Exploring re2, doesn't really work

src/actions.cr

@@ -1,3 +1,5 @@
+require "xml"
+
 # These are Lexer actions. When a rule matches, it will
 # perform a list of actions. These actions can emit tokens
 # or change the state machine.
@@ -24,11 +26,11 @@ module Tartrazine
     end
 
     # ameba:disable Metrics/CyclomaticComplexity
-    def emit(match : Regex::MatchData?, lexer : Lexer, match_group = 0) : Array(Token)
+    def emit(matches : Pointer(LibCre2::StringPiece), lexer : Lexer, match_group = 0) : Array(Token)
       case type
       when "token"
-        raise Exception.new "Can't have a token without a match" if match.nil?
+        raise Exception.new "Can't have a token without a match" if matches[0].length == 0
-        [Token.new(type: xml["type"], value: match[match_group])]
+        [Token.new(type: xml["type"], value: String.new(Slice.new(matches[0].data, matches[0].length)))]
       when "push"
         states_to_push = xml.attributes.select { |attrib|
           attrib.name == "state"
@@ -61,35 +63,37 @@ module Tartrazine
       when "bygroups"
         # FIXME: handle
         # ><bygroups>
-        # <token type="Punctuation"/>
+        # <token type="Punctuation"/>https://github.com/google/re2/wiki/Syntax
         # None
         # <token type="LiteralStringRegex"/>
         #
         # where that None means skipping a group
         #
-        raise Exception.new "Can't have a token without a match" if match.nil?
+        raise Exception.new "Can't have a bygroups without a match" if matches[0].length == 0
 
         # Each group matches an action. If the group match is empty,
         # the action is skipped.
         result = [] of Token
         @actions.each_with_index do |e, i|
-          next if match[i + 1]?.nil?
+          next if matches[i].length == 0
-          result += e.emit(match, lexer, i + 1)
+          result += e.emit(matches, lexer, i)
         end
         result
       when "using"
         # Shunt to another lexer entirely
-        return [] of Token if match.nil?
+        return [] of Token if matches[0].length == 0
         lexer_name = xml["lexer"].downcase
-        Log.trace { "to tokenize: #{match[match_group]}" }
+        # Log.trace { "to tokenize: #{match[match_group]}" }
-        Tartrazine.lexer(lexer_name).tokenize(match[match_group], usingself: true)
+        to_tokenize = String.new(Slice.new(matches[match_group].data, matches[match_group].length))
+        Tartrazine.lexer(lexer_name).tokenize(to_tokenize, usingself: true)
       when "usingself"
         # Shunt to another copy of this lexer
-        return [] of Token if match.nil?
+        return [] of Token if matches[0].length == 0
 
         new_lexer = Lexer.from_xml(lexer.xml)
-        Log.trace { "to tokenize: #{match[match_group]}" }
+        # Log.trace { "to tokenize: #{match[match_group]}" }
-        new_lexer.tokenize(match[match_group], usingself: true)
+        to_tokenize = String.new(Slice.new(matches[match_group].data, matches[match_group].length))
+        new_lexer.tokenize(to_tokenize, usingself: true)
       when "combined"
         # Combine two states into one anonymous state
         states = xml.attributes.select { |attrib|

src/cre2/Makefile

@@ -0,0 +1,5 @@
+all: cre2.o
+clean:
+	rm -f cre2.o
+cre2.o: cre2.cpp cre2.h
+	g++ -O3 -c -o cre2.o cre2.cpp

src/cre2/cre2.cpp

@@ -0,0 +1,121 @@
+#include <re2/re2.h>
+#include "cre2.h"
+
+#define TO_OPT(opt) (reinterpret_cast<RE2::Options *>(opt))
+
+cre2_options *cre2_opt_new(void) {
+    return reinterpret_cast<void*>(new RE2::Options());
+}
+
+void cre2_opt_delete(cre2_options *opt) {
+    delete TO_OPT(opt);
+}
+
+
+#define OPT_bool(name)  \
+void cre2_opt_##name(cre2_options *opt, int flag) {  \
+    TO_OPT(opt)->set_##name(bool(flag));             \
+}
+
+OPT_bool(posix_syntax)
+OPT_bool(longest_match)
+OPT_bool(log_errors)
+OPT_bool(literal)
+OPT_bool(never_nl)
+OPT_bool(case_sensitive)
+OPT_bool(perl_classes)
+OPT_bool(word_boundary)
+OPT_bool(one_line)
+
+#undef OPT_BOOL
+
+
+void cre2_opt_encoding(cre2_options *opt, encoding_t enc) {
+    switch (enc) {
+        case CRE2_UTF8:
+            TO_OPT(opt)->set_encoding(RE2::Options::EncodingUTF8);
+            break;
+        case CRE2_Latin1:
+            TO_OPT(opt)->set_encoding(RE2::Options::EncodingLatin1);
+            break;
+    }
+}
+
+void cre2_opt_max_mem(cre2_options *opt, int m) {
+    TO_OPT(opt)->set_max_mem(m);
+}
+
+
+#define TO_RE2(re)       (reinterpret_cast<RE2 *>(re))
+#define TO_CONST_RE2(re) (reinterpret_cast<const RE2 *>(re))
+
+cre2 *cre2_new(const char *pattern, int patternlen, const cre2_options *opt) {
+    re2::StringPiece pattern_re2(pattern, patternlen);
+    return reinterpret_cast<void*>(
+        new RE2(pattern_re2, *reinterpret_cast<const RE2::Options *>(opt)));
+}
+
+void cre2_delete(cre2 *re) {
+    delete TO_RE2(re);
+}
+
+
+int cre2_error_code(const cre2 *re) {
+    return int(TO_CONST_RE2(re)->error_code());
+}
+
+const char *cre2_error_string(const cre2 *re) {
+    return TO_CONST_RE2(re)->error().c_str();
+}
+
+void cre2_error_arg(const cre2 *re, struct string_piece *arg) {
+    const std::string &argstr = TO_CONST_RE2(re)->error_arg();
+    arg->data   = argstr.data();
+    arg->length = argstr.length();
+}
+
+int cre2_num_capturing_groups(const cre2 *re) {
+    return TO_CONST_RE2(re)->NumberOfCapturingGroups();
+}
+
+int cre2_program_size(const cre2 *re) {
+    return TO_CONST_RE2(re)->ProgramSize();
+}
+
+
+int cre2_match(
+    const cre2 *re
+  , const char *text
+  , int textlen
+  , int startpos
+  , int endpos
+  , anchor_t anchor
+  , struct string_piece *match
+  , int nmatch) {
+
+    re2::StringPiece text_re2(text, textlen);
+    // FIXME: exceptions?
+    re2::StringPiece *match_re2 = new re2::StringPiece[nmatch];
+
+    RE2::Anchor anchor_re2 = RE2::UNANCHORED;
+    switch (anchor) {
+        case CRE2_ANCHOR_START:
+            anchor_re2 = RE2::ANCHOR_START; break;
+        case CRE2_ANCHOR_BOTH:
+            anchor_re2 = RE2::ANCHOR_BOTH;  break;
+    }
+
+    bool ret = TO_CONST_RE2(re)
+        ->Match(text_re2, startpos, endpos, anchor_re2, match_re2, nmatch);
+
+    if (ret) {
+        for (int i=0; i<nmatch; i++) {
+            match[i].data   = match_re2[i].data();
+            match[i].length = match_re2[i].length();
+        }
+    }
+
+    delete [] match_re2;
+
+    return int(ret);
+}

src/cre2/cre2.cr

@@ -0,0 +1,69 @@
+@[Link(ldflags: "#{__DIR__}/cre2.o -Wl,--copy-dt-needed-entries `pkg-config --libs re2`")]
+lib LibCre2
+  type Options = Void*
+
+  fun opt_new = cre2_opt_new : Options
+  fun opt_delete = cre2_opt_delete(op : Options) : Nil
+
+  fun opt_posix_syntax = cre2_opt_posix_syntax(op : Options, flag : Bool) : Nil
+  fun opt_longest_match = cre2_opt_longest_match(op : Options, flag : Bool) : Nil
+  fun opt_log_errors = cre2_opt_log_errors(op : Options, flag : Bool) : Nil
+  fun opt_literal = cre2_opt_literal(op : Options, flag : Bool) : Nil
+  fun opt_never_nl = cre2_opt_never_nl(op : Options, flag : Bool) : Nil
+  fun opt_case_sensitive = cre2_opt_case_sensitive(op : Options, flag : Bool) : Nil
+  fun opt_perl_classes = cre2_opt_perl_classes(op : Options, flag : Bool) : Nil
+  fun opt_word_boundary = cre2_opt_word_boundary(op : Options, flag : Bool) : Nil
+  fun opt_one_line = cre2_opt_one_line(op : Options, flag : Bool) : Nil
+  fun opt_encoding = cre2_opt_encoding(op : Options, encoding : Int32) : Nil
+  fun opt_max_mem = cre2_opt_max_mem(op : Options, flag : Bool) : Nil
+
+  struct StringPiece
+    data : LibC::Char*
+    length : Int32
+  end
+
+  type CRe2 = Void*
+
+  fun new = cre2_new(pattern : LibC::Char*, patternlen : UInt32, opt : Options) : CRe2
+  fun del = cre2_delete(re : CRe2) : Nil
+  fun error_code = cre2_error_core(re : CRe2) : Int32
+  fun num_capturing_groups(re : CRe2) : Int32
+  fun program_size(re : CRe2) : Int32
+
+  # Invalidated by further re use
+  fun error_string = cre2_error_string(re : CRe2) : LibC::Char*
+  fun error_arg = cre2_error_arg(re : CRe2, arg : StringPiece*) : Nil
+
+  CRE2_UNANCHORED   = 1
+  CRE2_ANCHOR_START = 2
+  CRE2_ANCHOR_BOTH  = 3
+
+  fun match = cre2_match(
+    re : CRe2,
+    text : LibC::Char*,
+    textlen : UInt32,
+    startpos : UInt32,
+    endpos : UInt32,
+    anchor : Int32,
+    match : StringPiece*,
+    nmatch : Int32
+  ) : Int32
+end
+
+# match = Pointer(LibCre2::StringPiece).malloc(10)
+# opts = LibCre2.opt_new
+# LibCre2.opt_posix_syntax(opts, true)
+# LibCre2.opt_longest_match(opts, true)
+# LibCre2.opt_perl_classes(opts, true)
+# LibCre2.opt_encoding(opts, 1)
+# # LibCre2.opt_one_line(opts, false)
+# # LibCre2.opt_never_nl(opts, false)
+
+# pattern = "(\\s+)(foo)"
+# text = "  foo"
+# re = LibCre2.new(pattern, pattern.size, opts)
+# p! LibCre2.match(re, text, text.size, 0, text.size,
+#   LibCre2::CRE2_ANCHOR_START, match, 10)
+# (0...10).each do |i|
+#   p! String.new(Slice.new(match[i].data, match[i].length))
+# end

src/cre2/cre2.h

@@ -0,0 +1,66 @@
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+typedef void cre2_options;
+
+typedef int encoding_t;
+#define CRE2_UTF8   1
+#define CRE2_Latin1 2
+
+cre2_options *cre2_opt_new(void);
+void cre2_opt_delete(cre2_options *opt);
+
+void cre2_opt_posix_syntax(cre2_options *opt, int flag);
+void cre2_opt_longest_match(cre2_options *opt, int flag);
+void cre2_opt_log_errors(cre2_options *opt, int flag);
+void cre2_opt_literal(cre2_options *opt, int flag);
+void cre2_opt_never_nl(cre2_options *opt, int flag);
+void cre2_opt_case_sensitive(cre2_options *opt, int flag);
+void cre2_opt_perl_classes(cre2_options *opt, int flag);
+void cre2_opt_word_boundary(cre2_options *opt, int flag);
+void cre2_opt_one_line(cre2_options *opt, int flag);
+void cre2_opt_encoding(cre2_options *opt, encoding_t enc);
+void cre2_opt_max_mem(cre2_options *opt, int m);
+
+
+struct string_piece {
+    const char *data;
+    int length;
+};
+
+
+typedef void cre2;
+
+cre2 *cre2_new(const char *pattern, int patternlen, const cre2_options *opt);
+void cre2_delete(cre2 *re);
+
+int cre2_error_code(const cre2 *re);
+int cre2_num_capturing_groups(const cre2 *re);
+int cre2_program_size(const cre2 *re);
+
+// invalidated by further re use
+const char *cre2_error_string(const cre2 *re);
+void cre2_error_arg(const cre2 *re, struct string_piece *arg);
+
+
+typedef int anchor_t;
+#define CRE2_UNANCHORED   1
+#define CRE2_ANCHOR_START 2
+#define CRE2_ANCHOR_BOTH  3
+
+int cre2_match(
+      const cre2 *re
+    , const char *text
+    , int textlen
+    , int startpos
+    , int endpos
+    , anchor_t anchor
+    , struct string_piece *match
+    , int nmatch);
+
+
+#ifdef __cplusplus
+} // extern "C"
+#endif

src/rules.cr

@@ -1,3 +1,4 @@
+require "./cre2/cre2"
 require "./actions"
 
 # These are lexer rules. They match with the text being parsed
@@ -7,29 +8,32 @@ module Tartrazine
   # This rule matches via a regex pattern
 
   class Rule
-    property pattern : Regex = Re2.new ""
+    property pattern : Re3 = Re3.new ""
     property actions : Array(Action) = [] of Action
     property xml : String = "foo"
 
     def match(text, pos, lexer) : Tuple(Bool, Int32, Array(Token))
-      match = pattern.match(text, pos)
+      return false, pos, [] of Token
+      matched, matches = pattern.match(text, pos)
       # We don't match if the match doesn't move the cursor
       # because that causes infinite loops
-      return false, pos, [] of Token if match.nil? || match.end == 0
+
+      return false, pos, [] of Token unless matched
       # Log.trace { "#{match}, #{pattern.inspect}, #{text}, #{pos}" }
       tokens = [] of Token
       # Emit the tokens
       actions.each do |action|
         # Emit the token
-        tokens += action.emit(match, lexer)
+        tokens += action.emit(matches, lexer)
       end
-      Log.trace { "#{xml}, #{match.end}, #{tokens}" }
+      # Log.trace { "#{xml}, #{match.end}, #{tokens}" }
-      return true, match.end, tokens
+      return true, matches[0].length, tokens
     end
 
     def initialize(node : XML::Node, multiline, dotall, ignorecase)
       @xml = node.to_s
-      @pattern = Re2.new(
+      p! node["pattern"]
+      @pattern = Re3.new(
         node["pattern"],
         multiline,
         dotall,
@@ -76,7 +80,7 @@ module Tartrazine
     def match(text, pos, lexer) : Tuple(Bool, Int32, Array(Token))
       tokens = [] of Token
       actions.each do |action|
-        tokens += action.emit(nil, lexer)
+        tokens += action.emit(Pointer(LibCre2::StringPiece).malloc(1), lexer)
       end
       return true, pos, tokens
     end
@@ -106,4 +110,38 @@ module Tartrazine
       end
     end
   end
+
+  class Re3
+    @matches = Pointer(LibCre2::StringPiece).malloc(50)
+    @opts : LibCre2::Options
+
+    @re : LibCre2::CRe2
+
+    def group_count
+      LibCre2.num_capturing_groups(@re)
+    end
+
+    def initialize(pattern : String, multiline = false, dotall = false,
+                   ignorecase = false, anchored = false)
+      @opts = LibCre2.opt_new
+      LibCre2.opt_posix_syntax(@opts, false)
+      LibCre2.opt_longest_match(@opts, true)
+      LibCre2.opt_perl_classes(@opts, true)
+      LibCre2.opt_encoding(@opts, 1)
+      LibCre2.opt_one_line(@opts, !multiline)
+      LibCre2.opt_case_sensitive(@opts, !ignorecase)
+      pattern = "(m?)#{pattern}" if multiline
+      @re = LibCre2.new(pattern, pattern.size, @opts)
+    end
+
+    def match(text, pos)
+      matched = LibCre2.match(@re, text, text.size, pos, text.size,
+        LibCre2::CRE2_ANCHOR_START, @matches, 50)
+      return {matched != 0, @matches}
+    end
+  end
 end
+
+re = Tartrazine::Re3.new("(require-instance|fraction-digits|error-app-tag|error-message|min-elements|max-elements|yin-element|ordered-by|position|modifier|default|pattern|length|status|units|value|range|type|path|enum|base|bit)(?=[^\w\-\:])")
+p! re.match("value ", 0)
+

src/tartrazine.cr

@@ -63,8 +63,6 @@ module Tartrazine
       tokens = [] of Token
       pos = 0
       matched = false
-      time = 0
-      count = 0
 
       # Respect the `ensure_nl` config option
       if text.size > 0 && text[-1] != '\n' && config[:ensure_nl] && !usingself