aboutsummaryrefslogblamecommitdiffstats
path: root/lib/xmerl/src/xmerl_regexp.erl
blob: 566b77725f3027e835c6b81f36999327cd64fffd (plain) (tree)
1
2
3
4
5
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598


                   
                                                        
   










                                                                           



























                                                                              

                               







































































































































































































































































































































































































































































































































































                                                                               
 

























































































































































































































































                                                                               
                                           






















































                                                                            
                                   



























































                                                                       
 































































                                                                             
 





























































































                                                                             
                               


















                                                             
                                 




                                       
 


                                                                    
                                                                 







































                                                                               
                                                                      





































































                                                                                  
                                                                   

















                                                                   
 















































                                                                               
 
















































































                                                                                  
                                                 




















                                                                        
%%
%% %CopyrightBegin%
%% 
%% Copyright Ericsson AB 2006-2016. All Rights Reserved.
%% 
%% Licensed under the Apache License, Version 2.0 (the "License");
%% you may not use this file except in compliance with the License.
%% You may obtain a copy of the License at
%%
%%     http://www.apache.org/licenses/LICENSE-2.0
%%
%% Unless required by applicable law or agreed to in writing, software
%% distributed under the License is distributed on an "AS IS" BASIS,
%% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
%% See the License for the specific language governing permissions and
%% limitations under the License.
%% 
%% %CopyrightEnd%
%%

%%
-module(xmerl_regexp).

%% This module provides a basic set of regular expression functions
%% for strings. The functions provided are taken from AWK.
%%
%% Note that we interpret the syntax tree of a regular expression
%% directly instead of converting it to an NFA and then interpreting
%% that. This method seems to go significantly faster.

-export([sh_to_awk/1,parse/1,format_error/1,match/2,first_match/2,matches/2]).
-export([sub/3,gsub/3,split/2,sub_match/2,sub_first_match/2]).

-export([make_nfa/1,make_dfa/1,make_dfa/2,compile/1]).

-import(string, [substr/2,substr/3]).
-import(lists, [reverse/1,reverse/2,last/1,duplicate/2,seq/2]).
-import(lists, [member/2,keysearch/3,keysort/2,map/2,foldl/3]).
-import(ordsets, [is_element/2,add_element/2,union/2,subtract/2]).

%%-compile([export_all]).

-export([setup/1,compile_proc/2]).

-include("xmerl_internal.hrl").

setup(RE0) ->
    RE = setup(RE0, [$^]),
    Pid = spawn(?MODULE,compile_proc,[self(),RE]),
    receive
	{ok,Result} ->
	    Result
    after 2000 ->
	    exit(Pid,force),
	    parse(RE)
    end.
    %% compile(RE).
%%RE.
compile_proc(From,RE) ->
    Res = compile(RE),
    From ! {ok,Res}.


setup([$\\,$d|S],Acc) -> setup(S,"]9-0[" ++Acc);
setup([$\\,$D|S],Acc) -> setup(S,"]9-0^[" ++Acc);
setup([$\\,$s|S],Acc) -> setup(S,"]s\\t\\n\\r\\[" ++Acc);
setup([$\\,$S|S],Acc) -> setup(S,"]\\s\\t\\n\\r^[" ++Acc);
setup([$\\,$i|S],Acc) -> setup(S,"]z-aZ-A_:[" ++Acc);   %% Only Latin-1 now
setup([$\\,$I|S],Acc) -> setup(S,"]z-aZ-A_:^[" ++Acc);
setup([$\\,$c|S],Acc) -> setup(S,"]9-0z-aZ-A_:."++[183]++"-[" ++Acc); 
setup([$\\,$C|S],Acc) -> setup(S,"]9-0z-aZ-A_:."++[183]++"-^[" ++Acc);
%% fixme setup([$\\,$w|S]) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup([$\\,$W|S]) -> {{comp_class,"\s\t\n\r"},S};
%% Letter, Any
%% fixme setup(["\\p{L}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{L}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Letter, Uppercase
%% fixme setup(["\\p{Lu}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Lu}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Letter, Lowercase
%% fixme setup(["\\p{Ll}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Ll}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Letter, Titlecase
%% fixme setup(["\\p{Lt}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Lt}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Letter, Modifier
%% fixme setup(["\\p{Lm}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Lm}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Letter, Other
%% fixme setup(["\\p{Lo}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Lo}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Mark, Any
%% fixme setup(["\\p{M}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{M}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Mark, Nonspacing
%% fixme setup(["\\p{Mn}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Mn}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Mark, Spacing Combining
%% fixme setup(["\\p{Mc}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Mc}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Mark, Enclosing
%% fixme setup(["\\p{Me}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Me}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Number, Any
%% fixme setup(["\\p{N}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{N}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Number, Decimal Digit
%% fixme setup(["\\p{Nd}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Nd}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Number, Letter
%% fixme setup(["\\p{Nl}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Nl}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Number, Other
%% fixme setup(["\\p{No}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{No}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Any
%% fixme setup(["\\p{P}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{P}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Connector
%% fixme setup(["\\p{Pc}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Pc}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Dash
%% fixme setup(["\\p{Pd}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Pd}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Open
%% fixme setup(["\\p{Ps}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Ps}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Close
%% fixme setup(["\\p{Pe}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Pe}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Initial quote (may behave like Ps or Pe, depending on usage)
%% fixme setup(["\\p{Pi}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Pi}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Final quote (may behave like Ps or Pe, depending on usage)
%% fixme setup(["\\p{Pf}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Pf}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Punctuation, Other
%% fixme setup(["\\p{Po}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Po}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Symbol, Any
%% fixme setup(["\\p{S}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{S}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Symbol, Math
%% fixme setup(["\\p{Sm}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Sm}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Symbol, Currency
%% fixme setup(["\\p{Sc}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Sc}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Symbol, Modifier
%% fixme setup(["\\p{Sk}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Sk}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Symbol, Other
%% fixme setup(["\\p{So}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{So}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Separator, Any
%% fixme setup(["\\p{Z}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Z}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Separator, Space
%% fixme setup(["\\p{Zs}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Zs}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Separator, Line
%% fixme setup(["\\p{Zl}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Zl}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Separator, Paragraph
%% fixme setup(["\\p{Zp}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Zp}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Other, Any
%% fixme setup(["\\p{C}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{C}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Other, Control
%% fixme setup(["\\p{Cc}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Cc}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Other, Format
%% fixme setup(["\\p{Cf}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Cf}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Other, Surrogate not supported by schema recommendation
%% fixme setup(["\\p{Cs}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Cs}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Other, Private Use
%% fixme setup(["\\p{Co}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Co}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
%% Other, Not assigned (no characters in the file have this property)
%% fixme setup(["\\p{Cn}" ++ S) -> {{char_class,"\s\t\n\r"},S};
%% fixme setup(["\\P{Cn}" ++ S) -> {{comp_class,"\s\t\n\r"},S};
setup([A|S], Acc) -> setup(S, [A|Acc]);
setup([],Acc) ->  reverse([$$|Acc]).

%% sh_to_awk(ShellRegExp)
%%  Convert a sh style regexp into a full AWK one. The main difficulty is
%%  getting character sets right as the conventions are different.

sh_to_awk(Sh) -> "^(" ++ sh_to_awk_1(Sh).	%Fix the beginning

sh_to_awk_1([$*|Sh]) ->				%This matches any string
    ".*" ++ sh_to_awk_1(Sh);
sh_to_awk_1([$?|Sh]) ->				%This matches any character
    [$.|sh_to_awk_1(Sh)];
sh_to_awk_1([$[,$^,$]|Sh]) ->			%This takes careful handling
    "\\^" ++ sh_to_awk_1(Sh);
%% Must move '^' to end.
sh_to_awk_1("[^" ++ Sh) -> [$[|sh_to_awk_2(Sh, true)];
sh_to_awk_1("[!" ++ Sh) -> "[^" ++ sh_to_awk_2(Sh, false);
sh_to_awk_1([$[|Sh]) -> [$[|sh_to_awk_2(Sh, false)];
sh_to_awk_1([C|Sh]) ->
    %% Unspecialise everything else which is not an escape character.
    case sh_special_char(C) of
	true -> [$\\,C|sh_to_awk_1(Sh)];
	false -> [C|sh_to_awk_1(Sh)]
    end;
sh_to_awk_1([]) -> ")$".			%Fix the end

sh_to_awk_2([$]|Sh], UpArrow) -> [$]|sh_to_awk_3(Sh, UpArrow)];
sh_to_awk_2(Sh, UpArrow) -> sh_to_awk_3(Sh, UpArrow).

sh_to_awk_3([$]|Sh], true) -> "^]" ++ sh_to_awk_1(Sh);
sh_to_awk_3([$]|Sh], false) -> [$]|sh_to_awk_1(Sh)];
sh_to_awk_3([C|Sh], UpArrow) -> [C|sh_to_awk_3(Sh, UpArrow)];
sh_to_awk_3([], true) -> [$^|sh_to_awk_1([])];
sh_to_awk_3([], false) -> sh_to_awk_1([]).

%% -type sh_special_char(char()) -> bool().
%%  Test if a character is a special character.

sh_special_char($|) -> true;
sh_special_char($*) -> true;
sh_special_char($+) -> true;
sh_special_char($?) -> true;
sh_special_char($() -> true;
sh_special_char($)) -> true;
sh_special_char($\\) -> true;
sh_special_char($^) -> true;
sh_special_char($$) -> true;
sh_special_char($.) -> true;
sh_special_char($[) -> true;
sh_special_char($]) -> true;
sh_special_char($") -> true;
sh_special_char(_C) -> false.

%% parse(RegExp) -> {ok,RE} | {error,E}.
%%  Parse the regexp described in the string RegExp.

parse(S) ->
    case catch reg(S, 0) of
	{R,Sc,[]} -> {ok,{regexp,{R,Sc}}};
	{_R,_Sc,[C|_]} -> {error,{illegal,[C]}};
	{error,E} -> {error,E}
    end.

%% format_error(Error) -> String.

format_error({interval_range,What}) ->
    ["illegal interval range",io_lib:write_string(What)];
format_error({illegal,What}) -> ["illegal character `",What,"'"];
format_error({unterminated,What}) -> ["unterminated `",What,"'"];
format_error({posix_cc,What}) ->
    ["illegal POSIX character class ",io_lib:write_string(What)];
format_error({char_class,What}) ->
    ["illegal character class ",io_lib:write_string(What)].

%% match(String, RegExp) -> {match,Start,Length} | nomatch | {error,E}.
%%  Find the longest match of RegExp in String.

match(S, RegExp) when is_list(RegExp) ->
    case parse(RegExp) of
	{ok,RE} -> match(S, RE);
	{error,E} -> {error,E}
    end;
match(S, {regexp,RE}) ->
    case match_re(RE, S, 1, 0, -1) of
	{Start,Len} when Len >= 0 ->
	    {match,Start,Len};
	{_Start,_Len} -> nomatch
    end;
match(S, {comp_regexp,RE}) ->
    case match_comp(RE, S, 1, 0, -1) of
	{Start,Len} when Len >= 0 ->
	    {match,Start,Len};
	{_Start,_Len} -> nomatch
    end.

match_re(RE, [_|Cs]=S0, P0, Mst, Mlen) ->
    case re_apply(S0, P0, RE) of
	{match,P1,_S1,_Subs} ->
	    Len = P1-P0,
	    if Len > Mlen -> match_re(RE, Cs, P0+1, P0, Len);
	       true -> match_re(RE, Cs, P0+1, Mst, Mlen)
	    end;
	nomatch -> match_re(RE, Cs, P0+1, Mst, Mlen);
	never_match -> {Mst,Mlen}		%No need to go on
    end;
match_re(_RE, _S, _P, Mst, Mlen) -> {Mst,Mlen}.

match_comp(RE, [_|Cs]=S0, P0, Mst, Mlen) ->
    case comp_apply(S0, P0, RE) of
	{match,P1,_S1} ->
	    Len = P1-P0,
	    if Len > Mlen -> match_comp(RE, Cs, P0+1, P0, Len);
	       true -> match_comp(RE, Cs, P0+1, Mst, Mlen)
	    end;
	nomatch -> match_comp(RE, Cs, P0+1, Mst, Mlen)
    end;
match_comp(_RE, _S, _P, Mst, Mlen) -> {Mst,Mlen}.

%% match_re(RE, S0, Pos0, Mst, Mlen) ->
%%     case first_match_re(RE, S0, Pos0) of
%% 	{St,Len,_} ->				%Found a match
%% 	    Pos1 = St + 1,			%Where to start next match
%% 	    S1 = lists:nthtail(Pos1-Pos0, S0),
%% 	    if Len > Mlen -> match_re(RE, S1, Pos1, St, Len);
%% 	       true -> match_re(RE, S1, Pos1, Mst, Mlen)
%% 	    end;
%% 	nomatch -> {Mst,Mlen}
%%     end.

%% match_comp(RE, S0, Pos0, Mst, Mlen) ->
%%     case first_match_comp(RE, S0, Pos0) of
%% 	{St,Len} ->				%Found a match
%% 	    Pos1 = St + 1,			%Where to start next match
%% 	    S1 = lists:nthtail(Pos1-Pos0, S0),
%% 	    if Len > Mlen -> match_comp(RE, S1, Pos1, St, Len);
%% 	       true -> match_comp(RE, S1, Pos1, Mst, Mlen)
%% 	    end;
%% 	nomatch -> {Mst,Mlen}
%%     end.

%% first_match(String, RegExp) -> {match,Start,Length} | nomatch | {error,E}.
%%  Find the first match of RegExp in String.

first_match(S, RegExp) when is_list(RegExp) ->
    case parse(RegExp) of
	{ok,RE} -> first_match(S, RE);
	{error,E} -> {error,E}
    end;
first_match(S, {regexp,RE}) ->
    case first_match_re(RE, S, 1) of
	{Start,Len,_} -> {match,Start,Len};
	nomatch -> nomatch
    end;
first_match(S, {comp_regexp,RE}) ->
    case first_match_comp(RE, S, 1) of
	{Start,Len} -> {match,Start,Len};
	nomatch -> nomatch
    end.

first_match_re(RE, S, St) when S /= [] ->
    case re_apply(S, St, RE) of
	{match,P,_Rest,Subs} -> {St,P-St,Subs};
	nomatch -> first_match_re(RE, tl(S), St+1);
	never_match -> nomatch
    end;
first_match_re(_RE, [], _St) -> nomatch.

first_match_comp(RE, S, St) when S /= [] ->
    case comp_apply(S, St, RE) of
	{match,P,_Rest} -> {St,P-St};
	nomatch -> first_match_comp(RE, tl(S), St+1)
    end;
first_match_comp(_RE, [], _St) -> nomatch.

%% matches(String, RegExp) -> {match,[{Start,Length}]} | {error,E}.
%%  Return the all the non-overlapping matches of RegExp in String.

matches(S, RegExp) when is_list(RegExp) ->
    case parse(RegExp) of
	{ok,RE} -> matches(S, RE);
	{error,E} -> {error,E}
    end;
matches(S, {regexp,RE}) -> {match,matches_re(S, RE, 1)};
matches(S, {comp_regexp,RE}) -> {match,matches_comp(S, RE, 1)}.

matches_re([_|Cs]=S0, RE, P0) ->
    case re_apply(S0, P0, RE) of
	{match,P0,S1,_Subs} ->			%0 length match
	    [{P0,0}|matches_re(tl(S1), RE, P0+1)];
	{match,P1,S1,_Subs} ->
	    [{P0,P1-P0}|matches_re(S1, RE, P1)];
	nomatch -> matches_re(Cs, RE, P0+1);
	never_match -> []
    end;
matches_re([], _RE, _P) -> [].

matches_comp([_|Cs]=S0, RE, P0) ->
    case comp_apply(S0, P0, RE) of
	{match,P0,S1} ->			%0 length match
	    [{P0,0}|matches_comp(tl(S1), RE, P0+1)];
	{match,P1,S1} ->
	    [{P0,P1-P0}|matches_comp(S1, RE, P1)];
	nomatch -> matches_comp(Cs, RE, P0+1)
    end;
matches_comp([], _RE, _P) -> [].

%% sub(String, RegExp, Replace) -> {ok,RepString,RepCount} | {error,E}.
%%  Substitute the first match of the regular expression RegExp with
%%  the string Replace in String. Accept pre-parsed regular
%%  expressions.

sub(String, RegExp, Rep) when is_list(RegExp) ->
    case parse(RegExp) of
	{ok,RE} -> sub(String, RE, Rep);
	{error,E} -> {error,E}
    end;
sub(String, {regexp,RE}, Rep) ->
    case sub_re(String, 1, RE, [], Rep) of
	{yes,NewStr} -> {ok,NewStr,1};
	no -> {ok,String,0}
    end;
sub(String, {comp_regexp,RE}, Rep) ->
    case sub_comp(String, 1, RE, [], Rep) of
	{yes,NewStr} -> {ok,NewStr,1};
	no -> {ok,String,0}
    end.

%% sub_re(String, Position, Regexp, Before, Replacement) ->
%%      {NewString,Count}.
%% sub_comp(String, Position, Regexp, Before, Replacement) ->
%%      {NewString,Count}.
%% Step forward over String until a match is found saving stepped over
%% chars in Before. Return reversed Before prepended to replacement
%% and rest of string.

sub_re([C|Cs]=S0, P0, RE, Bef, Rep) ->
    case re_apply(S0, P0, RE) of
	{match,P0,_S1,_} ->			%Ignore 0 length match
	    sub_re(Cs, P0+1, RE, [C|Bef], Rep);
	{match,P1,Rest,_Gps} ->
	    {yes,reverse(Bef, sub_repl(Rep, substr(S0, 1, P1-P0), Rest))};
	nomatch -> sub_re(Cs, P0+1, RE, [C|Bef], Rep);
	never_match -> no			%No need to go on
    end;
sub_re([], _P, _RE, _Bef, _Rep) -> no.

sub_comp([C|Cs]=S0, P0, RE, Bef, Rep) ->
    case comp_apply(S0, P0, RE) of
	{match,P0,_S1} ->			%Ignore 0 length match
	    sub_comp(Cs, P0+1, RE, [C|Bef], Rep);
	{match,P1,Rest} ->
	    {yes,reverse(Bef, sub_repl(Rep, substr(S0, 1, P1-P0), Rest))};
	nomatch -> sub_comp(Cs, P0+1, RE, [C|Bef], Rep)
    end;
sub_comp([], _P, _RE, _Bef, _Rep) -> no.

sub_repl([$&|Rep], M, Rest) -> M ++ sub_repl(Rep, M, Rest);
sub_repl("\\&" ++ Rep, M, Rest) -> [$&|sub_repl(Rep, M, Rest)];
sub_repl([C|Rep], M, Rest) -> [C|sub_repl(Rep, M, Rest)];
sub_repl([], _M, Rest) -> Rest.

%%  gsub(String, RegExp, Replace) -> {ok,RepString,RepCount} | {error,E}.
%%  Substitute every match of the regular expression RegExp with the
%%  string New in String. Accept pre-parsed regular expressions.

gsub(String, RegExp, Rep) when is_list(RegExp) ->
    case parse(RegExp) of
	{ok,RE} -> gsub(String, RE, Rep);
	{error,E} -> {error,E}
    end;
gsub(String, {regexp,RE}, Rep) ->
    case gsub_re(String, 1, RE, [], Rep) of
	{NewStr,N} -> {ok,NewStr,N};
	no -> {ok,String,0}			%No substitutions
    end;
gsub(String, {comp_regexp,RE}, Rep) ->
    case gsub_comp(String, 1, RE, [], Rep) of
	{NewStr,N} -> {ok,NewStr,N};
	no -> {ok,String,0}			%No substitutions
    end.

%% gsub_re(String, Position, Regexp, Before, Replacement) ->
%%      {NewString,Count}.
%% gsub_comp(String, Position, Regexp, Before, Replacement) ->
%%      {NewString,Count}.
%% Step forward over String until a match is found saving stepped over
%% chars in Before. Call recursively to do rest of string after
%% match. Return reversed Before prepended to return from recursive
%% call.

gsub_re([C|Cs]=S0, P0, RE, Bef, Rep) ->
    case re_apply(S0, P0, RE) of
	{match,P0,_S1,_} ->			%Ignore 0 length match
	    gsub_re(Cs, P0+1, RE, [C|Bef], Rep);
	{match,P1,S1,_Gps} ->
	    case gsub_re(S1, P1, RE, [], Rep) of
		{NewStr,N0} ->			%Substituitions
		    {reverse(Bef, sub_repl(Rep, substr(S0, 1, P1-P0), NewStr)),
		     N0+1};
		no ->				%No substituitions.
		    {reverse(Bef, sub_repl(Rep, substr(S0, 1, P1-P0), S1)),1}
	    end;
	%%No match so step forward saving C on Bef.
	nomatch -> gsub_re(Cs, P0+1, RE, [C|Bef], Rep);
	never_match -> no			%No need to go on
    end;
gsub_re([], _P, _RE, _Bef, _Rep) -> no.

gsub_comp([C|Cs]=S0, P0, RE, Bef, Rep) ->
    case comp_apply(S0, P0, RE) of
	{match,P0,_S1} ->			%Ignore 0 length match
	    gsub_comp(Cs, P0+1, RE, [C|Bef], Rep);
	{match,P1,S1} ->
	    case gsub_comp(S1, P1, RE, [], Rep) of
		{NewStr,N0} ->			%Substituitions
		    {reverse(Bef, sub_repl(Rep, substr(S0, 1, P1-P0), NewStr)),
		     N0+1};
		no ->				%No substituitions.
		    {reverse(Bef, sub_repl(Rep, substr(S0, 1, P1-P0), S1)),1}
	    end;
	%%No match so step forward saving C on Bef.
	nomatch -> gsub_comp(Cs, P0+1, RE, [C|Bef], Rep)
    end;
gsub_comp([], _P, _RE, _Bef, _Rep) -> no.

%% split(String, RegExp) -> {ok,[SubString]} | {error,E}.
%%  Split a string into substrings where the RegExp describes the
%%  field seperator. The RegExp " " is specially treated.

split(String, " ") ->				%This is really special
    {ok,{regexp,RE}} = parse("[ \t]+"),
    case split_apply_re(String, RE, true) of
	[[]|Ss] -> {ok,Ss};
	Ss -> {ok,Ss}
    end;
split(String, RegExp) when is_list(RegExp) ->
    case parse(RegExp) of
	{ok,{regexp,RE}} -> {ok,split_apply_re(String, RE, false)};
	{error,E} -> {error,E}
    end;
split(String, {regexp,RE}) -> {ok,split_apply_re(String, RE, false)};
split(String, {comp_regexp,RE}) -> {ok,split_apply_comp(String, RE, false)}.

split_apply_re(S, RE, Trim) -> split_apply_re(S, 1, RE, Trim, []).

split_apply_re([], _P, _RE, true, []) -> [];
split_apply_re([], _P, _RE, _T, Sub) -> [reverse(Sub)];
split_apply_re([C|Cs]=S, P0, RE, T, Sub) ->
    case re_apply(S, P0, RE) of
	{match,P0,_S1,_} ->			%Ignore 0 length match
	    split_apply_re(Cs, P0+1, RE, T, [C|Sub]);
	{match,P1,S1,_} ->
	    [reverse(Sub)|split_apply_re(S1, P1, RE, T, [])];
	nomatch ->
	    split_apply_re(Cs, P0+1, RE, T, [C|Sub]);
	never_match -> [reverse(Sub, S)]	%No need to go on
    end.

split_apply_comp(S, RE, Trim) -> split_apply_comp(S, 1, RE, Trim, []).

%%split_apply_comp([], _P, _RE, true, []) -> [];
split_apply_comp([], _P, _RE, _T, Sub) -> [reverse(Sub)];
split_apply_comp([C|Cs]=S, P0, RE, T, Sub) ->
    case comp_apply(S, P0, RE) of
	{match,P0,_S1} ->			%Ignore 0 length match
	    split_apply_comp(Cs, P0+1, RE, T, [C|Sub]);
	{match,P1,S1} ->
	    [reverse(Sub)|split_apply_comp(S1, P1, RE, T, [])];
	nomatch ->
	    split_apply_comp(Cs, P0+1, RE, T, [C|Sub])
    end.

%% sub_match(String, RegExp) ->
%%      {match,Start,Length,SubExprs} | nomatch | {error,E}.
%%  Find the longest match of RegExp in String.

sub_match(S, RegExp) when is_list(RegExp) ->
    case parse(RegExp) of
	{ok,RE} -> sub_match(S, RE);
	{error,E} -> {error,E}
    end;
sub_match(S, {regexp,RE}) ->
    case sub_match_re(RE, S, 1, 0, -1, none) of
	{Start,Len,Subs} when Len >= 0 ->
	    {match,Start,Len,Subs};
	{_Start,_Len,_Subs} -> nomatch
    end.

sub_match_re(RE, S0, Pos0, Mst, Mlen, Msubs) ->
    case first_match_re(RE, S0, Pos0) of
	{St,Len,Subs} ->			%Found a match
	    Pos1 = St + 1,			%Where to start next match
	    S1 = lists:nthtail(Pos1-Pos0, S0),
	    if Len > Mlen -> sub_match_re(RE, S1, Pos1, St, Len, Subs);
	       true -> sub_match_re(RE, S1, Pos1, Mst, Mlen, Msubs)
	    end;
	nomatch -> {Mst,Mlen,Msubs}
    end.

%% sub_first_match(String, RegExp) ->
%%       {match,Start,Length,SubExprs} | nomatch | {error,E}.
%%  Find the longest match of RegExp in String, return Start and Length
%%  as well as tuple of sub-expression matches.

sub_first_match(S, RegExp) when is_list(RegExp) ->
    {ok,RE} = parse(RegExp),
    sub_first_match(S, RE);
sub_first_match(S, {regexp,RE}) ->
    case first_match_re(RE, S, 1) of
	{St,Len,Subs} -> {match,St,Len,Subs};
	nomatch -> nomatch
    end.


%% This is the regular expression grammar used. It is equivalent to the
%% one used in AWK, except that we allow ^ $ to be used anywhere and fail
%% in the matching.
%%
%% reg -> reg1 : '$1'.
%% reg1 -> reg1 "|" reg2 : {'or','$1','$2'}.
%% reg1 -> reg2 : '$1'.
%% reg2 -> reg2 reg3 : {concat,'$1','$2'}.
%% reg2 -> reg3 : '$1'.
%% reg3 -> reg3 "*" : {kclosure,'$1'}.
%% reg3 -> reg3 "+" : {pclosure,'$1'}.
%% reg3 -> reg3 "?" : {optional,'$1'}.
%% reg3 -> reg3 "{" [Min],[Max] "}" : {closure_range, Num, '$1'} see below
%% reg3 -> reg4 : '$1'.
%% reg4 -> "(" reg ")" : '$2'.
%% reg4 -> "\\" char : '$2'.
%% reg4 -> "^" : bos.
%% reg4 -> "$" : eos.
%% reg4 -> "." : char.
%% reg4 -> "[" class "]" : {char_class,char_class('$2')}
%% reg4 -> "[" "^" class "]" : {comp_class,char_class('$3')}
%% reg4 -> "\"" chars "\"" : char_string('$2')
%% reg4 -> char : '$1'.
%% reg4 -> empty : epsilon.
%%  The grammar of the current regular expressions. The actual parser
%%  is a recursive descent implementation of the grammar.

reg(S, Sc) -> reg1(S, Sc).

%% reg1 -> reg2 reg1'
%% reg1' -> "|" reg2
%% reg1' -> empty

reg1(S0, Sc0) ->
    {L,Sc1,S1} = reg2(S0, Sc0),
    reg1p(S1, L, Sc1).

reg1p([$||S0], L, Sc0) ->
    {R,Sc1,S1} = reg2(S0, Sc0),
    reg1p(S1, {'or',L,R}, Sc1);
reg1p(S, L, Sc) -> {L,Sc,S}.

%% reg2 -> reg3 reg2'
%% reg2' -> reg3
%% reg2' -> empty

reg2(S0, Sc0) ->
    {L,Sc1,S1} = reg3(S0, Sc0),
    reg2p(S1, L, Sc1).

reg2p([C|S0], L, Sc0) when C /= $|, C /= $) ->
    {R,Sc1,S1} = reg3([C|S0], Sc0),
    %% reg2p(S1, {concat,L,R}, Sc1);
    case is_integer(R) of
 	true -> 
 	    case L of
 		{literal,Lit} ->
 		    reg2p(S1, {literal,Lit ++[R]}, Sc1);
 		{concat,S2,Char} when is_integer(Char) ->
 		    reg2p(S1, {concat,S2,{literal,[Char,R]}}, Sc1);
 		{concat,S2,{literal,Lit}}  ->
 		    reg2p(S1, {concat,S2,{literal,Lit ++ [R]}}, Sc1);
 		Char when is_integer(Char) -> 
 		    reg2p(S1, {literal,[Char,R]}, Sc1);
 		_ ->
 		    reg2p(S1, {concat,L,R}, Sc1)
 	    end;
 	false ->
 	    reg2p(S1, {concat,L,R}, Sc1)
    end;
reg2p(S, L, Sc) -> {L,Sc,S}.

%% reg3 -> reg4 reg3'
%% reg3' -> "*" reg3'
%% reg3' -> "+" reg3'
%% reg3' -> "?" reg3'
%% reg3' -> "{" [Min],[Max] "}" reg3'
%% reg3' -> empty

reg3(S0, Sc0) ->
    {L,Sc1,S1} = reg4(S0, Sc0),
    reg3p(S1, L, Sc1).

reg3p([$*|S], L, Sc) -> reg3p(S, {kclosure,L}, Sc);
reg3p([$+|S], L, Sc) -> reg3p(S, {pclosure,L}, Sc);
reg3p([$?|S], L, Sc) -> reg3p(S, {optional,L}, Sc);
reg3p([${|Cs0], L, Sc) ->			% $}
    case interval_range(Cs0) of
	{none,none,_Cs1} -> parse_error({interval_range,[${|Cs0]});
	{N,M,[$}|Cs1]} -> reg3p(Cs1, {iclosure,L,N,M}, Sc);
	{_N,_M,_Cs1} -> parse_error({unterminated,"{"})
    end;
reg3p(S, L, Sc) -> {L,Sc,S}.

reg4([$(|S0], Sc0) ->
    Sc1 = Sc0+1,
    case reg(S0, Sc1) of
	{R,Sc2,[$)|S1]} -> {{subexpr,Sc1,R},Sc2,S1};
	{_R,_Sc,_S} -> parse_error({unterminated,"("})
    end;
reg4([$^|S], Sc) -> {bos,Sc,S};
reg4([$$|S], Sc) -> {eos,Sc,S};
reg4([$.|S], Sc) -> {{comp_class,"\n"},Sc,S};
reg4("[^" ++ S0, Sc) ->
    case char_class(S0) of
	{Cc,[$]|S1]} -> {{comp_class,Cc},Sc,S1};
	{_Cc,_S} -> parse_error({unterminated,"["})
    end;
reg4([$[|S0], Sc) ->
    case char_class(S0) of
	{Cc,[$]|S1]} -> {{char_class,Cc},Sc,S1};
	{_Cc,_S1} -> parse_error({unterminated,"["})
    end;
%reg4([$"|S0], Sc) ->
%    case char_string(S0) of
%	{St,[$"|S1]} -> {St,Sc,S1};
%	{St,S1} -> parse_error({unterminated,"\""})
%    end;
reg4([C0|S0], Sc) when
  is_integer(C0), C0 /= $*, C0 /= $+, C0 /= $?, C0 /= $], C0 /= $), C0 /= $} ->
    %% Handle \ quoted characters as well, at least those we see.
    {C1,S1} = char(C0, S0),
    {C1,Sc,S1};
reg4(S=[$)|_], Sc) -> {epsilon,Sc,S};
reg4([C|_S], _Sc) -> parse_error({illegal,[C]});
reg4([], Sc) -> {epsilon,Sc,[]}.

char($\\, [O1,O2,O3|S]) when
  O1 >= $0, O1 =< $7, O2 >= $0, O2 =< $7, O3 >= $0, O3 =< $7 ->
    {(O1*8 + O2)*8 + O3 - 73*$0,S};
char($\\, [C|S]) -> {escape_char(C),S};
char($\\, []) -> parse_error({unterminated,"\\"});
char(C, S) -> {C,S}.

escape_char($n) -> $\n;				%\n = LF
escape_char($r) -> $\r;				%\r = CR
escape_char($t) -> $\t;				%\t = TAB
escape_char($v) -> $\v;				%\v = VT
escape_char($b) -> $\b;				%\b = BS
escape_char($f) -> $\f;				%\f = FF
escape_char($e) -> $\e;				%\e = ESC
escape_char($s) -> $\s;				%\s = SPACE
escape_char($d) -> $\d;				%\d = DEL
escape_char(C) -> C.

char_class([$]|S0]) ->
    {Cc,S1} = char_class(S0, [$]]),
    {pack_cc(Cc),S1};
char_class(S0) ->
    {Cc,S1} = char_class(S0, []),
    {pack_cc(Cc),S1}.

pack_cc(Cc0) ->
    %% First sort the list.
    Cc1 = lists:usort(fun ({Cf1,_}, {Cf2,_}) -> Cf1 < Cf2;
			  ({Cf1,_}, C) -> Cf1 < C;
			  (C, {Cf,_}) -> C < Cf;
			  (C1, C2) -> C1 =< C2
		      end, Cc0),
    pack_cc1(Cc1).

pack_cc1([{Cf1,Cl1},{Cf2,Cl2}|Cc]) when Cl1 >= Cf2, Cl1 =< Cl2 ->
    pack_cc1([{Cf1,Cl2}|Cc]);
pack_cc1([{Cf1,Cl1},{Cf2,Cl2}|Cc]) when Cl1 >= Cf2, Cl1 >= Cl2 ->
    pack_cc1([{Cf1,Cl1}|Cc]);
pack_cc1([{Cf1,Cl1},{Cf2,Cl2}|Cc]) when Cl1+1 == Cf2 ->
    pack_cc1([{Cf1,Cl2}|Cc]);
pack_cc1([{Cf,Cl},C|Cc]) when Cl >= C -> pack_cc1([{Cf,Cl}|Cc]);
pack_cc1([{Cf,Cl},C|Cc]) when Cl+1 == C -> pack_cc1([{Cf,C}|Cc]);
pack_cc1([C,{Cf,Cl}|Cc]) when C == Cf-1 -> pack_cc1([{C,Cl}|Cc]);
pack_cc1([C1,C2|Cc]) when C1+1 == C2 -> pack_cc1([{C1,C2}|Cc]);
pack_cc1([C|Cc]) -> [C|pack_cc1(Cc)];
pack_cc1([]) -> [].

char_class("[:" ++ S0, Cc0) ->			%Start of POSIX char class
    case posix_cc(S0, Cc0) of
	{Cc1,":]" ++ S1} -> char_class(S1, Cc1);
	{_,_S1} -> parse_error({posix_cc,"[:" ++ S0})
    end;
char_class([C1|S0], Cc) when C1 /= $] ->
    case char(C1, S0) of
	{Cf,[$-,C2|S1]} when C2 /= $] ->
	    case char(C2, S1) of
		{Cl,S2} when Cf < Cl -> char_class(S2, [{Cf,Cl}|Cc]); 
		{_Cl,_S2} -> parse_error({char_class,[C1|S0]})
	    end;
	{C,S1} -> char_class(S1, [C|Cc])
    end;
char_class(S, Cc) -> {Cc,S}.

%% posix_cc(String, CharClass) -> {NewCharClass,RestString}.
%%  Handle POSIX character classes, use Latin-1 character set.

posix_cc("alnum" ++ S, Cc) ->
    {[{$0,$9},{$A,$Z},{192,214},{216,223},{$a,$z},{224,246},{248,255}|Cc],S};
posix_cc("alpha" ++ S, Cc) ->
    {[{$A,$Z},{192,214},{216,223},{$a,$z},{224,246},{248,255}|Cc],S};
posix_cc("blank" ++ S, Cc) -> {[$\s,$\t,160|Cc],S};
posix_cc("cntrl" ++ S, Cc) -> {[{0,31},{127,159}|Cc],S};
posix_cc("digit" ++ S, Cc) -> {[{$0,$9}|Cc],S};
posix_cc("graph" ++ S, Cc) -> {[{33,126},{161,255}|Cc],S};
posix_cc("lower" ++ S, Cc) -> {[{$a,$z},{224,246},{248,255}|Cc],S};
posix_cc("print" ++ S, Cc) -> {[{32,126},{160,255}|Cc],S};
posix_cc("punct" ++ S, Cc) -> {[{$!,$/},{$:,$?},{${,$~},{161,191}|Cc],S};
posix_cc("space" ++ S, Cc) -> {[$\s,$\t,$\f,$\r,$\v,160|Cc],S};
posix_cc("upper" ++ S, Cc) -> {[{$A,$Z},{192,214},{216,223}|Cc],S};
posix_cc("xdigit" ++ S, Cc) -> {[{$a,$f},{$A,$F},{$0,$9}|Cc],S};
posix_cc(S, _Cc) -> parse_error({posix_cc,"[:" ++ S}).

interval_range(Cs0) ->
    case number(Cs0) of
	{none,Cs1} -> {none,none,Cs1};
	{N,[$,|Cs1]} ->
	    case number(Cs1) of
		{none,Cs2} -> {N,any,Cs2};
		{M,Cs2} -> {N,M,Cs2}
	    end;
	{N,Cs1} -> {N,none,Cs1}
    end.

number([C|Cs]) when C >= $0, C =< $9 ->
    number(Cs, C - $0);
number(Cs) -> {none,Cs}.

number([C|Cs], Acc) when C >= $0, C =< $9 ->
    number(Cs, 10*Acc + (C - $0));
number(Cs, Acc) -> {Acc,Cs}.

parse_error(E) -> throw({error,E}).

%char_string([C|S]) when C /= $" -> char_string(S, C);
%char_string(S) -> {epsilon,S}.

%char_string([C|S0], L) when C /= $" ->
%    char_string(S0, {concat,L,C});
%char_string(S, L) -> {L,S}.

%% re_apply(String, StartPos, RegExp) ->
%%      {match,RestPos,Rest,SubExprs} | nomatch.
%%
%%  Apply the (parse of the) regular expression RegExp to String.  If
%%  there is a match return the position of the remaining string and
%%  the string if else return 'nomatch'.
%%
%%  StartPos should be the real start position as it is used to decide
%%  if we are at the beginning of the string.

re_apply(S, St, {RE,Sc}) ->
    Subs = erlang:make_tuple(Sc, none),		%Make a sub-regexp table.
    Res = re_apply(RE, [], S, St, Subs),
    %% ?dbg("~p x ~p -> ~p\n", [RE,S,Res]),
    Res.

re_apply(epsilon, More, S, P, Subs) ->		%This always matches
    re_apply_more(More, S, P, Subs);
re_apply({'or',RE1,RE2}, More, S, P, Subs) ->
    re_apply_or(re_apply(RE1, More, S, P, Subs),
		re_apply(RE2, More, S, P, Subs));
re_apply({concat,RE1,RE2}, More, S0, P, Subs) ->
    re_apply(RE1, [RE2|More], S0, P, Subs);
re_apply({literal,[C|Lcs]}, More, [C|S], P, Subs) ->
    re_apply_lit(Lcs, More, S, P+1, Subs);	%Have matched first char
re_apply({kclosure,RE}, More, S0, P0, Subs0) ->
    %% Greedy so try RE first, no difference here actually.
    Loop = case re_apply(RE, [], S0, P0, Subs0) of
	       {match,P0,_S1,_Subs1} ->		%0 length match, don't loop!
		   nomatch;
	       {match,P1,S1,Subs1} ->
		   re_apply_more([{kclosure,RE}|More], S1, P1, Subs1);
	       nomatch -> nomatch;
	       never_match -> never_match
	   end,
    re_apply_or(Loop, re_apply_more(More, S0, P0, Subs0));
re_apply({pclosure,RE}, More, S, P, Subs) ->
    re_apply(RE, [{kclosure,RE}|More], S, P, Subs);
re_apply({optional,RE}, More, S, P, Subs) ->
    %% Greedy so try RE first, no difference here actually.
    re_apply_or(re_apply(RE, More, S, P, Subs),
		re_apply_more(More, S, P, Subs));
re_apply({iclosure,RE,N,M}, More, S, P, Subs) when N > 0 ->
    re_apply(RE, [{iclosure,RE,N-1,M}|More], S, P, Subs);
re_apply({iclosure,RE,0,M}, More, S, P, Subs) ->
    Exp = expand_opt(RE, M),
    re_apply(Exp, More, S, P, Subs);
re_apply({subexpr,N,RE}, More, S, P, Subs) ->
    re_apply(RE, [{endsub,N,P}|More], S, P, Subs);
re_apply({endsub,N,St}, More, S, P, Subs0) ->
    Subs1 = setelement(N, Subs0, {St,P-St}),	%Record sub-expr
    re_apply_more(More, S, P, Subs1);
re_apply(bos, More, S, 1, Subs) -> re_apply_more(More, S, 1, Subs);
re_apply(bos, _More, _S, _, _) -> never_match;
re_apply(eos, More, [$\n], P, Subs) -> re_apply_more(More, [], P, Subs);
re_apply(eos, More, [], P, Subs) -> re_apply_more(More, [], P, Subs);
re_apply({char_class,Cc}, More, [C|S], P, Subs) ->
    case in_char_class(C, Cc) of
	true -> re_apply_more(More, S, P+1, Subs);
	false -> nomatch
    end;
re_apply({comp_class,Cc}, More, [C|S], P, Subs) ->
    case in_char_class(C, Cc) of
	true -> nomatch;
	false -> re_apply_more(More, S, P+1, Subs)
    end;
re_apply(C, More, [C|S], P, Subs) when is_integer(C) ->
    re_apply_more(More, S, P+1, Subs);
re_apply(_RE, _More, _S, _P, _Subs) ->
    %% ?dbg("~p : ~p\n", [_RE,_S]),
    nomatch.

%% re_apply_more([RegExp], String, Length, SubsExprs) ->
%%      {match,RestPos,Rest,SubExprs} | nomatch.

re_apply_more([RE|More], S, P, Subs) -> re_apply(RE, More, S, P, Subs);
re_apply_more([], S, P, Subs) -> {match,P,S,Subs}.

%% re_apply_lit(Literal, More, String, Position, SubExprs) ->
%%      {match,RestPos,Rest,SubExprs} | nomatch.
re_apply_lit([C|Lit], More, [C|Cs], P, Subs) ->
    re_apply_lit(Lit, More, Cs, P+1, Subs);
re_apply_lit([], More, Cs, P, Subs) ->
    re_apply_more(More, Cs, P, Subs);
re_apply_lit(_Lit, _More, _Cs, _P, _Subs) ->
    nomatch.

%% expand_iclosure(RE, N, M) -> RE.

expand_iclosure(RE, 0, M) -> expand_opt(RE, M);
expand_iclosure(RE, N, M) ->
    {concat,RE,expand_iclosure(RE, N-1, M)}.

%% expand_opt(RegExp, Count) -> RE.
%% Handle all the cases.

expand_opt(_RE, none) -> epsilon;
expand_opt(RE, any) -> {kclosure,RE};
expand_opt(_RE, 0) -> epsilon;
expand_opt(RE, 1) -> {optional,RE};
expand_opt(RE, N) ->
    {optional,{concat,RE,expand_opt(RE, N-1)}}.

%% find_prefix(PrefixStr, SourceStr)
%% if PrefixStr is a prefix of Str then return {ok,RemainingStr}
%% otherwise return false

%% find_prefix([C|Prest], [C|Rest]) ->
%%     find_prefix(Prest, Rest);
%% find_prefix([], Rest) -> {yes,Rest};
%% find_prefix(_, _) -> no.

%% in_char_class(Char, Class) -> bool().

in_char_class(C, [{C1,C2}|_Cc]) when C >= C1, C =< C2 -> true;
in_char_class(C, [C|_Cc]) -> true;
in_char_class(C, [_|Cc]) -> in_char_class(C, Cc);
in_char_class(_C, []) -> false.

%% re_apply_or(Match1, Match2, SubExprs) ->
%%      {match,RestPos,Rest,SubExprs} | nomatch.
%%  If we want the best match then choose the longest match, else just
%%  choose one by trying sequentially.

re_apply_or(M1={match,P1,_,_},{match,P2,_,_}) when P1 >= P2 -> M1;
re_apply_or({match,_,_,_},  M2={match,_,_,_}) -> M2;
re_apply_or(never_match, R2) -> R2;
re_apply_or(R1, never_match) -> R1;
re_apply_or(nomatch, R2) -> R2;
re_apply_or(R1, nomatch) -> R1.

%% Record definitions for the NFA, DFA and compiler.

-record(nfa_state, {no,edges=[],accept=no}).
-record(dfa_state, {no,nfa=[],trans=[],accept=no}).

-record(c_state, {no,trans=[],tmin=0,smin=none,tmax=0,smax=none,
		  accept=false,spec=[]}).

%% We use standard methods, Thompson's construction and subset
%% construction, to create first an NFA and then a DFA from the
%% regexps. A non-standard feature is that we work with sets of
%% character ranges (crs) instead sets of characters. This is most
%% noticeable when constructing DFAs. The major benefit is that we can
%% handle characters from any set, not just limited ASCII or 8859,
%% even 16/32 bit unicode.
%%
%% The whole range of characters is 0-maxchar, where maxchar is a BIG
%% number. We don't make any assumptions about the size of maxchar, it
%% is just bigger than any character.
%%
%% Using character ranges makes describing many regexps very simple,
%% for example the regexp "." just becomes the range
%% [{0-9},{11-maxchar}].

%% make_nfa(RegExpActions) -> {ok,{NFA,StartState}} | {error,E}.
%% Build a complete nfa from a list of {RegExp,Action}. The NFA field
%% accept has values {yes,Action}|no. The NFA is a list of states.

make_nfa(REAs0) ->
    case parse_reas(REAs0) of
	{ok,REAs1} ->
	    {NFA,Start} = build_combined_nfa(REAs1),
	    {ok,{NFA,Start}};
	{error,E} -> {error,E}
    end.

%% make_dfa(RegExpActions) -> {ok,{DFA,StartState}} | {error,E}.
%% make_dfa(RegExpActions, LowestState) -> {ok,{DFA,StartState}} | {error,E}.
%% Build a complete dfa from a list of {RegExp,Action}. The DFA field
%% accept has values {yes,Action}|no. If multiple Regexps can result
%% in same match string then RegExpActions list define priority.

make_dfa(REAs) -> make_dfa(REAs, 0).

make_dfa(REAs0, Low) ->
    case parse_reas(REAs0) of
	{ok,REAs1} ->
	    {NFA,Start0} = build_combined_nfa(REAs1),
	    {DFA0,Start1} = build_dfa(NFA, Start0),
	    {DFA,Start} = minimise_dfa(DFA0, Start1, Low),
	    {ok,{DFA,Start}};
	{error,E} -> {error,E}
    end.

parse_reas(REAs) -> parse_reas(REAs, []).

parse_reas([{{regexp,{R,_Sc}},A}|REAs], S) ->	%Already parsed
    parse_reas(REAs, [{R,A}|S]);
parse_reas([{RegExp,A}|REAs], S) ->
    case parse(RegExp) of
	{ok,{regexp,{R,_Sc}}} -> parse_reas(REAs, [{R,A}|S]);
	{error,E} -> {error,E}
    end;
parse_reas([], Stack) -> {ok,reverse(Stack)}.

%% build_combined_nfa(RegExpActionList) -> {NFA,StartState}.
%%  Build the combined NFA using Thompson's construction straight out
%%  of the book. Build the separate NFAs in the same order as the
%%  rules so that the accepting have ascending states have ascending
%%  state numbers.  Start numbering the states from 1 as we put the
%%  states in a tuple with the state number as the index.

build_combined_nfa(REAs) ->
    {NFA,Starts,Next} = build_nfa_list(REAs, [], [], 1),
    F = #nfa_state{no=Next,edges=epsilon_trans(Starts),accept=no},
    {[F|NFA],Next}.

build_nfa_list([{RE,Action}|REAs], NFA0, Starts, Next0) ->
    {NFA1,Next1,Start} = build_nfa(RE, Next0, Action),
    build_nfa_list(REAs, NFA1 ++ NFA0, [Start|Starts], Next1);
build_nfa_list([], NFA, Starts, Next) ->
    {NFA,reverse(Starts),Next}.

epsilon_trans(Sts) -> [ {epsilon,S} || S <- Sts ].

%% build_nfa(RegExp, NextState, Action) -> {NFA,NextFreeState,StartState}.
%%  When building the NFA states for a ??? we don't build the end
%%  state, just allocate a State for it and return this state
%%  number. This allows us to avoid building unnecessary states for
%%  concatenation which would then have to be removed by overwriting
%%  an existing state.

build_nfa(RE, Next, Action) ->
    {NFA,N,E} = build_nfa(RE, Next+1, Next, []),
    {[#nfa_state{no=E,accept={yes,Action}}|NFA],N,Next}.

%% build_nfa(RegExp, NextState, StartState, NFA) -> {NFA,NextState,EndState}.
%%  The NFA is a list of nfa_state is no predefined order. The state
%%  number of the returned EndState is already allocated!

build_nfa({'or',RE1,RE2}, N0, S, NFA0) ->
    {NFA1,N1,E1} = build_nfa(RE1, N0+1, N0, NFA0),
    {NFA2,N2,E2} = build_nfa(RE2, N1+1, N1, NFA1),
    E = N2,
    {[#nfa_state{no=S,edges=[{epsilon,N0},{epsilon,N1}]},
      #nfa_state{no=E1,edges=[{epsilon,E}]},
      #nfa_state{no=E2,edges=[{epsilon,E}]}|NFA2],
     N2+1,E};
build_nfa({literal,[]}, N, S, NFA) ->
    {NFA,N,S};
build_nfa({literal,[C|Cs]}, N0, S, NFA0) ->
    {NFA1,N1,E1} = build_nfa(C, N0, S, NFA0),
    build_nfa({literal,Cs}, N1, E1, NFA1);
build_nfa({concat,RE1,RE2}, N0, S, NFA0) ->
    {NFA1,N1,E1} = build_nfa(RE1, N0, S, NFA0),
    {NFA2,N2,E2} = build_nfa(RE2, N1, E1, NFA1),
    {NFA2,N2,E2};
build_nfa({kclosure,RE}, N0, S, NFA0) ->
    {NFA1,N1,E1} = build_nfa(RE, N0+1, N0, NFA0),
    E = N1,
    {[#nfa_state{no=S,edges=[{epsilon,N0},{epsilon,E}]},
      #nfa_state{no=E1,edges=[{epsilon,N0},{epsilon,E}]}|NFA1],
     N1+1,E};
build_nfa({pclosure,RE}, N0, S, NFA0) ->
    {NFA1,N1,E1} = build_nfa(RE, N0+1, N0, NFA0),
    E = N1,
    {[#nfa_state{no=S,edges=[{epsilon,N0}]},
      #nfa_state{no=E1,edges=[{epsilon,N0},{epsilon,E}]}|NFA1],
     N1+1,E};
build_nfa({optional,RE}, N0, S, NFA0) ->
    {NFA1,N1,E1} = build_nfa(RE, N0+1, N0, NFA0),
    E = N1,
    {[#nfa_state{no=S,edges=[{epsilon,N0},{epsilon,E}]},
      #nfa_state{no=E1,edges=[{epsilon,E}]}|NFA1],
     N1+1,E};
build_nfa({iclosure,RE,I1,I2}, N, S, NFA) ->
    Exp = expand_iclosure(RE, I1, I2),
    build_nfa(Exp, N, S, NFA);
build_nfa({char_class,Cc}, N, S, NFA) ->
    {[#nfa_state{no=S,edges=[{nfa_char_class(Cc),N}]}|NFA],N+1,N};
build_nfa({comp_class,Cc}, N, S, NFA) ->
    {[#nfa_state{no=S,edges=[{nfa_comp_class(Cc),N}]}|NFA],N+1,N};
build_nfa(epsilon, N, S, NFA) ->
    {NFA,N,S};
build_nfa({group,RE}, N, S, NFA) ->		%%% FIXME %%%%%%%
    build_nfa(RE, N, S, NFA);
build_nfa({subexpr,_N,RE}, N, S, NFA) ->	%%% FIXME %%%%%%%
    build_nfa(RE, N, S, NFA);
build_nfa(bos, N, S, NFA) ->
    {[#nfa_state{no=S,edges=[{[bos],N}]}|NFA],N+1,N};
build_nfa(eos, N, S, NFA) ->
    {[#nfa_state{no=S,edges=[{[eos],N}]}|NFA],N+1,N};
%%{[#nfa_state{no=S,edges=[{[eos],N}]}|NFA],N+1,N};
build_nfa(C, N, S, NFA) when is_integer(C) ->
    {[#nfa_state{no=S,edges=[{[{C,C}],N}]}|NFA],N+1,N}.

nfa_char_class(Cc) ->
    Crs = lists:foldl(fun({C1,C2}, Set) -> add_element({C1,C2}, Set);
			 (C, Set) -> add_element({C,C}, Set) end, [], Cc),
    %% ?dbg("cc: ~p\n", [Crs]),
    pack_crs(Crs).

pack_crs([{C1,C2}=Cr,{C3,C4}|Crs]) when C1 =< C3, C2 >= C4 ->
    %% C1      C2
    %%   C3  C4
    pack_crs([Cr|Crs]);
pack_crs([{C1,C2},{C3,C4}|Crs]) when C2 >= C3, C2 < C4 ->
    %% C1    C2
    %%    C3   C4
    pack_crs([{C1,C4}|Crs]);
pack_crs([{C1,C2},{C3,C4}|Crs]) when C2 + 1 == C3 ->
    %% C1   C2
    %%        C3  C4
    pack_crs([{C1,C4}|Crs]);
pack_crs([Cr|Crs]) -> [Cr|pack_crs(Crs)];
pack_crs([]) -> [].

nfa_comp_class(Cc) ->
    Crs = nfa_char_class(Cc),
    %% ?dbg("comp: ~p\n", [Crs]),
    comp_crs(Crs, 0).

comp_crs([{C1,C2}|Crs], Last) ->
    [{Last,C1-1}|comp_crs(Crs, C2+1)];
comp_crs([], Last) -> [{Last,maxchar}].

%% build_dfa(NFA, NfaStartState) -> {DFA,DfaStartState}.
%%  Build a DFA from an NFA using "subset construction". The major
%%  difference from the book is that we keep the marked and unmarked
%%  DFA states in separate lists. New DFA states are added to the
%%  unmarked list and states are marked by moving them to the marked
%%  list. We assume that the NFA accepting state numbers are in
%%  ascending order for the rules and use ordsets to keep this order.

build_dfa(NFA0, Start) ->
    %% We want NFA as sorted tuple for fast access, assume lowest state 1.
    NFA1 = list_to_tuple(keysort(#nfa_state.no, NFA0)),
    D = #dfa_state{no=0,nfa=eclosure([Start], NFA1),accept=no},
    {build_dfa([D], 1, [], NFA1),0}.

%% build_dfa([UnMarked], NextState, [Marked], NFA) -> DFA.
%%  Traverse the unmarked states. Temporarily add the current unmarked
%%  state to the marked list before calculating translation, this is
%%  to avoid adding too many duplicate states. Add it properly to the
%%  marked list afterwards with correct translations.

build_dfa([U|Us0], N0, Ms, NFA) ->
    {Ts,Us1,N1} = build_dfa(U#dfa_state.nfa, Us0, N0, [], [U|Ms], NFA),
    M = U#dfa_state{trans=Ts,accept=accept(U#dfa_state.nfa, NFA)},
    build_dfa(Us1, N1, [M|Ms], NFA);
build_dfa([], _N, Ms, _NFA) -> Ms.

%% build_dfa([NfaState], [Unmarked], NextState, [Transition], [Marked], NFA) ->
%%	{Transitions,UnmarkedStates,NextState}.
%%  Foreach NFA state set calculate the legal translations. N.B. must
%%  search *BOTH* the unmarked and marked lists to check if DFA state
%%  already exists. As the range of characters is potentially VERY
%%  large we cannot explicitly test all characters. Instead we first
%%  calculate the set of all disjoint character ranges which are
%%  possible candidates to the set of NFA states.

build_dfa(Set, Us, N, Ts, Ms, NFA) ->
    %% List of all transition sets.
    Crs0 = [Cr || S <- Set,
		  {Crs,_St} <- (element(S, NFA))#nfa_state.edges,
		 is_list(Crs),
		  Cr <- Crs ],
    Crs1 = lists:usort(Crs0),			%Must remove duplicates!
    %% Build list of disjoint test ranges.
    Test = disjoint_crs(Crs1),
    %% ?dbg("bd: ~p\n    ~p\n    ~p\n    ~p\n", [Set,Crs0,Crs1,Test]),
    build_dfa(Test, Set, Us, N, Ts, Ms, NFA).

%% disjoint_crs([CharRange]) -> [CharRange].
%%  Take a sorted list of char ranges and make a sorted list of
%%  disjoint char ranges. No new char range extends past an existing
%%  char range.

disjoint_crs([{_C1,C2}=Cr1,{C3,_C4}=Cr2|Crs]) when C2 < C3 ->
    %% C1  C2
    %%        C3  C4
    [Cr1|disjoint_crs([Cr2|Crs])];
disjoint_crs([{C1,C2},{C3,C4}|Crs]) when C1 == C3 ->
    %% C1     C2
    %% C3       C4
    [{C1,C2}|disjoint_crs(add_element({C2+1,C4}, Crs))];
disjoint_crs([{C1,C2},{C3,C4}|Crs]) when C1 < C3, C2 >= C3, C2 < C4 ->
    %% C1     C2
    %%    C3     C4
    [{C1,C3-1}|disjoint_crs(union([{C3,C2},{C2+1,C4}], Crs))];
disjoint_crs([{C1,C2},{C3,C4}|Crs]) when C1 < C3, C2 == C4 ->
    %% C1      C2
    %%    C3   C4
    [{C1,C3-1}|disjoint_crs(add_element({C3,C4}, Crs))];
disjoint_crs([{C1,C2},{C3,C4}|Crs]) when C1 < C3, C2 > C4 ->
    %% C1        C2
    %%    C3   C4
    [{C1,C3-1}|disjoint_crs(union([{C3,C4},{C4+1,C2}], Crs))];
disjoint_crs([Cr|Crs]) -> [Cr|disjoint_crs(Crs)];
disjoint_crs([]) -> [].

build_dfa([Cr|Crs], Set, Us, N, Ts, Ms, NFA) ->
    case eclosure(move(Set, Cr, NFA), NFA) of
	S when S /= [] ->
	    case keysearch(S, #dfa_state.nfa, Us) of
		{value,#dfa_state{no=T}} ->
		    build_dfa(Crs, Set, Us, N, [{Cr,T}|Ts], Ms, NFA);
		false ->
		    case keysearch(S, #dfa_state.nfa, Ms) of
			{value,#dfa_state{no=T}} ->
			    build_dfa(Crs, Set, Us, N, [{Cr,T}|Ts], Ms, NFA);
			false ->
			    U = #dfa_state{no=N,nfa=S},
			    build_dfa(Crs, Set, [U|Us], N+1, [{Cr,N}|Ts], Ms, NFA)
		    end
	    end;
	[] ->
	    build_dfa(Crs, Set, Us, N, Ts, Ms, NFA)
    end;
build_dfa([], _Set, Us, N, Ts, _Ms, _NFA) ->
    {Ts,Us,N}.
   
%% eclosure([State], NFA) -> [State].
%% move([State], Char, NFA) -> [State].
%%  These are straight out of the book. As eclosure uses ordsets then
%%  the generated state sets are in ascending order.

eclosure(Sts, NFA) -> eclosure(Sts, NFA, []).

eclosure([St|Sts], NFA, Ec) ->
    #nfa_state{edges=Es} = element(St, NFA),
    eclosure([ N || {epsilon,N} <- Es,
		    not is_element(N, Ec) ] ++ Sts,
	     NFA, add_element(St, Ec));
eclosure([], _NFA, Ec) -> Ec.

move(Sts, Cr, NFA) ->
    [ St || N <- Sts,
	    {Crs,St} <- (element(N, NFA))#nfa_state.edges,
	   is_list(Crs),
%% 	    begin
%% 		?dbg("move1: ~p\n", [{Sts,Cr,Crs,in_crs(Cr,Crs)}]),
%% 		true
%% 	    end,
	    in_crs(Cr, Crs) ].

in_crs({C1,C2}, [{C3,C4}|_Crs]) when C1 >= C3, C2 =< C4 -> true;
in_crs(Cr, [Cr|_Crs]) -> true;			%Catch bos and eos.
in_crs(Cr, [_|Crs]) -> in_crs(Cr, Crs);
in_crs(_Cr, []) -> false.

%% accept([State], NFA) -> true | false.
%%  Scan down the state list until we find an accepting state.

accept([St|Sts], NFA) ->
    case element(St, NFA) of
	#nfa_state{accept={yes,A}} -> {yes,A};
	#nfa_state{accept=no} -> accept(Sts, NFA)
    end;
accept([], _NFA) -> no.

%% minimise_dfa(DFA, StartState, FirstState) -> {DFA,StartState}.
%%  Minimise the DFA by removing equivalent states. We consider a
%%  state if both the transitions and the their accept state is the
%%  same.  First repeatedly run throught the DFA state list removing
%%  equivalent states and updating remaining transitions with
%%  remaining equivalent state numbers. When no more reductions are
%%  possible then pack the remaining state numbers to get consecutive
%%  states.

minimise_dfa(DFA0, Start, N) ->
    case min_dfa(DFA0) of
	{DFA1,[]} ->				%No reduction!
	    {DFA2,Rs} = pack_dfa(DFA1, N),
	    {min_update(DFA2, Rs),min_new_state(Start, Rs)};
	{DFA1,Rs} ->
	    minimise_dfa(min_update(DFA1, Rs), min_new_state(Start, Rs), N)
    end.

min_dfa(DFA) -> min_dfa(DFA, [], []).

min_dfa([D|DFA0], Rs0, MDFA) ->
    {DFA1,Rs1} = min_delete(DFA0, D#dfa_state.trans, D#dfa_state.accept, 
			    D#dfa_state.no, Rs0, []),
    min_dfa(DFA1, Rs1, [D|MDFA]);
min_dfa([], Rs, MDFA) -> {MDFA,Rs}.

min_delete([#dfa_state{no=N,trans=T,accept=A}|DFA], T, A, NewN, Rs, MDFA) ->
    min_delete(DFA, T, A, NewN, [{N,NewN}|Rs], MDFA);
min_delete([D|DFA], T, A, NewN, Rs, MDFA) ->
    min_delete(DFA, T, A, NewN, Rs, [D|MDFA]);
min_delete([], _T, _A, _NewN, Rs, MDFA) -> {MDFA,Rs}.

min_update(DFA, Rs) ->
    [ D#dfa_state{trans=min_update_trans(D#dfa_state.trans, Rs)} || D <- DFA ].

min_update_trans(Tr, Rs) ->
    [ {C,min_new_state(S, Rs)} || {C,S} <- Tr ].

min_new_state(Old, [{Old,New}|_Reds]) -> New;
min_new_state(Old, [_R|Reds]) -> min_new_state(Old, Reds);
min_new_state(Old, []) -> Old.

pack_dfa(DFA, N) -> pack_dfa(DFA, N, [], []).

pack_dfa([D|DFA], NewN, Rs, PDFA) ->
    pack_dfa(DFA, NewN+1, [{D#dfa_state.no,NewN}|Rs],
	     [D#dfa_state{no=NewN}|PDFA]);
pack_dfa([], _NewN, Rs, PDFA) -> {PDFA,Rs}.

%% comp_apply(String, StartPos, DFAReg) -> {match,RestPos,Rest} | nomatch.
%% Apply the DFA of a regular expression to a string.  If
%%  there is a match return the position of the remaining string and
%%  the string if else return 'nomatch'.
%%
%%  StartPos should be the real start position as it is used to decide
%%  if we are at the beginning of the string.

comp_apply(Cs, P, {DFA,Start,_Fail}) ->
    comp_apply(element(Start, DFA), Cs, P, DFA, nomatch).

comp_apply(#c_state{spec=[]}=St, Cs, P, DFA, Accept) ->
    comp_apply_tr(St, Cs, P, DFA, Accept);
comp_apply(#c_state{spec=Sp}=St, Cs, P, DFA, Accept) ->
    comp_apply_sp(St, Cs, P, DFA, Accept, Sp).

comp_apply_tr(#c_state{trans=none,accept=A}, Cs, P, _DFA, Accept) ->
    %% End state.
    accept_value(A, Cs, P, Accept);
comp_apply_tr(#c_state{trans=Tr,tmin=Tmin,smin=Smin,tmax=Tmax,smax=Smax,accept=A},
	      [C|Cs]=Cs0, P, DFA, Accept) ->
    %% Get the next state number to go to.
    NextSt = if  C =< Tmin -> Smin;		%Below transition table
		 C >= Tmax -> Smax;		%Above transition table
		 true ->			%Otherwise use table
 		     element(C - Tmin, Tr)
	     end,
    comp_apply(element(NextSt, DFA), Cs, P+1, DFA,
	       accept_value(A, Cs0, P, Accept));
comp_apply_tr(#c_state{trans=_Tr,accept=A}, [], P, _DFA, Accept) ->
    accept_value(A, [], P, Accept).

comp_apply_sp(_St, Cs, 1, DFA, Accept, [{bos,S}|_]) ->
    comp_apply(element(S, DFA), Cs, 1, DFA, Accept);
comp_apply_sp(_St, [$\n], P, DFA, Accept, [{eos,S}|_]) ->
    comp_apply(element(S, DFA), [], P, DFA, Accept);
comp_apply_sp(_St, [], P, DFA, Accept, [{eos,S}|_]) ->
    comp_apply(element(S, DFA), [], P, DFA, Accept);
comp_apply_sp(St, Cs, P, DFA, Accept, [_|Sp]) ->
    comp_apply_sp(St, Cs, P, DFA, Accept, Sp);
comp_apply_sp(St, Cs, P, DFA, Accept, []) ->
    comp_apply_tr(St, Cs, P, DFA, Accept).
    
accept_value(true, Cs, P, _Accept) -> {match,P,Cs};
accept_value(false, _Cs, _P, Accept) -> Accept.

%% compile(RegExp) -> {ok,RE} | {error,E}.
%%  Parse the regexp described in the string RegExp.

compile(RegExp) ->
    case make_dfa([{RegExp,yes}], 2) of
	{ok,{DFA0,Start}} ->
	    Fail = 1,
	    DFA1 = [#dfa_state{no=Fail,accept=no,trans=[]}|DFA0],
	    DFA = tuplelise_dfa(DFA1, 1),
	    {ok,{comp_regexp,{DFA,Start,Fail}}};
	{error,E} -> {error,E}
    end.

%% tuplelise_dfa(DFAstates, NoAcceptState) -> {{CompState},FirstState}.

tuplelise_dfa(DFA0, NoAccept) ->
    DFA1 = map(fun (#dfa_state{no=N,trans=Ts,accept=A}) ->
		       {Tr,Tmin,Smin,Tmax,Smax,Sp} = build_trans(Ts, NoAccept),
		       #c_state{no=N,trans=Tr,tmin=Tmin,smin=Smin,
				tmax=Tmax,smax=Smax,
				accept=fix_accept(A),spec=Sp}
	       end, DFA0),
    list_to_tuple(keysort(#dfa_state.no, DFA1)).

build_trans(Ts0, NoAccept) ->
    %% Split transitions into character ranges and specials.
    {Ts1,Sp1} = foldl(fun ({{_,_},_}=T, {Ts,Sp}) -> {[T|Ts],Sp};
			  ({_,_}=T, {Ts,Sp}) -> {Ts,[T|Sp]}
		      end, {[],[]}, Ts0),
    if Ts1 == [] ->
	    {none,none,none,none,none,Sp1};
       true ->
	    %% Have transitions, convert to tuple.
	    Ts2 = keysort(1, Ts1),
	    {Tmin,Smin,Ts3} = min_trans(Ts2, NoAccept),
	    %% ?dbg("exptr: ~p\n", [{Ts3,Tmin}]),
	    {Trans,Tmax,Smax} = expand_trans(Ts3, Tmin, NoAccept),
	    {list_to_tuple(Trans),Tmin,Smin,Tmax,Smax,Sp1}
    end.
   
min_trans([{{0,C2},S}|Crs], _Def) -> {C2,S,Crs};
min_trans([{{C1,_C2},_S}|_]=Crs, Def) -> {C1-1,Def,Crs}.

expand_trans([{{C1,maxchar},S}], Last, Def) ->
    Trs = duplicate(C1-(Last+1), Def),
    {Trs,C1,S};
expand_trans([{{C1,C2},S}], Last, Def) ->
    Trs = duplicate(C1-(Last+1), Def) ++ duplicate(C2-C1+1, S),
    {Trs,C2+1,Def};
expand_trans([{{C1,C2},S}|Crs], Last, Def) ->
    {Trs0,Tmax,Smax} = expand_trans(Crs, C2, Def),
    Trs1 = duplicate(C1-(Last+1), Def) ++ duplicate(C2-C1+1, S) ++ Trs0,
    {Trs1,Tmax,Smax}.

fix_accept({yes,_}) -> true;
fix_accept(no) -> false.