diff --git a/Makefile b/Makefile
index 48dd1d7..0875321 100644
--- a/Makefile
+++ b/Makefile
@@ -5,6 +5,7 @@ CXXDEPS  := $(CXXOBJS:.o=.dep)
 CXXFLAGS := -O2
 DEPS     := $(CXXDEPS)
 OBJS     := $(CXXOBJS)
+LDFLAGS  := -lpcrecpp
 
 all: $(TARGET)
 
diff --git a/parse-input.cc b/parse-input.cc
index ecce14e..86e302e 100644
--- a/parse-input.cc
+++ b/parse-input.cc
@@ -1,15 +1,91 @@
 
 #include <stdio.h>
-#include <ctype.h>  /* isspace() */
+#include <string.h>
+
+#include <iostream>
+#include <string>
+#include <map>
+#include <pcrecpp.h>
+
 #include "parse-input.h"
+
 using namespace std;
+using namespace pcrecpp;
 
-#define SET_ERROR(err, args...) \
-    do { \
-        error = true; \
-        sprintf(errstr, err " at line %d, column %d", ##args, lineno, colno); \
-    } while(0)
+enum Section{ none, tokens, rules };
 
-void parse_input(char * buff, int size)
+bool parse_input(char * buff, int size)
 {
+    RE empty("\\s*");
+    RE comment("\\s*#.*");
+    RE section_name("\\s*\\[([^\\]]+?)\\]\\s*");
+    RE token("\\s*"                             /* possible leading ws */
+             "([a-zA-Z_][a-zA-Z_0-9]*)"         /* token name */
+             "\\s+"                             /* required whitespace */
+             "((?:[^\\\\\\s]|\\\\.)+)"          /* token regular expression */
+             "(?:\\s+\\[([^\\]]+)\\])?"         /* optional token flags */
+             "\\s*");                           /* possible trailing ws */
+    RE rule("\\s*(\\S+)\\s*:=(.*)");
+
+    Section section = none;
+
+    int lineno = 1;
+    char * newline;
+    char * input = buff;
+    string sn;
+    map<string, Section> sections;
+    sections["none"] = none;
+    sections["tokens"] = tokens;
+    sections["rules"] = rules;
+    while ((newline = strstr(input, "\n")) != NULL)
+    {
+        int line_length = newline - input;
+        if (newline[-1] == '\r')
+        {
+            line_length--;
+        }
+        string line(input, line_length);
+        if (empty.FullMatch(line))
+            continue;
+        if (comment.FullMatch(line))
+            continue;
+        if (section_name.FullMatch(line, &sn))
+        {
+            if (sections.find(sn) != sections.end())
+            {
+                section = sections[sn];
+            }
+            else
+            {
+                cerr << "Unknown section name '" << sn << "'!" << endl;
+                return false;
+            }
+            continue;
+        }
+        switch (section)
+        {
+            case none:
+                break;
+            case tokens:
+                {
+                    string name, definition, flags;
+                    if (token.FullMatch(line, &name, &definition, &flags))
+                    {
+                        /* TODO: process token */
+                    }
+                }
+                break;
+            case rules:
+                {
+                    string name, definition;
+                    if (rule.FullMatch(line, &name, &definition))
+                    {
+                        /* TODO: process rule */
+                    }
+                }
+                break;
+        }
+        input = newline + 1;
+        lineno++;
+    }
 }
diff --git a/parse-input.h b/parse-input.h
index ede8a9e..dbf704a 100644
--- a/parse-input.h
+++ b/parse-input.h
@@ -5,6 +5,6 @@
 #include <vector>
 #include "refptr/refptr.h"
 
-void parse_input(char * buff, int size);
+bool parse_input(char * buff, int size);
 
 #endif