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 0, 0, 1, Testparser.TOKEN_int));
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 0, 2, 1, Testparser.TOKEN_plus));
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 0, 4, 1, Testparser.TOKEN_int));
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 0, 6, 1, Testparser.TOKEN_times));
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 1, 0, 3, Testparser.TOKEN_int));
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 1, 4, 1, Testparser.TOKEN_plus));
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 1, 6, 3, Testparser.TOKEN_int));
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 1, 9, 0, Testparser.TOKEN___EOF));
-
-    lexer = new Testparser.Lexer("");
-    assert(lexer.lex_token() == Result(Result.Type.TOKEN, 0, 0, 0, Testparser.TOKEN___EOF));
-}

spec/test_d_parser_identical_rules_lookahead.d

@@ -1,18 +0,0 @@
-import testparser;
-import std.stdio;
-
-int main()
-{
-    return 0;
-}
-
-unittest
-{
-    string input = "aba";
-    auto parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
-
-    input = "abb";
-    parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
-}

spec/test_d_parser_rule_from_multiple_states.d

@@ -1,22 +0,0 @@
-import testparser;
-import std.stdio;
-
-int main()
-{
-    return 0;
-}
-
-unittest
-{
-    string input = "a";
-    auto parser = new Testparser.Parser(input);
-    assert(parser.parse() == false);
-
-    input = "a b";
-    parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
-
-    input = "bb";
-    parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
-}

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);
+    }
+}

spec/test_input_index.rs

@@ -0,0 +1,28 @@
+use testparser::*;
+
+fn main() {
+    let mut c = p_context_new(b"ab");
+    assert_eq!(0, p_input_index(&c));
+    p_context_delete(c);
+
+    let mut c = p_context_new(b"a b");
+    let mut ti = p_token_info_t::default();
+    assert_eq!(P_SUCCESS, p_lex(&mut c, &mut ti));
+    assert_eq!(TOKEN_a, ti.token);
+    assert_eq!(1, p_input_index(&c));
+    assert_eq!(P_SUCCESS, p_lex(&mut c, &mut ti));
+    assert_eq!(TOKEN_b, ti.token);
+    assert_eq!(3, p_input_index(&c));
+    p_context_delete(c);
+
+    let mut c = p_context_new(b"ab");
+    assert_eq!(P_SUCCESS, p_parse_Start(&mut c));
+    assert_eq!(2, p_input_index(&c));
+    p_context_delete(c);
+
+    let mut c = p_context_new(b"abb");
+    let follow = [TOKEN_b];
+    assert_eq!(P_SUCCESS, p_parse_inner_Start(&mut c, &follow));
+    assert_eq!(2, p_input_index(&c));
+    p_context_delete(c);
+}

spec/test_lexer.c

@@ -0,0 +1,112 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+
+int main()
+{
+    size_t result;
+    p_code_point_t code_point;
+    uint8_t code_point_length;
+
+    result = p_decode_code_point((uint8_t const *)"5", 1u, &code_point, &code_point_length);
+    assert(result == P_SUCCESS);
+    assert(code_point == '5');
+    assert(code_point_length == 1u);
+
+    result = p_decode_code_point((uint8_t const *)"", 0u, &code_point, &code_point_length);
+    assert(result == P_EOF);
+
+    result = p_decode_code_point((uint8_t const *)"\xC2\xA9", 2u, &code_point, &code_point_length);
+    assert(result == P_SUCCESS);
+    assert(code_point == 0xA9u);
+    assert(code_point_length == 2u);
+
+    result = p_decode_code_point((uint8_t const *)"\xf0\x9f\xa7\xa1", 4u, &code_point, &code_point_length);
+    assert(result == P_SUCCESS);
+    assert(code_point == 0x1F9E1u);
+    assert(code_point_length == 4u);
+
+    result = p_decode_code_point((uint8_t const *)"\xf0\x9f\x27", 3u, &code_point, &code_point_length);
+    assert(result == P_DECODE_ERROR);
+
+    result = p_decode_code_point((uint8_t const *)"\xf0\x9f\xa7\xFF", 4u, &code_point, &code_point_length);
+    assert(result == P_DECODE_ERROR);
+
+    result = p_decode_code_point((uint8_t const *)"\xfe", 1u, &code_point, &code_point_length);
+    assert(result == P_DECODE_ERROR);
+
+
+    p_token_info_t token_info;
+    char const * input = "5 + 4 * \n677 + 567";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 1u);
+    assert(token_info.position.col == 1u);
+    assert(token_info.end_position.row == 1u);
+    assert(token_info.end_position.col == 1u);
+    assert(token_info.length == 1u);
+    assert(token_info.token == TOKEN_int);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 1u);
+    assert(token_info.position.col == 3u);
+    assert(token_info.end_position.row == 1u);
+    assert(token_info.end_position.col == 3u);
+    assert(token_info.length == 1u);
+    assert(token_info.token == TOKEN_plus);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 1u);
+    assert(token_info.position.col == 5u);
+    assert(token_info.end_position.row == 1u);
+    assert(token_info.end_position.col == 5u);
+    assert(token_info.length == 1u);
+    assert(token_info.token == TOKEN_int);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 1u);
+    assert(token_info.position.col == 7u);
+    assert(token_info.end_position.row == 1u);
+    assert(token_info.end_position.col == 7u);
+    assert(token_info.length == 1u);
+    assert(token_info.token == TOKEN_times);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 2u);
+    assert(token_info.position.col == 1u);
+    assert(token_info.end_position.row == 2u);
+    assert(token_info.end_position.col == 3u);
+    assert(token_info.length == 3u);
+    assert(token_info.token == TOKEN_int);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 2u);
+    assert(token_info.position.col == 5u);
+    assert(token_info.end_position.row == 2u);
+    assert(token_info.end_position.col == 5u);
+    assert(token_info.length == 1u);
+    assert(token_info.token == TOKEN_plus);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 2u);
+    assert(token_info.position.col == 7u);
+    assert(token_info.end_position.row == 2u);
+    assert(token_info.end_position.col == 9u);
+    assert(token_info.length == 3u);
+    assert(token_info.token == TOKEN_int);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 2u);
+    assert(token_info.position.col == 10u);
+    assert(token_info.end_position.row == 2u);
+    assert(token_info.end_position.col == 10u);
+    assert(token_info.length == 0u);
+    assert(token_info.token == TOKEN___EOF);
+    p_context_delete(context);
+
+    context = p_context_new((uint8_t const *)"", 0u);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.position.row == 1u);
+    assert(token_info.position.col == 1u);
+    assert(token_info.end_position.row == 1u);
+    assert(token_info.end_position.col == 1u);
+    assert(token_info.length == 0u);
+    assert(token_info.token == TOKEN___EOF);
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_lexer.d

@@ -0,0 +1,69 @@
+import testparser;
+import std.stdio;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    size_t result;
+    p_code_point_t code_point;
+    ubyte code_point_length;
+
+    result = p_decode_code_point("5", &code_point, &code_point_length);
+    assert(result == P_SUCCESS);
+    assert(code_point == '5');
+    assert(code_point_length == 1u);
+
+    result = p_decode_code_point("", &code_point, &code_point_length);
+    assert(result == P_EOF);
+
+    result = p_decode_code_point("\xC2\xA9", &code_point, &code_point_length);
+    assert(result == P_SUCCESS);
+    assert(code_point == 0xA9u);
+    assert(code_point_length == 2u);
+
+    result = p_decode_code_point("\xf0\x9f\xa7\xa1", &code_point, &code_point_length);
+    assert(result == P_SUCCESS);
+    assert(code_point == 0x1F9E1u);
+    assert(code_point_length == 4u);
+
+    result = p_decode_code_point("\xf0\x9f\x27", &code_point, &code_point_length);
+    assert(result == P_DECODE_ERROR);
+
+    result = p_decode_code_point("\xf0\x9f\xa7\xFF", &code_point, &code_point_length);
+    assert(result == P_DECODE_ERROR);
+
+    result = p_decode_code_point("\xfe", &code_point, &code_point_length);
+    assert(result == P_DECODE_ERROR);
+}
+
+unittest
+{
+    p_token_info_t token_info;
+    string input = "5 + 4 * \n677 + 567";
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(1, 1), p_position_t(1, 1), 1, TOKEN_int));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(1, 3), p_position_t(1, 3), 1, TOKEN_plus));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(1, 5), p_position_t(1, 5), 1, TOKEN_int));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(1, 7), p_position_t(1, 7), 1, TOKEN_times));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(2, 1), p_position_t(2, 3), 3, TOKEN_int));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(2, 5), p_position_t(2, 5), 1, TOKEN_plus));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(2, 7), p_position_t(2, 9), 3, TOKEN_int));
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(2, 10), p_position_t(2, 10), 0, TOKEN___EOF));
+
+    context = p_context_new("");
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info == p_token_info_t(p_position_t(1, 1), p_position_t(1, 1), 0, TOKEN___EOF));
+}

spec/test_lexer.rs

@@ -0,0 +1,49 @@
+use testparser::*;
+
+fn chk(ti: &p_token_info_t, row: u32, col: u32, erow: u32, ecol: u32, len: usize, token: p_token_t) {
+    assert_eq!(row, ti.position.row);
+    assert_eq!(col, ti.position.col);
+    assert_eq!(erow, ti.end_position.row);
+    assert_eq!(ecol, ti.end_position.col);
+    assert_eq!(len, ti.length);
+    assert_eq!(token, ti.token);
+}
+
+fn main() {
+    let mut cp: p_code_point_t = 0;
+    let mut cpl: u8 = 0;
+
+    assert_eq!(P_SUCCESS, p_decode_code_point(b"5", &mut cp, &mut cpl));
+    assert_eq!('5' as u32, cp);
+    assert_eq!(1, cpl);
+
+    assert_eq!(P_EOF, p_decode_code_point(b"", &mut cp, &mut cpl));
+
+    assert_eq!(P_SUCCESS, p_decode_code_point(b"\xC2\xA9", &mut cp, &mut cpl));
+    assert_eq!(0xA9, cp);
+    assert_eq!(2, cpl);
+
+    assert_eq!(P_SUCCESS, p_decode_code_point(b"\xf0\x9f\xa7\xa1", &mut cp, &mut cpl));
+    assert_eq!(0x1F9E1, cp);
+    assert_eq!(4, cpl);
+
+    assert_eq!(P_DECODE_ERROR, p_decode_code_point(b"\xf0\x9f\x27", &mut cp, &mut cpl));
+    assert_eq!(P_DECODE_ERROR, p_decode_code_point(b"\xf0\x9f\xa7\xFF", &mut cp, &mut cpl));
+    assert_eq!(P_DECODE_ERROR, p_decode_code_point(b"\xfe", &mut cp, &mut cpl));
+
+    let mut context = p_context_new(b"5 + 4 * \n677 + 567");
+    let mut ti = p_token_info_t::default();
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 1, 1, 1, 1, 1, TOKEN_int);
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 1, 3, 1, 3, 1, TOKEN_plus);
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 1, 5, 1, 5, 1, TOKEN_int);
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 1, 7, 1, 7, 1, TOKEN_times);
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 2, 1, 2, 3, 3, TOKEN_int);
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 2, 5, 2, 5, 1, TOKEN_plus);
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 2, 7, 2, 9, 3, TOKEN_int);
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 2, 10, 2, 10, 0, TOKEN___EOF);
+    p_context_delete(context);
+
+    let mut context = p_context_new(b"");
+    assert_eq!(P_SUCCESS, p_lex(&mut context, &mut ti)); chk(&ti, 1, 1, 1, 1, 0, TOKEN___EOF);
+    p_context_delete(context);
+}

spec/test_lexer_match_text.c

@@ -0,0 +1,16 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+#include <stdio.h>
+
+int main()
+{
+    char const * input = "identifier_123";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    printf("pass1\n");
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_lexer_match_text.d

@@ -9,7 +9,8 @@ int main()
 unittest
 {
     string input = `identifier_123`;
-    auto parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
     writeln("pass1");
 }

spec/test_lexer_match_text.rs

@@ -0,0 +1,8 @@
+use testparser::*;
+
+fn main() {
+    let mut context = p_context_new(b"identifier_123");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    println!("pass1");
+    p_context_delete(context);
+}

spec/test_lexer_modes.c

@@ -0,0 +1,22 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+#include <stdio.h>
+
+int main()
+{
+    char const * input = "abc \"a string\" def";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    printf("pass1\n");
+    p_context_delete(context);
+
+    input = "abc \"abc def\" def";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    printf("pass2\n");
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_lexer_modes.d

@@ -9,12 +9,13 @@ int main()
 unittest
 {
     string input = `abc "a string" def`;
-    auto parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
     writeln("pass1");
 
     input = `abc "abc def" def`;
-    parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
     writeln("pass2");
 }

spec/test_lexer_modes.rs

@@ -0,0 +1,13 @@
+use testparser::*;
+
+fn main() {
+    let mut context = p_context_new(b"abc \"a string\" def");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    println!("pass1");
+    p_context_delete(context);
+
+    let mut context = p_context_new(b"abc \"abc def\" def");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    println!("pass2");
+    p_context_delete(context);
+}

spec/test_lexer_multiple_modes.c

@@ -0,0 +1,22 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+#include <stdio.h>
+
+int main()
+{
+    char const * input = "abc.def";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    printf("pass1\n");
+    p_context_delete(context);
+
+    input = "abc .  abc";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    printf("pass2\n");
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_lexer_multiple_modes.d

@@ -0,0 +1,21 @@
+import testparser;
+import std.stdio;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = `abc.def`;
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+    writeln("pass1");
+
+    input = `abc .  abc`;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+    writeln("pass2");
+}

spec/test_lexer_multiple_modes.rs

@@ -0,0 +1,13 @@
+use testparser::*;
+
+fn main() {
+    let mut context = p_context_new(b"abc.def");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    println!("pass1");
+    p_context_delete(context);
+
+    let mut context = p_context_new(b"abc .  abc");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    println!("pass2");
+    p_context_delete(context);
+}

spec/test_lexer_positions.c

@@ -0,0 +1,50 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+
+int main()
+{
+    char const * input = "abc\n  defg hi\n!";
+    p_context_t * context = p_context_new((uint8_t const *)input, strlen(input));
+    p_token_info_t token_info;
+
+    /* First token "abc" on row 1, cols 1-3. */
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.token == TOKEN_word);
+    assert(context->last_start.row == 1u);
+    assert(context->last_start.col == 1u);
+    assert(context->last_end.row == 1u);
+    assert(context->last_end.col == 3u);
+    /* The lexer code block observed the same positions reported to the caller. */
+    assert(context->last_start.row == token_info.position.row);
+    assert(context->last_start.col == token_info.position.col);
+    assert(context->last_end.row == token_info.end_position.row);
+    assert(context->last_end.col == token_info.end_position.col);
+
+    /* Second token "defg" on row 2, cols 3-6. */
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.token == TOKEN_word);
+    assert(context->last_start.row == 2u);
+    assert(context->last_start.col == 3u);
+    assert(context->last_end.row == 2u);
+    assert(context->last_end.col == 6u);
+
+    /* Third token "hi" on row 2, cols 8-9. */
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.token == TOKEN_word);
+    assert(context->last_start.row == 2u);
+    assert(context->last_start.col == 8u);
+    assert(context->last_end.row == 2u);
+    assert(context->last_end.col == 9u);
+
+    /* The "!" stop token terminates the lexer. The context input text position
+     * must not be updated when the lexer user code requests termination, so it
+     * still points at the "!" token on row 3, col 1. */
+    assert(p_lex(context, &token_info) == P_USER_TERMINATED);
+    assert(p_user_terminate_code(context) == 42u);
+    assert(context->text_position.row == 3u);
+    assert(context->text_position.col == 1u);
+
+    p_context_delete(context);
+    return 0;
+}

spec/test_lexer_positions.d

@@ -0,0 +1,42 @@
+import testparser;
+import std.stdio;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = "abc\n  defg hi\n!";
+    p_context_t * context = p_context_new(input);
+    p_token_info_t token_info;
+
+    /* First token "abc" on row 1, cols 1-3. */
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.token == TOKEN_word);
+    assert(context.last_start == p_position_t(1, 1));
+    assert(context.last_end == p_position_t(1, 3));
+    /* The lexer code block observed the same positions reported to the caller. */
+    assert(context.last_start == token_info.position);
+    assert(context.last_end == token_info.end_position);
+
+    /* Second token "defg" on row 2, cols 3-6. */
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.token == TOKEN_word);
+    assert(context.last_start == p_position_t(2, 3));
+    assert(context.last_end == p_position_t(2, 6));
+
+    /* Third token "hi" on row 2, cols 8-9. */
+    assert(p_lex(context, &token_info) == P_SUCCESS);
+    assert(token_info.token == TOKEN_word);
+    assert(context.last_start == p_position_t(2, 8));
+    assert(context.last_end == p_position_t(2, 9));
+
+    /* The "!" stop token terminates the lexer. The context input text position
+     * must not be updated when the lexer user code requests termination, so it
+     * still points at the "!" token on row 3, col 1. */
+    assert(p_lex(context, &token_info) == P_USER_TERMINATED);
+    assert(p_user_terminate_code(context) == 42u);
+    assert(context.text_position == p_position_t(3, 1));
+}

spec/test_lexer_positions.rs

@@ -0,0 +1,38 @@
+use testparser::*;
+
+fn main() {
+    let mut c = p_context_new(b"abc\n  defg hi\n!");
+    let mut ti = p_token_info_t::default();
+
+    assert_eq!(P_SUCCESS, p_lex(&mut c, &mut ti));
+    assert_eq!(TOKEN_word, ti.token);
+    assert_eq!(1, c.last_start.row);
+    assert_eq!(1, c.last_start.col);
+    assert_eq!(1, c.last_end.row);
+    assert_eq!(3, c.last_end.col);
+    assert_eq!(c.last_start.row, ti.position.row);
+    assert_eq!(c.last_start.col, ti.position.col);
+    assert_eq!(c.last_end.row, ti.end_position.row);
+    assert_eq!(c.last_end.col, ti.end_position.col);
+
+    assert_eq!(P_SUCCESS, p_lex(&mut c, &mut ti));
+    assert_eq!(TOKEN_word, ti.token);
+    assert_eq!(2, c.last_start.row);
+    assert_eq!(3, c.last_start.col);
+    assert_eq!(2, c.last_end.row);
+    assert_eq!(6, c.last_end.col);
+
+    assert_eq!(P_SUCCESS, p_lex(&mut c, &mut ti));
+    assert_eq!(TOKEN_word, ti.token);
+    assert_eq!(2, c.last_start.row);
+    assert_eq!(8, c.last_start.col);
+    assert_eq!(2, c.last_end.row);
+    assert_eq!(9, c.last_end.col);
+
+    assert_eq!(P_USER_TERMINATED, p_lex(&mut c, &mut ti));
+    assert_eq!(42, p_user_terminate_code(&c));
+    assert_eq!(3, p_position(&c).row);
+    assert_eq!(1, p_position(&c).col);
+
+    p_context_delete(c);
+}

spec/test_lexer_result_value.c

@@ -0,0 +1,21 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+
+int main()
+{
+    char const * input = "x";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    assert(p_result(context) == 1u);
+    p_context_delete(context);
+
+    input = "fabulous";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    assert(p_result(context) == 8u);
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_lexer_result_value.d

@@ -9,12 +9,13 @@ int main()
 unittest
 {
     string input = `x`;
-    auto parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
-    assert(parser.result == 1u);
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+    assert(p_result(context) == 1u);
 
     input = `fabulous`;
-    parser = new Testparser.Parser(input);
-    assert(parser.parse() == true);
-    assert(parser.result == 8u);
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+    assert(p_result(context) == 8u);
 }

spec/test_lexer_result_value.rs

@@ -0,0 +1,13 @@
+use testparser::*;
+
+fn main() {
+    let mut context = p_context_new(b"x");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    assert_eq!(1, p_result(&context));
+    p_context_delete(context);
+
+    let mut context = p_context_new(b"fabulous");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    assert_eq!(8, p_result(&context));
+    p_context_delete(context);
+}

spec/test_lexer_unknown_character.c

@@ -0,0 +1,20 @@
+#include "testparser.h"
+#include <assert.h>
+#include <string.h>
+
+int main()
+{
+    char const * input = "x";
+    p_context_t * context;
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_UNEXPECTED_INPUT);
+    p_context_delete(context);
+
+    input = "123";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    assert(p_result(context) == 123u);
+    p_context_delete(context);
+
+    return 0;
+}

spec/test_lexer_unknown_character.d

@@ -0,0 +1,20 @@
+import testparser;
+import std.stdio;
+
+int main()
+{
+    return 0;
+}
+
+unittest
+{
+    string input = `x`;
+    p_context_t * context;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_UNEXPECTED_INPUT);
+
+    input = `123`;
+    context = p_context_new(input);
+    assert(p_parse(context) == P_SUCCESS);
+    assert(p_result(context) == 123u);
+}

spec/test_lexer_unknown_character.rs

@@ -0,0 +1,12 @@
+use testparser::*;
+
+fn main() {
+    let mut context = p_context_new(b"x");
+    assert_eq!(P_UNEXPECTED_INPUT, p_parse(&mut context));
+    p_context_delete(context);
+
+    let mut context = p_context_new(b"123");
+    assert_eq!(P_SUCCESS, p_parse(&mut context));
+    assert_eq!(123, p_result(&context));
+    p_context_delete(context);
+}

spec/test_macros.c

@@ -0,0 +1,118 @@
+#include "testparser.h"
+#include "testutils.h"
+#include <string.h>
+#include <assert.h>
+#include <stddef.h>
+#include <stdbool.h>
+
+static p_context_t * context;
+size_t n_tokens;
+p_token_info_t token_infos[10];
+
+/* Capture the macro body tokens (everything up to the closing '}') into
+ * token_infos[]. Called from mylexfn() right after the definition's '{' has
+ * been lexed, so the input cursor is positioned at the first body token. */
+static void capture_macro_body(void)
+{
+    n_tokens = 0u;
+    for (;;)
+    {
+        size_t result = p_lex(context, &token_infos[n_tokens]);
+        assert_eq(result, P_SUCCESS);
+        if (token_infos[n_tokens].token == TOKEN_rbrace)
+        {
+            break;
+        }
+        n_tokens++;
+        assert(n_tokens < sizeof(token_infos) / sizeof(token_infos[0]));
+    }
+}
+
+size_t mylexfn(p_context_t * context, p_token_info_t * out_token_info)
+{
+    static bool defining;
+    static bool expanding;
+    static size_t expand_i;
+
+    for (;;)
+    {
+        if (expanding)
+        {
+            size_t ei = expand_i++;
+            if (expand_i >= n_tokens)
+            {
+                expanding = false;
+            }
+            *out_token_info = token_infos[ei];
+            return P_SUCCESS;
+        }
+
+        size_t lex_result = p_lex(context, out_token_info);
+        if (lex_result != P_SUCCESS)
+        {
+            return lex_result;
+        }
+
+        switch (out_token_info->token)
+        {
+        case TOKEN_macro:
+            /* Start of a macro definition: "macro macroname { ... }". */
+            defining = true;
+            break;
+        case TOKEN_macroname:
+            if (!defining)
+            {
+                /* Use of a macro: replay its captured body tokens instead of
+                 * returning the macroname to the parser. */
+                expanding = true;
+                expand_i = 0u;
+                continue;
+            }
+            /* Definition name: pass through and keep waiting for '{'. */
+            break;
+        case TOKEN_lbrace:
+            if (defining)
+            {
+                /* Consume and store the macro body now, before the parser gets
+                 * a chance to read its lookahead token (which would otherwise
+                 * swallow the first body token). */
+                capture_macro_body();
+                defining = false;
+            }
+            break;
+        default:
+            defining = false;
+            break;
+        }
+        return lex_result;
+    }
+}
+
+size_t n_nums;
+int nums[10];
+
+void record(int v)
+{
+    nums[n_nums++] = v;
+}
+
+int main()
+{
+    char const * input =
+        "macro @m { 23 + 200 }\n"
+        "66 + 100\n"
+        "@m\n"
+        "33 + 55\n"
+        "@m\n";
+    context = p_context_new((uint8_t const *)input, strlen(input));
+    assert(p_parse(context) == P_SUCCESS);
+    p_context_delete(context);
+
+    assert_eq(n_nums, 4);
+    assert_eq(nums[0], 166);
+    assert_eq(nums[1], 223);
+    assert_eq(nums[2], 88);
+    assert_eq(nums[3], 223);
+
+    return 0;
+}