]>
Commit | Line | Data |
---|---|---|
38cbfe40 RK |
1 | ------------------------------------------------------------------------------ |
2 | -- -- | |
3 | -- GNAT LIBRARY COMPONENTS -- | |
4 | -- -- | |
5 | -- G N A T . R E G P A T -- | |
6 | -- -- | |
7 | -- B o d y -- | |
8 | -- -- | |
598c3446 | 9 | -- $Revision$ |
38cbfe40 RK |
10 | -- -- |
11 | -- Copyright (C) 1986 by University of Toronto. -- | |
12 | -- Copyright (C) 1996-2001 Ada Core Technologies, Inc. -- | |
13 | -- -- | |
14 | -- GNAT is free software; you can redistribute it and/or modify it under -- | |
15 | -- terms of the GNU General Public License as published by the Free Soft- -- | |
16 | -- ware Foundation; either version 2, or (at your option) any later ver- -- | |
17 | -- sion. GNAT is distributed in the hope that it will be useful, but WITH- -- | |
18 | -- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY -- | |
19 | -- or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License -- | |
20 | -- for more details. You should have received a copy of the GNU General -- | |
21 | -- Public License distributed with GNAT; see file COPYING. If not, write -- | |
22 | -- to the Free Software Foundation, 59 Temple Place - Suite 330, Boston, -- | |
23 | -- MA 02111-1307, USA. -- | |
24 | -- -- | |
25 | -- As a special exception, if other files instantiate generics from this -- | |
26 | -- unit, or you link this unit with other files to produce an executable, -- | |
27 | -- this unit does not by itself cause the resulting executable to be -- | |
28 | -- covered by the GNU General Public License. This exception does not -- | |
29 | -- however invalidate any other reasons why the executable file might be -- | |
30 | -- covered by the GNU Public License. -- | |
31 | -- -- | |
32 | -- GNAT is maintained by Ada Core Technologies Inc (http://www.gnat.com). -- | |
33 | -- -- | |
34 | ------------------------------------------------------------------------------ | |
35 | ||
36 | -- This is an altered Ada 95 version of the original V8 style regular | |
37 | -- expression library written in C by Henry Spencer. Apart from the | |
38 | -- translation to Ada, the interface has been considerably changed to | |
39 | -- use the Ada String type instead of C-style nul-terminated strings. | |
40 | ||
41 | -- Beware that some of this code is subtly aware of the way operator | |
42 | -- precedence is structured in regular expressions. Serious changes in | |
43 | -- regular-expression syntax might require a total rethink. | |
44 | ||
45 | with System.IO; use System.IO; | |
46 | with Ada.Characters.Handling; use Ada.Characters.Handling; | |
47 | with Unchecked_Conversion; | |
48 | ||
49 | package body GNAT.Regpat is | |
50 | ||
51 | MAGIC : constant Character := Character'Val (10#0234#); | |
52 | -- The first byte of the regexp internal "program" is actually | |
53 | -- this magic number; the start node begins in the second byte. | |
54 | -- | |
55 | -- This is used to make sure that a regular expression was correctly | |
56 | -- compiled. | |
57 | ||
58 | ---------------------------- | |
59 | -- Implementation details -- | |
60 | ---------------------------- | |
61 | ||
62 | -- This is essentially a linear encoding of a nondeterministic | |
63 | -- finite-state machine, also known as syntax charts or | |
64 | -- "railroad normal form" in parsing technology. | |
65 | ||
66 | -- Each node is an opcode plus a "next" pointer, possibly plus an | |
67 | -- operand. "Next" pointers of all nodes except BRANCH implement | |
68 | -- concatenation; a "next" pointer with a BRANCH on both ends of it | |
69 | -- is connecting two alternatives. | |
70 | ||
71 | -- The operand of some types of node is a literal string; for others, | |
72 | -- it is a node leading into a sub-FSM. In particular, the operand of | |
73 | -- a BRANCH node is the first node of the branch. | |
74 | -- (NB this is *not* a tree structure: the tail of the branch connects | |
75 | -- to the thing following the set of BRANCHes). | |
76 | ||
77 | -- You can see the exact byte-compiled version by using the Dump | |
78 | -- subprogram. However, here are a few examples: | |
79 | ||
80 | -- (a|b): 1 : MAGIC | |
81 | -- 2 : BRANCH (next at 10) | |
82 | -- 5 : EXACT (next at 18) operand=a | |
83 | -- 10 : BRANCH (next at 18) | |
84 | -- 13 : EXACT (next at 18) operand=b | |
85 | -- 18 : EOP (next at 0) | |
86 | -- | |
87 | -- (ab)*: 1 : MAGIC | |
88 | -- 2 : CURLYX (next at 26) { 0, 32767} | |
89 | -- 9 : OPEN 1 (next at 13) | |
90 | -- 13 : EXACT (next at 19) operand=ab | |
91 | -- 19 : CLOSE 1 (next at 23) | |
92 | -- 23 : WHILEM (next at 0) | |
93 | -- 26 : NOTHING (next at 29) | |
94 | -- 29 : EOP (next at 0) | |
95 | ||
96 | -- The opcodes are: | |
97 | ||
98 | type Opcode is | |
99 | ||
100 | -- Name Operand? Meaning | |
101 | ||
102 | (EOP, -- no End of program | |
103 | MINMOD, -- no Next operator is not greedy | |
104 | ||
105 | -- Classes of characters | |
106 | ||
107 | ANY, -- no Match any one character except newline | |
108 | SANY, -- no Match any character, including new line | |
109 | ANYOF, -- class Match any character in this class | |
110 | EXACT, -- str Match this string exactly | |
111 | EXACTF, -- str Match this string (case-folding is one) | |
112 | NOTHING, -- no Match empty string | |
113 | SPACE, -- no Match any whitespace character | |
114 | NSPACE, -- no Match any non-whitespace character | |
115 | DIGIT, -- no Match any numeric character | |
116 | NDIGIT, -- no Match any non-numeric character | |
117 | ALNUM, -- no Match any alphanumeric character | |
118 | NALNUM, -- no Match any non-alphanumeric character | |
119 | ||
120 | -- Branches | |
121 | ||
122 | BRANCH, -- node Match this alternative, or the next | |
123 | ||
124 | -- Simple loops (when the following node is one character in length) | |
125 | ||
126 | STAR, -- node Match this simple thing 0 or more times | |
127 | PLUS, -- node Match this simple thing 1 or more times | |
128 | CURLY, -- 2num node Match this simple thing between n and m times. | |
129 | ||
130 | -- Complex loops | |
131 | ||
132 | CURLYX, -- 2num node Match this complex thing {n,m} times | |
133 | -- The nums are coded on two characters each. | |
134 | ||
135 | WHILEM, -- no Do curly processing and see if rest matches | |
136 | ||
137 | -- Matches after or before a word | |
138 | ||
139 | BOL, -- no Match "" at beginning of line | |
140 | MBOL, -- no Same, assuming mutiline (match after \n) | |
141 | SBOL, -- no Same, assuming single line (don't match at \n) | |
142 | EOL, -- no Match "" at end of line | |
143 | MEOL, -- no Same, assuming mutiline (match before \n) | |
144 | SEOL, -- no Same, assuming single line (don't match at \n) | |
145 | ||
146 | BOUND, -- no Match "" at any word boundary | |
147 | NBOUND, -- no Match "" at any word non-boundary | |
148 | ||
149 | -- Parenthesis groups handling | |
150 | ||
151 | REFF, -- num Match some already matched string, folded | |
152 | OPEN, -- num Mark this point in input as start of #n | |
153 | CLOSE); -- num Analogous to OPEN | |
154 | ||
155 | for Opcode'Size use 8; | |
156 | ||
157 | -- Opcode notes: | |
158 | ||
159 | -- BRANCH | |
160 | -- The set of branches constituting a single choice are hooked | |
161 | -- together with their "next" pointers, since precedence prevents | |
162 | -- anything being concatenated to any individual branch. The | |
163 | -- "next" pointer of the last BRANCH in a choice points to the | |
164 | -- thing following the whole choice. This is also where the | |
165 | -- final "next" pointer of each individual branch points; each | |
166 | -- branch starts with the operand node of a BRANCH node. | |
167 | ||
168 | -- STAR,PLUS | |
169 | -- '?', and complex '*' and '+', are implemented with CURLYX. | |
170 | -- branches. Simple cases (one character per match) are implemented with | |
171 | -- STAR and PLUS for speed and to minimize recursive plunges. | |
172 | ||
173 | -- OPEN,CLOSE | |
174 | -- ...are numbered at compile time. | |
175 | ||
176 | -- EXACT, EXACTF | |
177 | -- There are in fact two arguments, the first one is the length (minus | |
178 | -- one of the string argument), coded on one character, the second | |
179 | -- argument is the string itself, coded on length + 1 characters. | |
180 | ||
181 | -- A node is one char of opcode followed by two chars of "next" pointer. | |
182 | -- "Next" pointers are stored as two 8-bit pieces, high order first. The | |
183 | -- value is a positive offset from the opcode of the node containing it. | |
184 | -- An operand, if any, simply follows the node. (Note that much of the | |
185 | -- code generation knows about this implicit relationship.) | |
186 | ||
187 | -- Using two bytes for the "next" pointer is vast overkill for most | |
188 | -- things, but allows patterns to get big without disasters. | |
189 | ||
190 | ----------------------- | |
191 | -- Character classes -- | |
192 | ----------------------- | |
193 | -- This is the implementation for character classes ([...]) in the | |
194 | -- syntax for regular expressions. Each character (0..256) has an | |
195 | -- entry into the table. This makes for a very fast matching | |
196 | -- algorithm. | |
197 | ||
198 | type Class_Byte is mod 256; | |
199 | type Character_Class is array (Class_Byte range 0 .. 31) of Class_Byte; | |
200 | ||
201 | type Bit_Conversion_Array is array (Class_Byte range 0 .. 7) of Class_Byte; | |
202 | Bit_Conversion : constant Bit_Conversion_Array := | |
203 | (1, 2, 4, 8, 16, 32, 64, 128); | |
204 | ||
205 | type Std_Class is (ANYOF_NONE, | |
206 | ANYOF_ALNUM, -- Alphanumeric class [a-zA-Z0-9] | |
207 | ANYOF_NALNUM, | |
208 | ANYOF_SPACE, -- Space class [ \t\n\r\f] | |
209 | ANYOF_NSPACE, | |
210 | ANYOF_DIGIT, -- Digit class [0-9] | |
211 | ANYOF_NDIGIT, | |
212 | ANYOF_ALNUMC, -- Alphanumeric class [a-zA-Z0-9] | |
213 | ANYOF_NALNUMC, | |
214 | ANYOF_ALPHA, -- Alpha class [a-zA-Z] | |
215 | ANYOF_NALPHA, | |
216 | ANYOF_ASCII, -- Ascii class (7 bits) 0..127 | |
217 | ANYOF_NASCII, | |
218 | ANYOF_CNTRL, -- Control class | |
219 | ANYOF_NCNTRL, | |
220 | ANYOF_GRAPH, -- Graphic class | |
221 | ANYOF_NGRAPH, | |
222 | ANYOF_LOWER, -- Lower case class [a-z] | |
223 | ANYOF_NLOWER, | |
224 | ANYOF_PRINT, -- printable class | |
225 | ANYOF_NPRINT, | |
226 | ANYOF_PUNCT, -- | |
227 | ANYOF_NPUNCT, | |
228 | ANYOF_UPPER, -- Upper case class [A-Z] | |
229 | ANYOF_NUPPER, | |
230 | ANYOF_XDIGIT, -- Hexadecimal digit | |
231 | ANYOF_NXDIGIT | |
232 | ); | |
233 | ||
234 | procedure Set_In_Class | |
235 | (Bitmap : in out Character_Class; | |
236 | C : Character); | |
237 | -- Set the entry to True for C in the class Bitmap. | |
238 | ||
239 | function Get_From_Class | |
240 | (Bitmap : Character_Class; | |
241 | C : Character) | |
242 | return Boolean; | |
243 | -- Return True if the entry is set for C in the class Bitmap. | |
244 | ||
245 | procedure Reset_Class (Bitmap : in out Character_Class); | |
246 | -- Clear all the entries in the class Bitmap. | |
247 | ||
248 | pragma Inline_Always (Set_In_Class); | |
249 | pragma Inline_Always (Get_From_Class); | |
250 | pragma Inline_Always (Reset_Class); | |
251 | ||
252 | ----------------------- | |
253 | -- Local Subprograms -- | |
254 | ----------------------- | |
255 | ||
256 | function "+" (Left : Opcode; Right : Integer) return Opcode; | |
257 | function "-" (Left : Opcode; Right : Opcode) return Integer; | |
258 | function "=" (Left : Character; Right : Opcode) return Boolean; | |
259 | ||
260 | function Is_Alnum (C : Character) return Boolean; | |
261 | -- Return True if C is an alphanum character or an underscore ('_') | |
262 | ||
263 | function Is_Space (C : Character) return Boolean; | |
264 | -- Return True if C is a whitespace character | |
265 | ||
266 | function Is_Printable (C : Character) return Boolean; | |
267 | -- Return True if C is a printable character | |
268 | ||
269 | function Operand (P : Pointer) return Pointer; | |
270 | -- Return a pointer to the first operand of the node at P | |
271 | ||
272 | function String_Length | |
273 | (Program : Program_Data; | |
274 | P : Pointer) | |
275 | return Program_Size; | |
276 | -- Return the length of the string argument of the node at P | |
277 | ||
278 | function String_Operand (P : Pointer) return Pointer; | |
279 | -- Return a pointer to the string argument of the node at P | |
280 | ||
281 | procedure Bitmap_Operand | |
282 | (Program : Program_Data; | |
283 | P : Pointer; | |
284 | Op : out Character_Class); | |
285 | -- Return a pointer to the string argument of the node at P | |
286 | ||
287 | function Get_Next_Offset | |
288 | (Program : Program_Data; | |
289 | IP : Pointer) | |
290 | return Pointer; | |
291 | -- Get the offset field of a node. Used by Get_Next. | |
292 | ||
293 | function Get_Next | |
294 | (Program : Program_Data; | |
295 | IP : Pointer) | |
296 | return Pointer; | |
297 | -- Dig the next instruction pointer out of a node | |
298 | ||
299 | procedure Optimize (Self : in out Pattern_Matcher); | |
300 | -- Optimize a Pattern_Matcher by noting certain special cases | |
301 | ||
302 | function Read_Natural | |
303 | (Program : Program_Data; | |
304 | IP : Pointer) | |
305 | return Natural; | |
306 | -- Return the 2-byte natural coded at position IP. | |
307 | ||
308 | -- All of the subprograms above are tiny and should be inlined | |
309 | ||
310 | pragma Inline ("+"); | |
311 | pragma Inline ("-"); | |
312 | pragma Inline ("="); | |
313 | pragma Inline (Is_Alnum); | |
314 | pragma Inline (Is_Space); | |
315 | pragma Inline (Get_Next); | |
316 | pragma Inline (Get_Next_Offset); | |
317 | pragma Inline (Operand); | |
318 | pragma Inline (Read_Natural); | |
319 | pragma Inline (String_Length); | |
320 | pragma Inline (String_Operand); | |
321 | ||
322 | type Expression_Flags is record | |
323 | Has_Width, -- Known never to match null string | |
324 | Simple, -- Simple enough to be STAR/PLUS operand | |
325 | SP_Start : Boolean; -- Starts with * or + | |
326 | end record; | |
327 | ||
328 | Worst_Expression : constant Expression_Flags := (others => False); | |
329 | -- Worst case | |
330 | ||
331 | --------- | |
332 | -- "+" -- | |
333 | --------- | |
334 | ||
335 | function "+" (Left : Opcode; Right : Integer) return Opcode is | |
336 | begin | |
337 | return Opcode'Val (Opcode'Pos (Left) + Right); | |
338 | end "+"; | |
339 | ||
340 | --------- | |
341 | -- "-" -- | |
342 | --------- | |
343 | ||
344 | function "-" (Left : Opcode; Right : Opcode) return Integer is | |
345 | begin | |
346 | return Opcode'Pos (Left) - Opcode'Pos (Right); | |
347 | end "-"; | |
348 | ||
349 | --------- | |
350 | -- "=" -- | |
351 | --------- | |
352 | ||
353 | function "=" (Left : Character; Right : Opcode) return Boolean is | |
354 | begin | |
355 | return Character'Pos (Left) = Opcode'Pos (Right); | |
356 | end "="; | |
357 | ||
358 | -------------------- | |
359 | -- Bitmap_Operand -- | |
360 | -------------------- | |
361 | ||
362 | procedure Bitmap_Operand | |
363 | (Program : Program_Data; | |
364 | P : Pointer; | |
365 | Op : out Character_Class) | |
366 | is | |
367 | function Convert is new Unchecked_Conversion | |
368 | (Program_Data, Character_Class); | |
369 | ||
370 | begin | |
371 | Op (0 .. 31) := Convert (Program (P + 3 .. P + 34)); | |
372 | end Bitmap_Operand; | |
373 | ||
374 | ------------- | |
375 | -- Compile -- | |
376 | ------------- | |
377 | ||
378 | procedure Compile | |
379 | (Matcher : out Pattern_Matcher; | |
380 | Expression : String; | |
381 | Final_Code_Size : out Program_Size; | |
382 | Flags : Regexp_Flags := No_Flags) | |
383 | is | |
384 | -- We can't allocate space until we know how big the compiled form | |
385 | -- will be, but we can't compile it (and thus know how big it is) | |
386 | -- until we've got a place to put the code. So we cheat: we compile | |
387 | -- it twice, once with code generation turned off and size counting | |
388 | -- turned on, and once "for real". | |
389 | ||
390 | -- This also means that we don't allocate space until we are sure | |
391 | -- that the thing really will compile successfully, and we never | |
392 | -- have to move the code and thus invalidate pointers into it. | |
393 | ||
394 | -- Beware that the optimization-preparation code in here knows | |
395 | -- about some of the structure of the compiled regexp. | |
396 | ||
397 | PM : Pattern_Matcher renames Matcher; | |
398 | Program : Program_Data renames PM.Program; | |
399 | ||
400 | Emit_Code : constant Boolean := PM.Size > 0; | |
401 | Emit_Ptr : Pointer := Program_First; | |
402 | ||
403 | Parse_Pos : Natural := Expression'First; -- Input-scan pointer | |
404 | Parse_End : Natural := Expression'Last; | |
405 | ||
406 | ---------------------------- | |
407 | -- Subprograms for Create -- | |
408 | ---------------------------- | |
409 | ||
410 | procedure Emit (B : Character); | |
411 | -- Output the Character to the Program. | |
412 | -- If code-generation is disables, simply increments the program | |
413 | -- counter. | |
414 | ||
415 | function Emit_Node (Op : Opcode) return Pointer; | |
416 | -- If code-generation is enabled, Emit_Node outputs the | |
417 | -- opcode and reserves space for a pointer to the next node. | |
418 | -- Return value is the location of new opcode, ie old Emit_Ptr. | |
419 | ||
420 | procedure Emit_Natural (IP : Pointer; N : Natural); | |
421 | -- Split N on two characters at position IP. | |
422 | ||
423 | procedure Emit_Class (Bitmap : Character_Class); | |
424 | -- Emits a character class. | |
425 | ||
426 | procedure Case_Emit (C : Character); | |
427 | -- Emit C, after converting is to lower-case if the regular | |
428 | -- expression is case insensitive. | |
429 | ||
430 | procedure Parse | |
431 | (Parenthesized : Boolean; | |
432 | Flags : in out Expression_Flags; | |
433 | IP : out Pointer); | |
434 | -- Parse regular expression, i.e. main body or parenthesized thing | |
435 | -- Caller must absorb opening parenthesis. | |
436 | ||
437 | procedure Parse_Branch | |
438 | (Flags : in out Expression_Flags; | |
439 | First : Boolean; | |
440 | IP : out Pointer); | |
441 | -- Implements the concatenation operator and handles '|' | |
442 | -- First should be true if this is the first item of the alternative. | |
443 | ||
444 | procedure Parse_Piece | |
445 | (Expr_Flags : in out Expression_Flags; IP : out Pointer); | |
446 | -- Parse something followed by possible [*+?] | |
447 | ||
448 | procedure Parse_Atom | |
449 | (Expr_Flags : in out Expression_Flags; IP : out Pointer); | |
450 | -- Parse_Atom is the lowest level parse procedure. | |
451 | -- Optimization: gobbles an entire sequence of ordinary characters | |
452 | -- so that it can turn them into a single node, which is smaller to | |
453 | -- store and faster to run. Backslashed characters are exceptions, | |
454 | -- each becoming a separate node; the code is simpler that way and | |
455 | -- it's not worth fixing. | |
456 | ||
457 | procedure Insert_Operator | |
458 | (Op : Opcode; | |
459 | Operand : Pointer; | |
460 | Greedy : Boolean := True); | |
461 | -- Insert_Operator inserts an operator in front of an | |
462 | -- already-emitted operand and relocates the operand. | |
463 | -- This applies to PLUS and STAR. | |
464 | -- If Minmod is True, then the operator is non-greedy. | |
465 | ||
466 | procedure Insert_Curly_Operator | |
467 | (Op : Opcode; | |
468 | Min : Natural; | |
469 | Max : Natural; | |
470 | Operand : Pointer; | |
471 | Greedy : Boolean := True); | |
472 | -- Insert an operator for CURLY ({Min}, {Min,} or {Min,Max}). | |
473 | -- If Minmod is True, then the operator is non-greedy. | |
474 | ||
475 | procedure Link_Tail (P, Val : Pointer); | |
476 | -- Link_Tail sets the next-pointer at the end of a node chain | |
477 | ||
478 | procedure Link_Operand_Tail (P, Val : Pointer); | |
479 | -- Link_Tail on operand of first argument; nop if operandless | |
480 | ||
481 | function Next_Instruction (P : Pointer) return Pointer; | |
482 | -- Dig the "next" pointer out of a node | |
483 | ||
484 | procedure Fail (M : in String); | |
485 | -- Fail with a diagnostic message, if possible | |
486 | ||
487 | function Is_Curly_Operator (IP : Natural) return Boolean; | |
488 | -- Return True if IP is looking at a '{' that is the beginning | |
489 | -- of a curly operator, ie it matches {\d+,?\d*} | |
490 | ||
491 | function Is_Mult (IP : Natural) return Boolean; | |
492 | -- Return True if C is a regexp multiplier: '+', '*' or '?' | |
493 | ||
494 | procedure Get_Curly_Arguments | |
495 | (IP : Natural; | |
496 | Min : out Natural; | |
497 | Max : out Natural; | |
498 | Greedy : out Boolean); | |
499 | -- Parse the argument list for a curly operator. | |
500 | -- It is assumed that IP is indeed pointing at a valid operator. | |
501 | ||
502 | procedure Parse_Character_Class (IP : out Pointer); | |
503 | -- Parse a character class. | |
504 | -- The calling subprogram should consume the opening '[' before. | |
505 | ||
506 | procedure Parse_Literal (Expr_Flags : in out Expression_Flags; | |
507 | IP : out Pointer); | |
508 | -- Parse_Literal encodes a string of characters | |
509 | -- to be matched exactly. | |
510 | ||
511 | function Parse_Posix_Character_Class return Std_Class; | |
512 | -- Parse a posic character class, like [:alpha:] or [:^alpha:]. | |
513 | -- The called is suppoed to absorbe the opening [. | |
514 | ||
515 | pragma Inline_Always (Is_Mult); | |
516 | pragma Inline_Always (Emit_Natural); | |
517 | pragma Inline_Always (Parse_Character_Class); -- since used only once | |
518 | ||
519 | --------------- | |
520 | -- Case_Emit -- | |
521 | --------------- | |
522 | ||
523 | procedure Case_Emit (C : Character) is | |
524 | begin | |
525 | if (Flags and Case_Insensitive) /= 0 then | |
526 | Emit (To_Lower (C)); | |
527 | ||
528 | else | |
529 | -- Dump current character | |
530 | ||
531 | Emit (C); | |
532 | end if; | |
533 | end Case_Emit; | |
534 | ||
535 | ---------- | |
536 | -- Emit -- | |
537 | ---------- | |
538 | ||
539 | procedure Emit (B : Character) is | |
540 | begin | |
541 | if Emit_Code then | |
542 | Program (Emit_Ptr) := B; | |
543 | end if; | |
544 | ||
545 | Emit_Ptr := Emit_Ptr + 1; | |
546 | end Emit; | |
547 | ||
548 | ---------------- | |
549 | -- Emit_Class -- | |
550 | ---------------- | |
551 | ||
552 | procedure Emit_Class (Bitmap : Character_Class) is | |
553 | subtype Program31 is Program_Data (0 .. 31); | |
554 | ||
555 | function Convert is new Unchecked_Conversion | |
556 | (Character_Class, Program31); | |
557 | ||
558 | begin | |
559 | if Emit_Code then | |
560 | Program (Emit_Ptr .. Emit_Ptr + 31) := Convert (Bitmap); | |
561 | end if; | |
562 | ||
563 | Emit_Ptr := Emit_Ptr + 32; | |
564 | end Emit_Class; | |
565 | ||
566 | ------------------ | |
567 | -- Emit_Natural -- | |
568 | ------------------ | |
569 | ||
570 | procedure Emit_Natural (IP : Pointer; N : Natural) is | |
571 | begin | |
572 | if Emit_Code then | |
573 | Program (IP + 1) := Character'Val (N / 256); | |
574 | Program (IP) := Character'Val (N mod 256); | |
575 | end if; | |
576 | end Emit_Natural; | |
577 | ||
578 | --------------- | |
579 | -- Emit_Node -- | |
580 | --------------- | |
581 | ||
582 | function Emit_Node (Op : Opcode) return Pointer is | |
583 | Result : constant Pointer := Emit_Ptr; | |
584 | ||
585 | begin | |
586 | if Emit_Code then | |
587 | Program (Emit_Ptr) := Character'Val (Opcode'Pos (Op)); | |
588 | Program (Emit_Ptr + 1) := ASCII.NUL; | |
589 | Program (Emit_Ptr + 2) := ASCII.NUL; | |
590 | end if; | |
591 | ||
592 | Emit_Ptr := Emit_Ptr + 3; | |
593 | return Result; | |
594 | end Emit_Node; | |
595 | ||
596 | ---------- | |
597 | -- Fail -- | |
598 | ---------- | |
599 | ||
600 | procedure Fail (M : in String) is | |
601 | begin | |
602 | raise Expression_Error; | |
603 | end Fail; | |
604 | ||
605 | ------------------------- | |
606 | -- Get_Curly_Arguments -- | |
607 | ------------------------- | |
608 | ||
609 | procedure Get_Curly_Arguments | |
610 | (IP : Natural; | |
611 | Min : out Natural; | |
612 | Max : out Natural; | |
613 | Greedy : out Boolean) | |
614 | is | |
615 | Save_Pos : Natural := Parse_Pos + 1; | |
616 | ||
617 | begin | |
618 | Min := 0; | |
619 | Max := Max_Curly_Repeat; | |
620 | ||
621 | while Expression (Parse_Pos) /= '}' | |
622 | and then Expression (Parse_Pos) /= ',' | |
623 | loop | |
624 | Parse_Pos := Parse_Pos + 1; | |
625 | end loop; | |
626 | ||
627 | Min := Natural'Value (Expression (Save_Pos .. Parse_Pos - 1)); | |
628 | ||
629 | if Expression (Parse_Pos) = ',' then | |
630 | Save_Pos := Parse_Pos + 1; | |
631 | while Expression (Parse_Pos) /= '}' loop | |
632 | Parse_Pos := Parse_Pos + 1; | |
633 | end loop; | |
634 | ||
635 | if Save_Pos /= Parse_Pos then | |
636 | Max := Natural'Value (Expression (Save_Pos .. Parse_Pos - 1)); | |
637 | end if; | |
638 | ||
639 | else | |
640 | Max := Min; | |
641 | end if; | |
642 | ||
643 | if Parse_Pos < Expression'Last | |
644 | and then Expression (Parse_Pos + 1) = '?' | |
645 | then | |
646 | Greedy := False; | |
647 | Parse_Pos := Parse_Pos + 1; | |
648 | ||
649 | else | |
650 | Greedy := True; | |
651 | end if; | |
652 | end Get_Curly_Arguments; | |
653 | ||
654 | --------------------------- | |
655 | -- Insert_Curly_Operator -- | |
656 | --------------------------- | |
657 | ||
658 | procedure Insert_Curly_Operator | |
659 | (Op : Opcode; | |
660 | Min : Natural; | |
661 | Max : Natural; | |
662 | Operand : Pointer; | |
663 | Greedy : Boolean := True) | |
664 | is | |
665 | Dest : constant Pointer := Emit_Ptr; | |
666 | Old : Pointer; | |
667 | Size : Pointer := 7; | |
668 | ||
669 | begin | |
670 | -- If the operand is not greedy, insert an extra operand before it | |
671 | ||
672 | if not Greedy then | |
673 | Size := Size + 3; | |
674 | end if; | |
675 | ||
676 | -- Move the operand in the byte-compilation, so that we can insert | |
677 | -- the operator before it. | |
678 | ||
679 | if Emit_Code then | |
680 | Program (Operand + Size .. Emit_Ptr + Size) := | |
681 | Program (Operand .. Emit_Ptr); | |
682 | end if; | |
683 | ||
684 | -- Insert the operator at the position previously occupied by the | |
685 | -- operand. | |
686 | ||
687 | Emit_Ptr := Operand; | |
688 | ||
689 | if not Greedy then | |
690 | Old := Emit_Node (MINMOD); | |
691 | Link_Tail (Old, Old + 3); | |
692 | end if; | |
693 | ||
694 | Old := Emit_Node (Op); | |
695 | Emit_Natural (Old + 3, Min); | |
696 | Emit_Natural (Old + 5, Max); | |
697 | ||
698 | Emit_Ptr := Dest + Size; | |
699 | end Insert_Curly_Operator; | |
700 | ||
701 | --------------------- | |
702 | -- Insert_Operator -- | |
703 | --------------------- | |
704 | ||
705 | procedure Insert_Operator | |
706 | (Op : Opcode; | |
707 | Operand : Pointer; | |
708 | Greedy : Boolean := True) | |
709 | is | |
710 | Dest : constant Pointer := Emit_Ptr; | |
711 | Old : Pointer; | |
712 | Size : Pointer := 3; | |
713 | ||
714 | begin | |
715 | -- If not greedy, we have to emit another opcode first | |
716 | ||
717 | if not Greedy then | |
718 | Size := Size + 3; | |
719 | end if; | |
720 | ||
721 | -- Move the operand in the byte-compilation, so that we can insert | |
722 | -- the operator before it. | |
723 | ||
724 | if Emit_Code then | |
725 | Program (Operand + Size .. Emit_Ptr + Size) | |
726 | := Program (Operand .. Emit_Ptr); | |
727 | end if; | |
728 | ||
729 | -- Insert the operator at the position previously occupied by the | |
730 | -- operand. | |
731 | ||
732 | Emit_Ptr := Operand; | |
733 | ||
734 | if not Greedy then | |
735 | Old := Emit_Node (MINMOD); | |
736 | Link_Tail (Old, Old + 3); | |
737 | end if; | |
738 | ||
739 | Old := Emit_Node (Op); | |
740 | Emit_Ptr := Dest + Size; | |
741 | end Insert_Operator; | |
742 | ||
743 | ----------------------- | |
744 | -- Is_Curly_Operator -- | |
745 | ----------------------- | |
746 | ||
747 | function Is_Curly_Operator (IP : Natural) return Boolean is | |
748 | Scan : Natural := IP; | |
749 | ||
750 | begin | |
751 | if Expression (Scan) /= '{' | |
752 | or else Scan + 2 > Expression'Last | |
753 | or else not Is_Digit (Expression (Scan + 1)) | |
754 | then | |
755 | return False; | |
756 | end if; | |
757 | ||
758 | Scan := Scan + 1; | |
759 | ||
760 | -- The first digit | |
761 | ||
762 | loop | |
763 | Scan := Scan + 1; | |
764 | ||
765 | if Scan > Expression'Last then | |
766 | return False; | |
767 | end if; | |
768 | ||
769 | exit when not Is_Digit (Expression (Scan)); | |
770 | end loop; | |
771 | ||
772 | if Expression (Scan) = ',' then | |
773 | loop | |
774 | Scan := Scan + 1; | |
775 | ||
776 | if Scan > Expression'Last then | |
777 | return False; | |
778 | end if; | |
779 | ||
780 | exit when not Is_Digit (Expression (Scan)); | |
781 | end loop; | |
782 | end if; | |
783 | ||
784 | return Expression (Scan) = '}'; | |
785 | end Is_Curly_Operator; | |
786 | ||
787 | ------------- | |
788 | -- Is_Mult -- | |
789 | ------------- | |
790 | ||
791 | function Is_Mult (IP : Natural) return Boolean is | |
792 | C : constant Character := Expression (IP); | |
793 | ||
794 | begin | |
795 | return C = '*' | |
796 | or else C = '+' | |
797 | or else C = '?' | |
798 | or else (C = '{' and then Is_Curly_Operator (IP)); | |
799 | end Is_Mult; | |
800 | ||
801 | ----------------------- | |
802 | -- Link_Operand_Tail -- | |
803 | ----------------------- | |
804 | ||
805 | procedure Link_Operand_Tail (P, Val : Pointer) is | |
806 | begin | |
807 | if Emit_Code and then Program (P) = BRANCH then | |
808 | Link_Tail (Operand (P), Val); | |
809 | end if; | |
810 | end Link_Operand_Tail; | |
811 | ||
812 | --------------- | |
813 | -- Link_Tail -- | |
814 | --------------- | |
815 | ||
816 | procedure Link_Tail (P, Val : Pointer) is | |
817 | Scan : Pointer; | |
818 | Temp : Pointer; | |
819 | Offset : Pointer; | |
820 | ||
821 | begin | |
822 | if not Emit_Code then | |
823 | return; | |
824 | end if; | |
825 | ||
826 | -- Find last node | |
827 | ||
828 | Scan := P; | |
829 | loop | |
830 | Temp := Next_Instruction (Scan); | |
831 | exit when Temp = 0; | |
832 | Scan := Temp; | |
833 | end loop; | |
834 | ||
835 | Offset := Val - Scan; | |
836 | ||
837 | Emit_Natural (Scan + 1, Natural (Offset)); | |
838 | end Link_Tail; | |
839 | ||
840 | ---------------------- | |
841 | -- Next_Instruction -- | |
842 | ---------------------- | |
843 | ||
844 | function Next_Instruction (P : Pointer) return Pointer is | |
845 | Offset : Pointer; | |
846 | ||
847 | begin | |
848 | if not Emit_Code then | |
849 | return 0; | |
850 | end if; | |
851 | ||
852 | Offset := Get_Next_Offset (Program, P); | |
853 | ||
854 | if Offset = 0 then | |
855 | return 0; | |
856 | end if; | |
857 | ||
858 | return P + Offset; | |
859 | end Next_Instruction; | |
860 | ||
861 | ----------- | |
862 | -- Parse -- | |
863 | ----------- | |
864 | ||
865 | -- Combining parenthesis handling with the base level | |
866 | -- of regular expression is a trifle forced, but the | |
867 | -- need to tie the tails of the branches to what follows | |
868 | -- makes it hard to avoid. | |
869 | ||
870 | procedure Parse | |
871 | (Parenthesized : in Boolean; | |
872 | Flags : in out Expression_Flags; | |
873 | IP : out Pointer) | |
874 | is | |
875 | E : String renames Expression; | |
876 | Br : Pointer; | |
877 | Ender : Pointer; | |
878 | Par_No : Natural; | |
879 | New_Flags : Expression_Flags; | |
880 | Have_Branch : Boolean := False; | |
881 | ||
882 | begin | |
883 | Flags := (Has_Width => True, others => False); -- Tentatively | |
884 | ||
885 | -- Make an OPEN node, if parenthesized | |
886 | ||
887 | if Parenthesized then | |
888 | if Matcher.Paren_Count > Max_Paren_Count then | |
889 | Fail ("too many ()"); | |
890 | end if; | |
891 | ||
892 | Par_No := Matcher.Paren_Count + 1; | |
893 | Matcher.Paren_Count := Matcher.Paren_Count + 1; | |
894 | IP := Emit_Node (OPEN); | |
895 | Emit (Character'Val (Par_No)); | |
896 | ||
897 | else | |
898 | IP := 0; | |
899 | end if; | |
900 | ||
901 | -- Pick up the branches, linking them together | |
902 | ||
903 | Parse_Branch (New_Flags, True, Br); | |
904 | ||
905 | if Br = 0 then | |
906 | IP := 0; | |
907 | return; | |
908 | end if; | |
909 | ||
910 | if Parse_Pos <= Parse_End | |
911 | and then E (Parse_Pos) = '|' | |
912 | then | |
913 | Insert_Operator (BRANCH, Br); | |
914 | Have_Branch := True; | |
915 | end if; | |
916 | ||
917 | if IP /= 0 then | |
918 | Link_Tail (IP, Br); -- OPEN -> first | |
919 | else | |
920 | IP := Br; | |
921 | end if; | |
922 | ||
923 | if not New_Flags.Has_Width then | |
924 | Flags.Has_Width := False; | |
925 | end if; | |
926 | ||
927 | Flags.SP_Start := Flags.SP_Start or New_Flags.SP_Start; | |
928 | ||
929 | while Parse_Pos <= Parse_End | |
930 | and then (E (Parse_Pos) = '|') | |
931 | loop | |
932 | Parse_Pos := Parse_Pos + 1; | |
933 | Parse_Branch (New_Flags, False, Br); | |
934 | ||
935 | if Br = 0 then | |
936 | IP := 0; | |
937 | return; | |
938 | end if; | |
939 | ||
940 | Link_Tail (IP, Br); -- BRANCH -> BRANCH | |
941 | ||
942 | if not New_Flags.Has_Width then | |
943 | Flags.Has_Width := False; | |
944 | end if; | |
945 | ||
946 | Flags.SP_Start := Flags.SP_Start or New_Flags.SP_Start; | |
947 | end loop; | |
948 | ||
949 | -- Make a closing node, and hook it on the end | |
950 | ||
951 | if Parenthesized then | |
952 | Ender := Emit_Node (CLOSE); | |
953 | Emit (Character'Val (Par_No)); | |
954 | else | |
955 | Ender := Emit_Node (EOP); | |
956 | end if; | |
957 | ||
958 | Link_Tail (IP, Ender); | |
959 | ||
960 | if Have_Branch then | |
961 | ||
962 | -- Hook the tails of the branches to the closing node | |
963 | ||
964 | Br := IP; | |
965 | loop | |
966 | exit when Br = 0; | |
967 | Link_Operand_Tail (Br, Ender); | |
968 | Br := Next_Instruction (Br); | |
969 | end loop; | |
970 | end if; | |
971 | ||
972 | -- Check for proper termination | |
973 | ||
974 | if Parenthesized then | |
975 | if Parse_Pos > Parse_End or else E (Parse_Pos) /= ')' then | |
976 | Fail ("unmatched ()"); | |
977 | end if; | |
978 | ||
979 | Parse_Pos := Parse_Pos + 1; | |
980 | ||
981 | elsif Parse_Pos <= Parse_End then | |
982 | if E (Parse_Pos) = ')' then | |
983 | Fail ("unmatched ()"); | |
984 | else | |
985 | Fail ("junk on end"); -- "Can't happen" | |
986 | end if; | |
987 | end if; | |
988 | end Parse; | |
989 | ||
990 | ---------------- | |
991 | -- Parse_Atom -- | |
992 | ---------------- | |
993 | ||
994 | procedure Parse_Atom | |
995 | (Expr_Flags : in out Expression_Flags; | |
996 | IP : out Pointer) | |
997 | is | |
998 | C : Character; | |
999 | ||
1000 | begin | |
1001 | -- Tentatively set worst expression case | |
1002 | ||
1003 | Expr_Flags := Worst_Expression; | |
1004 | ||
1005 | C := Expression (Parse_Pos); | |
1006 | Parse_Pos := Parse_Pos + 1; | |
1007 | ||
1008 | case (C) is | |
1009 | when '^' => | |
1010 | if (Flags and Multiple_Lines) /= 0 then | |
1011 | IP := Emit_Node (MBOL); | |
1012 | elsif (Flags and Single_Line) /= 0 then | |
1013 | IP := Emit_Node (SBOL); | |
1014 | else | |
1015 | IP := Emit_Node (BOL); | |
1016 | end if; | |
1017 | ||
1018 | when '$' => | |
1019 | if (Flags and Multiple_Lines) /= 0 then | |
1020 | IP := Emit_Node (MEOL); | |
1021 | elsif (Flags and Single_Line) /= 0 then | |
1022 | IP := Emit_Node (SEOL); | |
1023 | else | |
1024 | IP := Emit_Node (EOL); | |
1025 | end if; | |
1026 | ||
1027 | when '.' => | |
1028 | if (Flags and Single_Line) /= 0 then | |
1029 | IP := Emit_Node (SANY); | |
1030 | else | |
1031 | IP := Emit_Node (ANY); | |
1032 | end if; | |
1033 | Expr_Flags.Has_Width := True; | |
1034 | Expr_Flags.Simple := True; | |
1035 | ||
1036 | when '[' => | |
1037 | Parse_Character_Class (IP); | |
1038 | Expr_Flags.Has_Width := True; | |
1039 | Expr_Flags.Simple := True; | |
1040 | ||
1041 | when '(' => | |
1042 | declare | |
1043 | New_Flags : Expression_Flags; | |
1044 | ||
1045 | begin | |
1046 | Parse (True, New_Flags, IP); | |
1047 | ||
1048 | if IP = 0 then | |
1049 | return; | |
1050 | end if; | |
1051 | ||
1052 | Expr_Flags.Has_Width := | |
1053 | Expr_Flags.Has_Width or New_Flags.Has_Width; | |
1054 | Expr_Flags.SP_Start := | |
1055 | Expr_Flags.SP_Start or New_Flags.SP_Start; | |
1056 | end; | |
1057 | ||
1058 | when '|' | ASCII.LF | ')' => | |
1059 | Fail ("internal urp"); -- Supposed to be caught earlier | |
1060 | ||
1061 | when '?' | '+' | '*' | '{' => | |
1062 | Fail ("?+*{ follows nothing"); | |
1063 | ||
1064 | when '\' => | |
1065 | if Parse_Pos > Parse_End then | |
1066 | Fail ("trailing \"); | |
1067 | end if; | |
1068 | ||
1069 | Parse_Pos := Parse_Pos + 1; | |
1070 | ||
1071 | case Expression (Parse_Pos - 1) is | |
1072 | when 'b' => | |
1073 | IP := Emit_Node (BOUND); | |
1074 | ||
1075 | when 'B' => | |
1076 | IP := Emit_Node (NBOUND); | |
1077 | ||
1078 | when 's' => | |
1079 | IP := Emit_Node (SPACE); | |
1080 | Expr_Flags.Simple := True; | |
1081 | Expr_Flags.Has_Width := True; | |
1082 | ||
1083 | when 'S' => | |
1084 | IP := Emit_Node (NSPACE); | |
1085 | Expr_Flags.Simple := True; | |
1086 | Expr_Flags.Has_Width := True; | |
1087 | ||
1088 | when 'd' => | |
1089 | IP := Emit_Node (DIGIT); | |
1090 | Expr_Flags.Simple := True; | |
1091 | Expr_Flags.Has_Width := True; | |
1092 | ||
1093 | when 'D' => | |
1094 | IP := Emit_Node (NDIGIT); | |
1095 | Expr_Flags.Simple := True; | |
1096 | Expr_Flags.Has_Width := True; | |
1097 | ||
1098 | when 'w' => | |
1099 | IP := Emit_Node (ALNUM); | |
1100 | Expr_Flags.Simple := True; | |
1101 | Expr_Flags.Has_Width := True; | |
1102 | ||
1103 | when 'W' => | |
1104 | IP := Emit_Node (NALNUM); | |
1105 | Expr_Flags.Simple := True; | |
1106 | Expr_Flags.Has_Width := True; | |
1107 | ||
1108 | when 'A' => | |
1109 | IP := Emit_Node (SBOL); | |
1110 | ||
1111 | when 'G' => | |
1112 | IP := Emit_Node (SEOL); | |
1113 | ||
1114 | when '0' .. '9' => | |
1115 | IP := Emit_Node (REFF); | |
1116 | ||
1117 | declare | |
1118 | Save : Natural := Parse_Pos - 1; | |
1119 | ||
1120 | begin | |
1121 | while Parse_Pos <= Expression'Last | |
1122 | and then Is_Digit (Expression (Parse_Pos)) | |
1123 | loop | |
1124 | Parse_Pos := Parse_Pos + 1; | |
1125 | end loop; | |
1126 | ||
1127 | Emit (Character'Val (Natural'Value | |
1128 | (Expression (Save .. Parse_Pos - 1)))); | |
1129 | end; | |
1130 | ||
1131 | when others => | |
1132 | Parse_Pos := Parse_Pos - 1; | |
1133 | Parse_Literal (Expr_Flags, IP); | |
1134 | end case; | |
1135 | ||
1136 | when others => Parse_Literal (Expr_Flags, IP); | |
1137 | end case; | |
1138 | end Parse_Atom; | |
1139 | ||
1140 | ------------------ | |
1141 | -- Parse_Branch -- | |
1142 | ------------------ | |
1143 | ||
1144 | procedure Parse_Branch | |
1145 | (Flags : in out Expression_Flags; | |
1146 | First : Boolean; | |
1147 | IP : out Pointer) | |
1148 | is | |
1149 | E : String renames Expression; | |
1150 | Chain : Pointer; | |
1151 | Last : Pointer; | |
1152 | New_Flags : Expression_Flags; | |
1153 | Dummy : Pointer; | |
1154 | ||
1155 | begin | |
1156 | Flags := Worst_Expression; -- Tentatively | |
1157 | ||
1158 | if First then | |
1159 | IP := Emit_Ptr; | |
1160 | else | |
1161 | IP := Emit_Node (BRANCH); | |
1162 | end if; | |
1163 | ||
1164 | Chain := 0; | |
1165 | ||
1166 | while Parse_Pos <= Parse_End | |
1167 | and then E (Parse_Pos) /= ')' | |
1168 | and then E (Parse_Pos) /= ASCII.LF | |
1169 | and then E (Parse_Pos) /= '|' | |
1170 | loop | |
1171 | Parse_Piece (New_Flags, Last); | |
1172 | ||
1173 | if Last = 0 then | |
1174 | IP := 0; | |
1175 | return; | |
1176 | end if; | |
1177 | ||
1178 | Flags.Has_Width := Flags.Has_Width or New_Flags.Has_Width; | |
1179 | ||
1180 | if Chain = 0 then -- First piece | |
1181 | Flags.SP_Start := Flags.SP_Start or New_Flags.SP_Start; | |
1182 | else | |
1183 | Link_Tail (Chain, Last); | |
1184 | end if; | |
1185 | ||
1186 | Chain := Last; | |
1187 | end loop; | |
1188 | ||
1189 | if Chain = 0 then -- Loop ran zero CURLY | |
1190 | Dummy := Emit_Node (NOTHING); | |
1191 | end if; | |
1192 | ||
1193 | end Parse_Branch; | |
1194 | ||
1195 | --------------------------- | |
1196 | -- Parse_Character_Class -- | |
1197 | --------------------------- | |
1198 | ||
1199 | procedure Parse_Character_Class (IP : out Pointer) is | |
1200 | Bitmap : Character_Class; | |
1201 | Invert : Boolean := False; | |
1202 | In_Range : Boolean := False; | |
1203 | Named_Class : Std_Class := ANYOF_NONE; | |
1204 | Value : Character; | |
1205 | Last_Value : Character := ASCII.Nul; | |
1206 | ||
1207 | begin | |
1208 | Reset_Class (Bitmap); | |
1209 | ||
1210 | -- Do we have an invert character class ? | |
1211 | ||
1212 | if Parse_Pos <= Parse_End | |
1213 | and then Expression (Parse_Pos) = '^' | |
1214 | then | |
1215 | Invert := True; | |
1216 | Parse_Pos := Parse_Pos + 1; | |
1217 | end if; | |
1218 | ||
1219 | -- First character can be ] or -, without closing the class. | |
1220 | ||
1221 | if Parse_Pos <= Parse_End | |
1222 | and then (Expression (Parse_Pos) = ']' | |
1223 | or else Expression (Parse_Pos) = '-') | |
1224 | then | |
1225 | Set_In_Class (Bitmap, Expression (Parse_Pos)); | |
1226 | Parse_Pos := Parse_Pos + 1; | |
1227 | end if; | |
1228 | ||
1229 | -- While we don't have the end of the class | |
1230 | ||
1231 | while Parse_Pos <= Parse_End | |
1232 | and then Expression (Parse_Pos) /= ']' | |
1233 | loop | |
1234 | Named_Class := ANYOF_NONE; | |
1235 | Value := Expression (Parse_Pos); | |
1236 | Parse_Pos := Parse_Pos + 1; | |
1237 | ||
1238 | -- Do we have a Posix character class | |
1239 | if Value = '[' then | |
1240 | Named_Class := Parse_Posix_Character_Class; | |
1241 | ||
1242 | elsif Value = '\' then | |
1243 | if Parse_Pos = Parse_End then | |
1244 | Fail ("Trailing \"); | |
1245 | end if; | |
1246 | Value := Expression (Parse_Pos); | |
1247 | Parse_Pos := Parse_Pos + 1; | |
1248 | ||
1249 | case Value is | |
1250 | when 'w' => Named_Class := ANYOF_ALNUM; | |
1251 | when 'W' => Named_Class := ANYOF_NALNUM; | |
1252 | when 's' => Named_Class := ANYOF_SPACE; | |
1253 | when 'S' => Named_Class := ANYOF_NSPACE; | |
1254 | when 'd' => Named_Class := ANYOF_DIGIT; | |
1255 | when 'D' => Named_Class := ANYOF_NDIGIT; | |
1256 | when 'n' => Value := ASCII.LF; | |
1257 | when 'r' => Value := ASCII.CR; | |
1258 | when 't' => Value := ASCII.HT; | |
1259 | when 'f' => Value := ASCII.FF; | |
1260 | when 'e' => Value := ASCII.ESC; | |
1261 | when 'a' => Value := ASCII.BEL; | |
1262 | ||
1263 | -- when 'x' => ??? hexadecimal value | |
1264 | -- when 'c' => ??? control character | |
1265 | -- when '0'..'9' => ??? octal character | |
1266 | ||
1267 | when others => null; | |
1268 | end case; | |
1269 | end if; | |
1270 | ||
1271 | -- Do we have a character class? | |
1272 | ||
1273 | if Named_Class /= ANYOF_NONE then | |
1274 | ||
1275 | -- A range like 'a-\d' or 'a-[:digit:] is not a range | |
1276 | ||
1277 | if In_Range then | |
1278 | Set_In_Class (Bitmap, Last_Value); | |
1279 | Set_In_Class (Bitmap, '-'); | |
1280 | In_Range := False; | |
1281 | end if; | |
1282 | ||
1283 | -- Expand the range | |
1284 | ||
1285 | case Named_Class is | |
1286 | when ANYOF_NONE => null; | |
1287 | ||
1288 | when ANYOF_ALNUM | ANYOF_ALNUMC => | |
1289 | for Value in Class_Byte'Range loop | |
1290 | if Is_Alnum (Character'Val (Value)) then | |
1291 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1292 | end if; | |
1293 | end loop; | |
1294 | ||
1295 | when ANYOF_NALNUM | ANYOF_NALNUMC => | |
1296 | for Value in Class_Byte'Range loop | |
1297 | if not Is_Alnum (Character'Val (Value)) then | |
1298 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1299 | end if; | |
1300 | end loop; | |
1301 | ||
1302 | when ANYOF_SPACE => | |
1303 | for Value in Class_Byte'Range loop | |
1304 | if Is_Space (Character'Val (Value)) then | |
1305 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1306 | end if; | |
1307 | end loop; | |
1308 | ||
1309 | when ANYOF_NSPACE => | |
1310 | for Value in Class_Byte'Range loop | |
1311 | if not Is_Space (Character'Val (Value)) then | |
1312 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1313 | end if; | |
1314 | end loop; | |
1315 | ||
1316 | when ANYOF_DIGIT => | |
1317 | for Value in Class_Byte'Range loop | |
1318 | if Is_Digit (Character'Val (Value)) then | |
1319 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1320 | end if; | |
1321 | end loop; | |
1322 | ||
1323 | when ANYOF_NDIGIT => | |
1324 | for Value in Class_Byte'Range loop | |
1325 | if not Is_Digit (Character'Val (Value)) then | |
1326 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1327 | end if; | |
1328 | end loop; | |
1329 | ||
1330 | when ANYOF_ALPHA => | |
1331 | for Value in Class_Byte'Range loop | |
1332 | if Is_Letter (Character'Val (Value)) then | |
1333 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1334 | end if; | |
1335 | end loop; | |
1336 | ||
1337 | when ANYOF_NALPHA => | |
1338 | for Value in Class_Byte'Range loop | |
1339 | if not Is_Letter (Character'Val (Value)) then | |
1340 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1341 | end if; | |
1342 | end loop; | |
1343 | ||
1344 | when ANYOF_ASCII => | |
1345 | for Value in 0 .. 127 loop | |
1346 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1347 | end loop; | |
1348 | ||
1349 | when ANYOF_NASCII => | |
1350 | for Value in 128 .. 255 loop | |
1351 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1352 | end loop; | |
1353 | ||
1354 | when ANYOF_CNTRL => | |
1355 | for Value in Class_Byte'Range loop | |
1356 | if Is_Control (Character'Val (Value)) then | |
1357 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1358 | end if; | |
1359 | end loop; | |
1360 | ||
1361 | when ANYOF_NCNTRL => | |
1362 | for Value in Class_Byte'Range loop | |
1363 | if not Is_Control (Character'Val (Value)) then | |
1364 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1365 | end if; | |
1366 | end loop; | |
1367 | ||
1368 | when ANYOF_GRAPH => | |
1369 | for Value in Class_Byte'Range loop | |
1370 | if Is_Graphic (Character'Val (Value)) then | |
1371 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1372 | end if; | |
1373 | end loop; | |
1374 | ||
1375 | when ANYOF_NGRAPH => | |
1376 | for Value in Class_Byte'Range loop | |
1377 | if not Is_Graphic (Character'Val (Value)) then | |
1378 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1379 | end if; | |
1380 | end loop; | |
1381 | ||
1382 | when ANYOF_LOWER => | |
1383 | for Value in Class_Byte'Range loop | |
1384 | if Is_Lower (Character'Val (Value)) then | |
1385 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1386 | end if; | |
1387 | end loop; | |
1388 | ||
1389 | when ANYOF_NLOWER => | |
1390 | for Value in Class_Byte'Range loop | |
1391 | if not Is_Lower (Character'Val (Value)) then | |
1392 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1393 | end if; | |
1394 | end loop; | |
1395 | ||
1396 | when ANYOF_PRINT => | |
1397 | for Value in Class_Byte'Range loop | |
1398 | if Is_Printable (Character'Val (Value)) then | |
1399 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1400 | end if; | |
1401 | end loop; | |
1402 | ||
1403 | when ANYOF_NPRINT => | |
1404 | for Value in Class_Byte'Range loop | |
1405 | if not Is_Printable (Character'Val (Value)) then | |
1406 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1407 | end if; | |
1408 | end loop; | |
1409 | ||
1410 | when ANYOF_PUNCT => | |
1411 | for Value in Class_Byte'Range loop | |
1412 | if Is_Printable (Character'Val (Value)) | |
1413 | and then not Is_Space (Character'Val (Value)) | |
1414 | and then not Is_Alnum (Character'Val (Value)) | |
1415 | then | |
1416 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1417 | end if; | |
1418 | end loop; | |
1419 | ||
1420 | when ANYOF_NPUNCT => | |
1421 | for Value in Class_Byte'Range loop | |
1422 | if not Is_Printable (Character'Val (Value)) | |
1423 | or else Is_Space (Character'Val (Value)) | |
1424 | or else Is_Alnum (Character'Val (Value)) | |
1425 | then | |
1426 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1427 | end if; | |
1428 | end loop; | |
1429 | ||
1430 | when ANYOF_UPPER => | |
1431 | for Value in Class_Byte'Range loop | |
1432 | if Is_Upper (Character'Val (Value)) then | |
1433 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1434 | end if; | |
1435 | end loop; | |
1436 | ||
1437 | when ANYOF_NUPPER => | |
1438 | for Value in Class_Byte'Range loop | |
1439 | if not Is_Upper (Character'Val (Value)) then | |
1440 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1441 | end if; | |
1442 | end loop; | |
1443 | ||
1444 | when ANYOF_XDIGIT => | |
1445 | for Value in Class_Byte'Range loop | |
1446 | if Is_Hexadecimal_Digit (Character'Val (Value)) then | |
1447 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1448 | end if; | |
1449 | end loop; | |
1450 | ||
1451 | when ANYOF_NXDIGIT => | |
1452 | for Value in Class_Byte'Range loop | |
1453 | if not Is_Hexadecimal_Digit | |
1454 | (Character'Val (Value)) | |
1455 | then | |
1456 | Set_In_Class (Bitmap, Character'Val (Value)); | |
1457 | end if; | |
1458 | end loop; | |
1459 | ||
1460 | end case; | |
1461 | ||
1462 | -- Not a character range | |
1463 | ||
1464 | elsif not In_Range then | |
1465 | Last_Value := Value; | |
1466 | ||
1467 | if Expression (Parse_Pos) = '-' | |
1468 | and then Parse_Pos < Parse_End | |
1469 | and then Expression (Parse_Pos + 1) /= ']' | |
1470 | then | |
1471 | Parse_Pos := Parse_Pos + 1; | |
1472 | ||
1473 | -- Do we have a range like '\d-a' and '[:space:]-a' | |
1474 | -- which is not a real range | |
1475 | ||
1476 | if Named_Class /= ANYOF_NONE then | |
1477 | Set_In_Class (Bitmap, '-'); | |
1478 | else | |
1479 | In_Range := True; | |
1480 | end if; | |
1481 | ||
1482 | else | |
1483 | Set_In_Class (Bitmap, Value); | |
1484 | ||
1485 | end if; | |
1486 | ||
1487 | -- Else in a character range | |
1488 | ||
1489 | else | |
1490 | if Last_Value > Value then | |
1491 | Fail ("Invalid Range [" & Last_Value'Img | |
1492 | & "-" & Value'Img & "]"); | |
1493 | end if; | |
1494 | ||
1495 | while Last_Value <= Value loop | |
1496 | Set_In_Class (Bitmap, Last_Value); | |
1497 | Last_Value := Character'Succ (Last_Value); | |
1498 | end loop; | |
1499 | ||
1500 | In_Range := False; | |
1501 | ||
1502 | end if; | |
1503 | ||
1504 | end loop; | |
1505 | ||
1506 | -- Optimize case-insensitive ranges (put the upper case or lower | |
1507 | -- case character into the bitmap) | |
1508 | ||
1509 | if (Flags and Case_Insensitive) /= 0 then | |
1510 | for C in Character'Range loop | |
1511 | if Get_From_Class (Bitmap, C) then | |
1512 | Set_In_Class (Bitmap, To_Lower (C)); | |
1513 | Set_In_Class (Bitmap, To_Upper (C)); | |
1514 | end if; | |
1515 | end loop; | |
1516 | end if; | |
1517 | ||
1518 | -- Optimize inverted classes | |
1519 | ||
1520 | if Invert then | |
1521 | for J in Bitmap'Range loop | |
1522 | Bitmap (J) := not Bitmap (J); | |
1523 | end loop; | |
1524 | end if; | |
1525 | ||
1526 | Parse_Pos := Parse_Pos + 1; | |
1527 | ||
1528 | -- Emit the class | |
1529 | ||
1530 | IP := Emit_Node (ANYOF); | |
1531 | Emit_Class (Bitmap); | |
1532 | end Parse_Character_Class; | |
1533 | ||
1534 | ------------------- | |
1535 | -- Parse_Literal -- | |
1536 | ------------------- | |
1537 | ||
1538 | -- This is a bit tricky due to quoted chars and due to | |
1539 | -- the multiplier characters '*', '+', and '?' that | |
1540 | -- take the SINGLE char previous as their operand. | |
1541 | -- | |
1542 | -- On entry, the character at Parse_Pos - 1 is going to go | |
1543 | -- into the string, no matter what it is. It could be | |
1544 | -- following a \ if Parse_Atom was entered from the '\' case. | |
1545 | -- | |
1546 | -- Basic idea is to pick up a good char in C and examine | |
1547 | -- the next char. If Is_Mult (C) then twiddle, if it's a \ | |
1548 | -- then frozzle and if it's another magic char then push C and | |
1549 | -- terminate the string. If none of the above, push C on the | |
1550 | -- string and go around again. | |
1551 | -- | |
1552 | -- Start_Pos is used to remember where "the current character" | |
1553 | -- starts in the string, if due to an Is_Mult we need to back | |
1554 | -- up and put the current char in a separate 1-character string. | |
1555 | -- When Start_Pos is 0, C is the only char in the string; | |
1556 | -- this is used in Is_Mult handling, and in setting the SIMPLE | |
1557 | -- flag at the end. | |
1558 | ||
1559 | procedure Parse_Literal | |
1560 | (Expr_Flags : in out Expression_Flags; | |
1561 | IP : out Pointer) | |
1562 | is | |
1563 | Start_Pos : Natural := 0; | |
1564 | C : Character; | |
1565 | Length_Ptr : Pointer; | |
598c3446 | 1566 | Has_Special_Operator : Boolean := False; |
38cbfe40 RK |
1567 | |
1568 | begin | |
1569 | Parse_Pos := Parse_Pos - 1; -- Look at current character | |
1570 | ||
1571 | if (Flags and Case_Insensitive) /= 0 then | |
1572 | IP := Emit_Node (EXACTF); | |
1573 | else | |
1574 | IP := Emit_Node (EXACT); | |
1575 | end if; | |
1576 | ||
1577 | Length_Ptr := Emit_Ptr; | |
1578 | Emit_Ptr := String_Operand (IP); | |
1579 | ||
1580 | Parse_Loop : | |
1581 | loop | |
1582 | ||
1583 | C := Expression (Parse_Pos); -- Get current character | |
1584 | ||
1585 | case C is | |
1586 | when '.' | '[' | '(' | ')' | '|' | ASCII.LF | '$' | '^' => | |
1587 | ||
1588 | if Start_Pos = 0 then | |
598c3446 | 1589 | Start_Pos := Parse_Pos; |
38cbfe40 RK |
1590 | Emit (C); -- First character is always emitted |
1591 | else | |
1592 | exit Parse_Loop; -- Else we are done | |
1593 | end if; | |
1594 | ||
1595 | when '?' | '+' | '*' | '{' => | |
1596 | ||
1597 | if Start_Pos = 0 then | |
598c3446 | 1598 | Start_Pos := Parse_Pos; |
38cbfe40 RK |
1599 | Emit (C); -- First character is always emitted |
1600 | ||
1601 | -- Are we looking at an operator, or is this | |
1602 | -- simply a normal character ? | |
1603 | elsif not Is_Mult (Parse_Pos) then | |
598c3446 GB |
1604 | Start_Pos := Parse_Pos; |
1605 | Case_Emit (C); | |
38cbfe40 RK |
1606 | else |
1607 | -- We've got something like "abc?d". Mark this as a | |
1608 | -- special case. What we want to emit is a first | |
1609 | -- constant string for "ab", then one for "c" that will | |
1610 | -- ultimately be transformed with a CURLY operator, A | |
1611 | -- special case has to be handled for "a?", since there | |
1612 | -- is no initial string to emit. | |
598c3446 | 1613 | Has_Special_Operator := True; |
38cbfe40 RK |
1614 | exit Parse_Loop; |
1615 | end if; | |
1616 | ||
1617 | when '\' => | |
598c3446 | 1618 | Start_Pos := Parse_Pos; |
38cbfe40 RK |
1619 | if Parse_Pos = Parse_End then |
1620 | Fail ("Trailing \"); | |
1621 | else | |
1622 | case Expression (Parse_Pos + 1) is | |
1623 | when 'b' | 'B' | 's' | 'S' | 'd' | 'D' | |
1624 | | 'w' | 'W' | '0' .. '9' | 'G' | 'A' | |
1625 | => exit Parse_Loop; | |
1626 | when 'n' => Emit (ASCII.LF); | |
1627 | when 't' => Emit (ASCII.HT); | |
1628 | when 'r' => Emit (ASCII.CR); | |
1629 | when 'f' => Emit (ASCII.FF); | |
1630 | when 'e' => Emit (ASCII.ESC); | |
1631 | when 'a' => Emit (ASCII.BEL); | |
1632 | when others => Emit (Expression (Parse_Pos + 1)); | |
1633 | end case; | |
1634 | Parse_Pos := Parse_Pos + 1; | |
1635 | end if; | |
1636 | ||
598c3446 GB |
1637 | when others => |
1638 | Start_Pos := Parse_Pos; | |
1639 | Case_Emit (C); | |
38cbfe40 RK |
1640 | end case; |
1641 | ||
1642 | exit Parse_Loop when Emit_Ptr - Length_Ptr = 254; | |
1643 | ||
38cbfe40 RK |
1644 | Parse_Pos := Parse_Pos + 1; |
1645 | ||
1646 | exit Parse_Loop when Parse_Pos > Parse_End; | |
1647 | end loop Parse_Loop; | |
1648 | ||
1649 | -- Is the string followed by a '*+?{' operator ? If yes, and if there | |
1650 | -- is an initial string to emit, do it now. | |
1651 | ||
598c3446 | 1652 | if Has_Special_Operator |
38cbfe40 RK |
1653 | and then Emit_Ptr >= Length_Ptr + 3 |
1654 | then | |
1655 | Emit_Ptr := Emit_Ptr - 1; | |
598c3446 | 1656 | Parse_Pos := Start_Pos; |
38cbfe40 RK |
1657 | end if; |
1658 | ||
1659 | if Emit_Code then | |
1660 | Program (Length_Ptr) := Character'Val (Emit_Ptr - Length_Ptr - 2); | |
1661 | end if; | |
1662 | ||
1663 | Expr_Flags.Has_Width := True; | |
1664 | ||
1665 | -- Slight optimization when there is a single character | |
1666 | ||
1667 | if Emit_Ptr = Length_Ptr + 2 then | |
1668 | Expr_Flags.Simple := True; | |
1669 | end if; | |
1670 | end Parse_Literal; | |
1671 | ||
1672 | ----------------- | |
1673 | -- Parse_Piece -- | |
1674 | ----------------- | |
1675 | ||
1676 | -- Note that the branching code sequences used for '?' and the | |
1677 | -- general cases of '*' and + are somewhat optimized: they use | |
1678 | -- the same NOTHING node as both the endmarker for their branch | |
1679 | -- list and the body of the last branch. It might seem that | |
1680 | -- this node could be dispensed with entirely, but the endmarker | |
1681 | -- role is not redundant. | |
1682 | ||
1683 | procedure Parse_Piece | |
1684 | (Expr_Flags : in out Expression_Flags; | |
1685 | IP : out Pointer) | |
1686 | is | |
1687 | Op : Character; | |
1688 | New_Flags : Expression_Flags; | |
1689 | Greedy : Boolean := True; | |
1690 | ||
1691 | begin | |
1692 | Parse_Atom (New_Flags, IP); | |
1693 | ||
1694 | if IP = 0 then | |
1695 | return; | |
1696 | end if; | |
1697 | ||
1698 | if Parse_Pos > Parse_End | |
1699 | or else not Is_Mult (Parse_Pos) | |
1700 | then | |
1701 | Expr_Flags := New_Flags; | |
1702 | return; | |
1703 | end if; | |
1704 | ||
1705 | Op := Expression (Parse_Pos); | |
1706 | ||
1707 | if Op /= '+' then | |
1708 | Expr_Flags := (SP_Start => True, others => False); | |
1709 | else | |
1710 | Expr_Flags := (Has_Width => True, others => False); | |
1711 | end if; | |
1712 | ||
1713 | -- Detect non greedy operators in the easy cases | |
1714 | ||
1715 | if Op /= '{' | |
1716 | and then Parse_Pos + 1 <= Parse_End | |
1717 | and then Expression (Parse_Pos + 1) = '?' | |
1718 | then | |
1719 | Greedy := False; | |
1720 | Parse_Pos := Parse_Pos + 1; | |
1721 | end if; | |
1722 | ||
1723 | -- Generate the byte code | |
1724 | ||
1725 | case Op is | |
1726 | when '*' => | |
1727 | ||
1728 | if New_Flags.Simple then | |
1729 | Insert_Operator (STAR, IP, Greedy); | |
1730 | else | |
1731 | Link_Tail (IP, Emit_Node (WHILEM)); | |
1732 | Insert_Curly_Operator | |
1733 | (CURLYX, 0, Max_Curly_Repeat, IP, Greedy); | |
1734 | Link_Tail (IP, Emit_Node (NOTHING)); | |
1735 | end if; | |
1736 | ||
1737 | when '+' => | |
1738 | ||
1739 | if New_Flags.Simple then | |
1740 | Insert_Operator (PLUS, IP, Greedy); | |
1741 | else | |
1742 | Link_Tail (IP, Emit_Node (WHILEM)); | |
1743 | Insert_Curly_Operator | |
1744 | (CURLYX, 1, Max_Curly_Repeat, IP, Greedy); | |
1745 | Link_Tail (IP, Emit_Node (NOTHING)); | |
1746 | end if; | |
1747 | ||
1748 | when '?' => | |
1749 | if New_Flags.Simple then | |
1750 | Insert_Curly_Operator (CURLY, 0, 1, IP, Greedy); | |
1751 | else | |
1752 | Link_Tail (IP, Emit_Node (WHILEM)); | |
1753 | Insert_Curly_Operator (CURLYX, 0, 1, IP, Greedy); | |
1754 | Link_Tail (IP, Emit_Node (NOTHING)); | |
1755 | end if; | |
1756 | ||
1757 | when '{' => | |
1758 | declare | |
1759 | Min, Max : Natural; | |
1760 | ||
1761 | begin | |
1762 | Get_Curly_Arguments (Parse_Pos, Min, Max, Greedy); | |
1763 | ||
1764 | if New_Flags.Simple then | |
1765 | Insert_Curly_Operator (CURLY, Min, Max, IP, Greedy); | |
1766 | else | |
1767 | Link_Tail (IP, Emit_Node (WHILEM)); | |
1768 | Insert_Curly_Operator (CURLYX, Min, Max, IP, Greedy); | |
1769 | Link_Tail (IP, Emit_Node (NOTHING)); | |
1770 | end if; | |
1771 | end; | |
1772 | ||
1773 | when others => | |
1774 | null; | |
1775 | end case; | |
1776 | ||
1777 | Parse_Pos := Parse_Pos + 1; | |
1778 | ||
1779 | if Parse_Pos <= Parse_End | |
1780 | and then Is_Mult (Parse_Pos) | |
1781 | then | |
1782 | Fail ("nested *+{"); | |
1783 | end if; | |
1784 | end Parse_Piece; | |
1785 | ||
1786 | --------------------------------- | |
1787 | -- Parse_Posix_Character_Class -- | |
1788 | --------------------------------- | |
1789 | ||
1790 | function Parse_Posix_Character_Class return Std_Class is | |
1791 | Invert : Boolean := False; | |
1792 | Class : Std_Class := ANYOF_NONE; | |
1793 | E : String renames Expression; | |
1794 | ||
1795 | begin | |
1796 | if Parse_Pos <= Parse_End | |
1797 | and then Expression (Parse_Pos) = ':' | |
1798 | then | |
1799 | Parse_Pos := Parse_Pos + 1; | |
1800 | ||
1801 | -- Do we have something like: [[:^alpha:]] | |
1802 | ||
1803 | if Parse_Pos <= Parse_End | |
1804 | and then Expression (Parse_Pos) = '^' | |
1805 | then | |
1806 | Invert := True; | |
1807 | Parse_Pos := Parse_Pos + 1; | |
1808 | end if; | |
1809 | ||
1810 | -- All classes have 6 characters at least | |
1811 | -- ??? magid constant 6 should have a name! | |
1812 | ||
1813 | if Parse_Pos + 6 <= Parse_End then | |
1814 | ||
1815 | case Expression (Parse_Pos) is | |
1816 | when 'a' => | |
1817 | if E (Parse_Pos .. Parse_Pos + 4) = "alnum:]" then | |
1818 | if Invert then | |
1819 | Class := ANYOF_NALNUMC; | |
1820 | else | |
1821 | Class := ANYOF_ALNUMC; | |
1822 | end if; | |
1823 | ||
1824 | elsif E (Parse_Pos .. Parse_Pos + 6) = "alpha:]" then | |
1825 | if Invert then | |
1826 | Class := ANYOF_NALPHA; | |
1827 | else | |
1828 | Class := ANYOF_ALPHA; | |
1829 | end if; | |
1830 | ||
1831 | elsif E (Parse_Pos .. Parse_Pos + 6) = "ascii:]" then | |
1832 | if Invert then | |
1833 | Class := ANYOF_NASCII; | |
1834 | else | |
1835 | Class := ANYOF_ASCII; | |
1836 | end if; | |
1837 | ||
1838 | end if; | |
1839 | ||
1840 | when 'c' => | |
1841 | if E (Parse_Pos .. Parse_Pos + 6) = "cntrl:]" then | |
1842 | if Invert then | |
1843 | Class := ANYOF_NCNTRL; | |
1844 | else | |
1845 | Class := ANYOF_CNTRL; | |
1846 | end if; | |
1847 | end if; | |
1848 | ||
1849 | when 'd' => | |
1850 | ||
1851 | if E (Parse_Pos .. Parse_Pos + 6) = "digit:]" then | |
1852 | if Invert then | |
1853 | Class := ANYOF_NDIGIT; | |
1854 | else | |
1855 | Class := ANYOF_DIGIT; | |
1856 | end if; | |
1857 | end if; | |
1858 | ||
1859 | when 'g' => | |
1860 | ||
1861 | if E (Parse_Pos .. Parse_Pos + 6) = "graph:]" then | |
1862 | if Invert then | |
1863 | Class := ANYOF_NGRAPH; | |
1864 | else | |
1865 | Class := ANYOF_GRAPH; | |
1866 | end if; | |
1867 | end if; | |
1868 | ||
1869 | when 'l' => | |
1870 | ||
1871 | if E (Parse_Pos .. Parse_Pos + 6) = "lower:]" then | |
1872 | if Invert then | |
1873 | Class := ANYOF_NLOWER; | |
1874 | else | |
1875 | Class := ANYOF_LOWER; | |
1876 | end if; | |
1877 | end if; | |
1878 | ||
1879 | when 'p' => | |
1880 | ||
1881 | if E (Parse_Pos .. Parse_Pos + 6) = "print:]" then | |
1882 | if Invert then | |
1883 | Class := ANYOF_NPRINT; | |
1884 | else | |
1885 | Class := ANYOF_PRINT; | |
1886 | end if; | |
1887 | ||
1888 | elsif E (Parse_Pos .. Parse_Pos + 6) = "punct:]" then | |
1889 | if Invert then | |
1890 | Class := ANYOF_NPUNCT; | |
1891 | else | |
1892 | Class := ANYOF_PUNCT; | |
1893 | end if; | |
1894 | end if; | |
1895 | ||
1896 | when 's' => | |
1897 | ||
1898 | if E (Parse_Pos .. Parse_Pos + 6) = "space:]" then | |
1899 | if Invert then | |
1900 | Class := ANYOF_NSPACE; | |
1901 | else | |
1902 | Class := ANYOF_SPACE; | |
1903 | end if; | |
1904 | end if; | |
1905 | ||
1906 | when 'u' => | |
1907 | ||
1908 | if E (Parse_Pos .. Parse_Pos + 6) = "upper:]" then | |
1909 | if Invert then | |
1910 | Class := ANYOF_NUPPER; | |
1911 | else | |
1912 | Class := ANYOF_UPPER; | |
1913 | end if; | |
1914 | end if; | |
1915 | ||
1916 | when 'w' => | |
1917 | ||
1918 | if E (Parse_Pos .. Parse_Pos + 5) = "word:]" then | |
1919 | if Invert then | |
1920 | Class := ANYOF_NALNUM; | |
1921 | else | |
1922 | Class := ANYOF_ALNUM; | |
1923 | end if; | |
1924 | ||
1925 | Parse_Pos := Parse_Pos - 1; | |
1926 | end if; | |
1927 | ||
1928 | when 'x' => | |
1929 | ||
1930 | if Parse_Pos + 7 <= Parse_End | |
1931 | and then E (Parse_Pos .. Parse_Pos + 7) = "xdigit:]" | |
1932 | then | |
1933 | if Invert then | |
1934 | Class := ANYOF_NXDIGIT; | |
1935 | else | |
1936 | Class := ANYOF_XDIGIT; | |
1937 | end if; | |
1938 | ||
1939 | Parse_Pos := Parse_Pos + 1; | |
1940 | end if; | |
1941 | ||
1942 | when others => | |
1943 | Class := ANYOF_NONE; | |
1944 | ||
1945 | end case; | |
1946 | ||
1947 | if Class /= ANYOF_NONE then | |
1948 | Parse_Pos := Parse_Pos + 7; | |
1949 | end if; | |
1950 | ||
1951 | else | |
1952 | Fail ("Invalid character class"); | |
1953 | end if; | |
1954 | ||
1955 | else | |
1956 | return ANYOF_NONE; | |
1957 | end if; | |
1958 | ||
1959 | return Class; | |
1960 | end Parse_Posix_Character_Class; | |
1961 | ||
1962 | Expr_Flags : Expression_Flags; | |
1963 | Result : Pointer; | |
1964 | ||
1965 | -- Start of processing for Compile | |
1966 | ||
1967 | begin | |
1968 | Emit (MAGIC); | |
1969 | Parse (False, Expr_Flags, Result); | |
1970 | ||
1971 | if Result = 0 then | |
1972 | Fail ("Couldn't compile expression"); | |
1973 | end if; | |
1974 | ||
1975 | Final_Code_Size := Emit_Ptr - 1; | |
1976 | ||
1977 | -- Do we want to actually compile the expression, or simply get the | |
1978 | -- code size ??? | |
1979 | ||
1980 | if Emit_Code then | |
1981 | Optimize (PM); | |
1982 | end if; | |
1983 | ||
1984 | PM.Flags := Flags; | |
1985 | end Compile; | |
1986 | ||
1987 | function Compile | |
1988 | (Expression : String; | |
1989 | Flags : Regexp_Flags := No_Flags) | |
1990 | return Pattern_Matcher | |
1991 | is | |
1992 | Size : Program_Size; | |
1993 | Dummy : Pattern_Matcher (0); | |
1994 | ||
1995 | begin | |
1996 | Compile (Dummy, Expression, Size, Flags); | |
1997 | ||
1998 | declare | |
1999 | Result : Pattern_Matcher (Size); | |
2000 | begin | |
2001 | Compile (Result, Expression, Size, Flags); | |
2002 | return Result; | |
2003 | end; | |
2004 | end Compile; | |
2005 | ||
2006 | procedure Compile | |
2007 | (Matcher : out Pattern_Matcher; | |
2008 | Expression : String; | |
2009 | Flags : Regexp_Flags := No_Flags) | |
2010 | is | |
2011 | Size : Program_Size; | |
2012 | ||
2013 | begin | |
2014 | Compile (Matcher, Expression, Size, Flags); | |
2015 | end Compile; | |
2016 | ||
2017 | ---------- | |
2018 | -- Dump -- | |
2019 | ---------- | |
2020 | ||
2021 | procedure Dump (Self : Pattern_Matcher) is | |
2022 | ||
2023 | -- Index : Pointer := Program_First + 1; | |
2024 | -- What is the above line for ??? | |
2025 | ||
2026 | Op : Opcode; | |
2027 | Program : Program_Data renames Self.Program; | |
2028 | ||
2029 | procedure Dump_Until | |
2030 | (Start : Pointer; | |
2031 | Till : Pointer; | |
2032 | Indent : Natural := 0); | |
2033 | -- Dump the program until the node Till (not included) is met. | |
2034 | -- Every line is indented with Index spaces at the beginning | |
2035 | -- Dumps till the end if Till is 0. | |
2036 | ||
2037 | ---------------- | |
2038 | -- Dump_Until -- | |
2039 | ---------------- | |
2040 | ||
2041 | procedure Dump_Until | |
2042 | (Start : Pointer; | |
2043 | Till : Pointer; | |
2044 | Indent : Natural := 0) | |
2045 | is | |
2046 | Next : Pointer; | |
2047 | Index : Pointer := Start; | |
2048 | Local_Indent : Natural := Indent; | |
2049 | Length : Pointer; | |
2050 | ||
2051 | begin | |
2052 | while Index < Till loop | |
2053 | ||
2054 | Op := Opcode'Val (Character'Pos ((Self.Program (Index)))); | |
2055 | ||
2056 | if Op = CLOSE then | |
2057 | Local_Indent := Local_Indent - 3; | |
2058 | end if; | |
2059 | ||
2060 | declare | |
2061 | Point : String := Pointer'Image (Index); | |
2062 | ||
2063 | begin | |
2064 | for J in 1 .. 6 - Point'Length loop | |
2065 | Put (' '); | |
2066 | end loop; | |
2067 | ||
2068 | Put (Point | |
2069 | & " : " | |
2070 | & (1 .. Local_Indent => ' ') | |
2071 | & Opcode'Image (Op)); | |
2072 | end; | |
2073 | ||
2074 | -- Print the parenthesis number | |
2075 | ||
2076 | if Op = OPEN or else Op = CLOSE or else Op = REFF then | |
2077 | Put (Natural'Image (Character'Pos (Program (Index + 3)))); | |
2078 | end if; | |
2079 | ||
2080 | Next := Index + Get_Next_Offset (Program, Index); | |
2081 | ||
2082 | if Next = Index then | |
2083 | Put (" (next at 0)"); | |
2084 | else | |
2085 | Put (" (next at " & Pointer'Image (Next) & ")"); | |
2086 | end if; | |
2087 | ||
2088 | case Op is | |
2089 | ||
2090 | -- Character class operand | |
2091 | ||
2092 | when ANYOF => null; | |
2093 | declare | |
2094 | Bitmap : Character_Class; | |
2095 | Last : Character := ASCII.Nul; | |
2096 | Current : Natural := 0; | |
2097 | ||
2098 | Current_Char : Character; | |
2099 | ||
2100 | begin | |
2101 | Bitmap_Operand (Program, Index, Bitmap); | |
2102 | Put (" operand="); | |
2103 | ||
2104 | while Current <= 255 loop | |
2105 | Current_Char := Character'Val (Current); | |
2106 | ||
2107 | -- First item in a range | |
2108 | ||
2109 | if Get_From_Class (Bitmap, Current_Char) then | |
2110 | Last := Current_Char; | |
2111 | ||
2112 | -- Search for the last item in the range | |
2113 | ||
2114 | loop | |
2115 | Current := Current + 1; | |
2116 | exit when Current > 255; | |
2117 | Current_Char := Character'Val (Current); | |
2118 | exit when | |
2119 | not Get_From_Class (Bitmap, Current_Char); | |
2120 | ||
2121 | end loop; | |
2122 | ||
2123 | if Last <= ' ' then | |
2124 | Put (Last'Img); | |
2125 | else | |
2126 | Put (Last); | |
2127 | end if; | |
2128 | ||
2129 | if Character'Succ (Last) /= Current_Char then | |
2130 | Put ("-" & Character'Pred (Current_Char)); | |
2131 | end if; | |
2132 | ||
2133 | else | |
2134 | Current := Current + 1; | |
2135 | end if; | |
2136 | end loop; | |
2137 | ||
2138 | New_Line; | |
2139 | Index := Index + 3 + Bitmap'Length; | |
2140 | end; | |
2141 | ||
2142 | -- string operand | |
2143 | ||
2144 | when EXACT | EXACTF => | |
2145 | Length := String_Length (Program, Index); | |
2146 | Put (" operand (length:" & Program_Size'Image (Length + 1) | |
2147 | & ") =" | |
2148 | & String (Program (String_Operand (Index) | |
2149 | .. String_Operand (Index) | |
2150 | + Length))); | |
2151 | Index := String_Operand (Index) + Length + 1; | |
2152 | New_Line; | |
2153 | ||
2154 | -- Node operand | |
2155 | ||
2156 | when BRANCH => | |
2157 | New_Line; | |
2158 | Dump_Until (Index + 3, Next, Local_Indent + 3); | |
2159 | Index := Next; | |
2160 | ||
2161 | when STAR | PLUS => | |
2162 | New_Line; | |
2163 | ||
2164 | -- Only one instruction | |
2165 | ||
2166 | Dump_Until (Index + 3, Index + 4, Local_Indent + 3); | |
2167 | Index := Next; | |
2168 | ||
2169 | when CURLY | CURLYX => | |
2170 | Put (" {" | |
2171 | & Natural'Image (Read_Natural (Program, Index + 3)) | |
2172 | & "," | |
2173 | & Natural'Image (Read_Natural (Program, Index + 5)) | |
2174 | & "}"); | |
2175 | New_Line; | |
2176 | Dump_Until (Index + 7, Next, Local_Indent + 3); | |
2177 | Index := Next; | |
2178 | ||
2179 | when OPEN => | |
2180 | New_Line; | |
2181 | Index := Index + 4; | |
2182 | Local_Indent := Local_Indent + 3; | |
2183 | ||
2184 | when CLOSE | REFF => | |
2185 | New_Line; | |
2186 | Index := Index + 4; | |
2187 | ||
2188 | when EOP => | |
2189 | Index := Index + 3; | |
2190 | New_Line; | |
2191 | exit; | |
2192 | ||
2193 | -- No operand | |
2194 | ||
2195 | when others => | |
2196 | Index := Index + 3; | |
2197 | New_Line; | |
2198 | end case; | |
2199 | end loop; | |
2200 | end Dump_Until; | |
2201 | ||
2202 | -- Start of processing for Dump | |
2203 | ||
2204 | begin | |
2205 | pragma Assert (Self.Program (Program_First) = MAGIC, | |
2206 | "Corrupted Pattern_Matcher"); | |
2207 | ||
2208 | Put_Line ("Must start with (Self.First) = " | |
2209 | & Character'Image (Self.First)); | |
2210 | ||
2211 | if (Self.Flags and Case_Insensitive) /= 0 then | |
2212 | Put_Line (" Case_Insensitive mode"); | |
2213 | end if; | |
2214 | ||
2215 | if (Self.Flags and Single_Line) /= 0 then | |
2216 | Put_Line (" Single_Line mode"); | |
2217 | end if; | |
2218 | ||
2219 | if (Self.Flags and Multiple_Lines) /= 0 then | |
2220 | Put_Line (" Multiple_Lines mode"); | |
2221 | end if; | |
2222 | ||
2223 | Put_Line (" 1 : MAGIC"); | |
2224 | Dump_Until (Program_First + 1, Self.Program'Last + 1); | |
2225 | end Dump; | |
2226 | ||
2227 | -------------------- | |
2228 | -- Get_From_Class -- | |
2229 | -------------------- | |
2230 | ||
2231 | function Get_From_Class | |
2232 | (Bitmap : Character_Class; | |
2233 | C : Character) | |
2234 | return Boolean | |
2235 | is | |
2236 | Value : constant Class_Byte := Character'Pos (C); | |
2237 | ||
2238 | begin | |
2239 | return (Bitmap (Value / 8) | |
2240 | and Bit_Conversion (Value mod 8)) /= 0; | |
2241 | end Get_From_Class; | |
2242 | ||
2243 | -------------- | |
2244 | -- Get_Next -- | |
2245 | -------------- | |
2246 | ||
2247 | function Get_Next (Program : Program_Data; IP : Pointer) return Pointer is | |
2248 | Offset : constant Pointer := Get_Next_Offset (Program, IP); | |
2249 | ||
2250 | begin | |
2251 | if Offset = 0 then | |
2252 | return 0; | |
2253 | else | |
2254 | return IP + Offset; | |
2255 | end if; | |
2256 | end Get_Next; | |
2257 | ||
2258 | --------------------- | |
2259 | -- Get_Next_Offset -- | |
2260 | --------------------- | |
2261 | ||
2262 | function Get_Next_Offset | |
2263 | (Program : Program_Data; | |
2264 | IP : Pointer) | |
2265 | return Pointer | |
2266 | is | |
2267 | begin | |
2268 | return Pointer (Read_Natural (Program, IP + 1)); | |
2269 | end Get_Next_Offset; | |
2270 | ||
2271 | -------------- | |
2272 | -- Is_Alnum -- | |
2273 | -------------- | |
2274 | ||
2275 | function Is_Alnum (C : Character) return Boolean is | |
2276 | begin | |
2277 | return Is_Alphanumeric (C) or else C = '_'; | |
2278 | end Is_Alnum; | |
2279 | ||
2280 | ------------------ | |
2281 | -- Is_Printable -- | |
2282 | ------------------ | |
2283 | ||
2284 | function Is_Printable (C : Character) return Boolean is | |
2285 | Value : constant Natural := Character'Pos (C); | |
2286 | ||
2287 | begin | |
2288 | return (Value > 32 and then Value < 127) | |
2289 | or else Is_Space (C); | |
2290 | end Is_Printable; | |
2291 | ||
2292 | -------------- | |
2293 | -- Is_Space -- | |
2294 | -------------- | |
2295 | ||
2296 | function Is_Space (C : Character) return Boolean is | |
2297 | begin | |
2298 | return C = ' ' | |
2299 | or else C = ASCII.HT | |
2300 | or else C = ASCII.CR | |
2301 | or else C = ASCII.LF | |
2302 | or else C = ASCII.VT | |
2303 | or else C = ASCII.FF; | |
2304 | end Is_Space; | |
2305 | ||
2306 | ----------- | |
2307 | -- Match -- | |
2308 | ----------- | |
2309 | ||
2310 | procedure Match | |
2311 | (Self : Pattern_Matcher; | |
2312 | Data : String; | |
2313 | Matches : out Match_Array) | |
2314 | is | |
2315 | Program : Program_Data renames Self.Program; -- Shorter notation | |
2316 | ||
2317 | -- Global work variables | |
2318 | ||
2319 | Input_Pos : Natural; -- String-input pointer | |
2320 | BOL_Pos : Natural; -- Beginning of input, for ^ check | |
2321 | Matched : Boolean := False; -- Until proven True | |
2322 | ||
2323 | Matches_Full : Match_Array (0 .. Natural'Max (Self.Paren_Count, | |
2324 | Matches'Last)); | |
2325 | -- Stores the value of all the parenthesis pairs. | |
2326 | -- We do not use directly Matches, so that we can also use back | |
2327 | -- references (REFF) even if Matches is too small. | |
2328 | ||
2329 | type Natural_Array is array (Match_Count range <>) of Natural; | |
2330 | Matches_Tmp : Natural_Array (Matches_Full'Range); | |
2331 | -- Save the opening position of parenthesis. | |
2332 | ||
2333 | Last_Paren : Natural := 0; | |
2334 | -- Last parenthesis seen | |
2335 | ||
2336 | Greedy : Boolean := True; | |
2337 | -- True if the next operator should be greedy | |
2338 | ||
2339 | type Current_Curly_Record; | |
2340 | type Current_Curly_Access is access all Current_Curly_Record; | |
2341 | type Current_Curly_Record is record | |
2342 | Paren_Floor : Natural; -- How far back to strip parenthesis data | |
2343 | Cur : Integer; -- How many instances of scan we've matched | |
2344 | Min : Natural; -- Minimal number of scans to match | |
2345 | Max : Natural; -- Maximal number of scans to match | |
2346 | Greedy : Boolean; -- Whether to work our way up or down | |
2347 | Scan : Pointer; -- The thing to match | |
2348 | Next : Pointer; -- What has to match after it | |
2349 | Lastloc : Natural; -- Where we started matching this scan | |
2350 | Old_Cc : Current_Curly_Access; -- Before we started this one | |
2351 | end record; | |
2352 | -- Data used to handle the curly operator and the plus and star | |
2353 | -- operators for complex expressions. | |
2354 | ||
2355 | Current_Curly : Current_Curly_Access := null; | |
2356 | -- The curly currently being processed. | |
2357 | ||
2358 | ----------------------- | |
2359 | -- Local Subprograms -- | |
2360 | ----------------------- | |
2361 | ||
2362 | function Index (Start : Positive; C : Character) return Natural; | |
2363 | -- Find character C in Data starting at Start and return position | |
2364 | ||
2365 | function Repeat | |
2366 | (IP : Pointer; | |
2367 | Max : Natural := Natural'Last) | |
2368 | return Natural; | |
2369 | -- Repeatedly match something simple, report how many | |
2370 | -- It only matches on things of length 1. | |
2371 | -- Starting from Input_Pos, it matches at most Max CURLY. | |
2372 | ||
2373 | function Try (Pos : in Positive) return Boolean; | |
2374 | -- Try to match at specific point | |
2375 | ||
2376 | function Match (IP : Pointer) return Boolean; | |
2377 | -- This is the main matching routine. Conceptually the strategy | |
2378 | -- is simple: check to see whether the current node matches, | |
2379 | -- call self recursively to see whether the rest matches, | |
2380 | -- and then act accordingly. | |
2381 | -- | |
2382 | -- In practice Match makes some effort to avoid recursion, in | |
2383 | -- particular by going through "ordinary" nodes (that don't | |
2384 | -- need to know whether the rest of the match failed) by | |
2385 | -- using a loop instead of recursion. | |
2386 | ||
2387 | function Match_Whilem (IP : Pointer) return Boolean; | |
2388 | -- Return True if a WHILEM matches | |
2389 | ||
2390 | function Recurse_Match (IP : Pointer; From : Natural) return Boolean; | |
2391 | pragma Inline (Recurse_Match); | |
2392 | -- Calls Match recursively. It saves and restores the parenthesis | |
2393 | -- status and location in the input stream correctly, so that | |
2394 | -- backtracking is possible | |
2395 | ||
2396 | function Match_Simple_Operator | |
2397 | (Op : Opcode; | |
2398 | Scan : Pointer; | |
2399 | Next : Pointer; | |
2400 | Greedy : Boolean) | |
2401 | return Boolean; | |
2402 | -- Return True it the simple operator (possibly non-greedy) matches | |
2403 | ||
2404 | pragma Inline_Always (Index); | |
2405 | pragma Inline_Always (Repeat); | |
2406 | ||
2407 | -- These are two complex functions, but used only once. | |
2408 | pragma Inline_Always (Match_Whilem); | |
2409 | pragma Inline_Always (Match_Simple_Operator); | |
2410 | ||
2411 | ----------- | |
2412 | -- Index -- | |
2413 | ----------- | |
2414 | ||
2415 | function Index | |
2416 | (Start : Positive; | |
2417 | C : Character) | |
2418 | return Natural | |
2419 | is | |
2420 | begin | |
2421 | for J in Start .. Data'Last loop | |
2422 | if Data (J) = C then | |
2423 | return J; | |
2424 | end if; | |
2425 | end loop; | |
2426 | ||
2427 | return 0; | |
2428 | end Index; | |
2429 | ||
2430 | ------------------- | |
2431 | -- Recurse_Match -- | |
2432 | ------------------- | |
2433 | ||
2434 | function Recurse_Match (IP : Pointer; From : Natural) return Boolean is | |
2435 | L : constant Natural := Last_Paren; | |
2436 | Tmp_F : constant Match_Array := | |
2437 | Matches_Full (From + 1 .. Matches_Full'Last); | |
2438 | Start : constant Natural_Array := | |
2439 | Matches_Tmp (From + 1 .. Matches_Tmp'Last); | |
2440 | Input : constant Natural := Input_Pos; | |
2441 | begin | |
2442 | if Match (IP) then | |
2443 | return True; | |
2444 | end if; | |
2445 | Last_Paren := L; | |
2446 | Matches_Full (Tmp_F'Range) := Tmp_F; | |
2447 | Matches_Tmp (Start'Range) := Start; | |
2448 | Input_Pos := Input; | |
2449 | return False; | |
2450 | end Recurse_Match; | |
2451 | ||
2452 | ----------- | |
2453 | -- Match -- | |
2454 | ----------- | |
2455 | ||
2456 | function Match (IP : Pointer) return Boolean is | |
2457 | Scan : Pointer := IP; | |
2458 | Next : Pointer; | |
2459 | Op : Opcode; | |
2460 | ||
2461 | begin | |
2462 | State_Machine : | |
2463 | loop | |
2464 | pragma Assert (Scan /= 0); | |
2465 | ||
2466 | -- Determine current opcode and count its usage in debug mode | |
2467 | ||
2468 | Op := Opcode'Val (Character'Pos (Program (Scan))); | |
2469 | ||
2470 | -- Calculate offset of next instruction. | |
2471 | -- Second character is most significant in Program_Data. | |
2472 | ||
2473 | Next := Get_Next (Program, Scan); | |
2474 | ||
2475 | case Op is | |
2476 | when EOP => | |
2477 | return True; -- Success ! | |
2478 | ||
2479 | when BRANCH => | |
2480 | if Program (Next) /= BRANCH then | |
2481 | Next := Operand (Scan); -- No choice, avoid recursion | |
2482 | ||
2483 | else | |
2484 | loop | |
2485 | if Recurse_Match (Operand (Scan), 0) then | |
2486 | return True; | |
2487 | end if; | |
2488 | ||
2489 | Scan := Get_Next (Program, Scan); | |
2490 | exit when Scan = 0 or Program (Scan) /= BRANCH; | |
2491 | end loop; | |
2492 | ||
2493 | exit State_Machine; | |
2494 | end if; | |
2495 | ||
2496 | when NOTHING => | |
2497 | null; | |
2498 | ||
2499 | when BOL => | |
2500 | exit State_Machine when | |
2501 | Input_Pos /= BOL_Pos | |
2502 | and then ((Self.Flags and Multiple_Lines) = 0 | |
2503 | or else Data (Input_Pos - 1) /= ASCII.LF); | |
2504 | ||
2505 | when MBOL => | |
2506 | exit State_Machine when | |
2507 | Input_Pos /= BOL_Pos | |
2508 | and then Data (Input_Pos - 1) /= ASCII.LF; | |
2509 | ||
2510 | when SBOL => | |
2511 | exit State_Machine when Input_Pos /= BOL_Pos; | |
2512 | ||
2513 | when EOL => | |
2514 | exit State_Machine when | |
2515 | Input_Pos <= Data'Last | |
2516 | and then ((Self.Flags and Multiple_Lines) = 0 | |
2517 | or else Data (Input_Pos) /= ASCII.LF); | |
2518 | ||
2519 | when MEOL => | |
2520 | exit State_Machine when | |
2521 | Input_Pos <= Data'Last | |
2522 | and then Data (Input_Pos) /= ASCII.LF; | |
2523 | ||
2524 | when SEOL => | |
2525 | exit State_Machine when Input_Pos <= Data'Last; | |
2526 | ||
2527 | when BOUND | NBOUND => | |
2528 | ||
2529 | -- Was last char in word ? | |
2530 | ||
2531 | declare | |
2532 | N : Boolean := False; | |
2533 | Ln : Boolean := False; | |
2534 | ||
2535 | begin | |
2536 | if Input_Pos /= Data'First then | |
2537 | N := Is_Alnum (Data (Input_Pos - 1)); | |
2538 | end if; | |
2539 | ||
2540 | if Input_Pos > Data'Last then | |
2541 | Ln := False; | |
2542 | else | |
2543 | Ln := Is_Alnum (Data (Input_Pos)); | |
2544 | end if; | |
2545 | ||
2546 | if Op = BOUND then | |
2547 | if N = Ln then | |
2548 | exit State_Machine; | |
2549 | end if; | |
2550 | else | |
2551 | if N /= Ln then | |
2552 | exit State_Machine; | |
2553 | end if; | |
2554 | end if; | |
2555 | end; | |
2556 | ||
2557 | when SPACE => | |
2558 | exit State_Machine when | |
2559 | Input_Pos > Data'Last | |
2560 | or else not Is_Space (Data (Input_Pos)); | |
2561 | Input_Pos := Input_Pos + 1; | |
2562 | ||
2563 | when NSPACE => | |
2564 | exit State_Machine when | |
2565 | Input_Pos > Data'Last | |
2566 | or else Is_Space (Data (Input_Pos)); | |
2567 | Input_Pos := Input_Pos + 1; | |
2568 | ||
2569 | when DIGIT => | |
2570 | exit State_Machine when | |
2571 | Input_Pos > Data'Last | |
2572 | or else not Is_Digit (Data (Input_Pos)); | |
2573 | Input_Pos := Input_Pos + 1; | |
2574 | ||
2575 | when NDIGIT => | |
2576 | exit State_Machine when | |
2577 | Input_Pos > Data'Last | |
2578 | or else Is_Digit (Data (Input_Pos)); | |
2579 | Input_Pos := Input_Pos + 1; | |
2580 | ||
2581 | when ALNUM => | |
2582 | exit State_Machine when | |
2583 | Input_Pos > Data'Last | |
2584 | or else not Is_Alnum (Data (Input_Pos)); | |
2585 | Input_Pos := Input_Pos + 1; | |
2586 | ||
2587 | when NALNUM => | |
2588 | exit State_Machine when | |
2589 | Input_Pos > Data'Last | |
2590 | or else Is_Alnum (Data (Input_Pos)); | |
2591 | Input_Pos := Input_Pos + 1; | |
2592 | ||
2593 | when ANY => | |
2594 | exit State_Machine when Input_Pos > Data'Last | |
2595 | or else Data (Input_Pos) = ASCII.LF; | |
2596 | Input_Pos := Input_Pos + 1; | |
2597 | ||
2598 | when SANY => | |
2599 | exit State_Machine when Input_Pos > Data'Last; | |
2600 | Input_Pos := Input_Pos + 1; | |
2601 | ||
2602 | when EXACT => | |
2603 | declare | |
2604 | Opnd : Pointer := String_Operand (Scan); | |
2605 | Current : Positive := Input_Pos; | |
2606 | Last : constant Pointer := | |
2607 | Opnd + String_Length (Program, Scan); | |
2608 | ||
2609 | begin | |
2610 | while Opnd <= Last loop | |
2611 | exit State_Machine when Current > Data'Last | |
2612 | or else Program (Opnd) /= Data (Current); | |
2613 | Current := Current + 1; | |
2614 | Opnd := Opnd + 1; | |
2615 | end loop; | |
2616 | ||
2617 | Input_Pos := Current; | |
2618 | end; | |
2619 | ||
2620 | when EXACTF => | |
2621 | declare | |
2622 | Opnd : Pointer := String_Operand (Scan); | |
2623 | Current : Positive := Input_Pos; | |
2624 | Last : constant Pointer := | |
2625 | Opnd + String_Length (Program, Scan); | |
2626 | ||
2627 | begin | |
2628 | while Opnd <= Last loop | |
2629 | exit State_Machine when Current > Data'Last | |
2630 | or else Program (Opnd) /= To_Lower (Data (Current)); | |
2631 | Current := Current + 1; | |
2632 | Opnd := Opnd + 1; | |
2633 | end loop; | |
2634 | ||
2635 | Input_Pos := Current; | |
2636 | end; | |
2637 | ||
2638 | when ANYOF => | |
2639 | declare | |
2640 | Bitmap : Character_Class; | |
2641 | ||
2642 | begin | |
2643 | Bitmap_Operand (Program, Scan, Bitmap); | |
2644 | exit State_Machine when | |
2645 | Input_Pos > Data'Last | |
2646 | or else not Get_From_Class (Bitmap, Data (Input_Pos)); | |
2647 | Input_Pos := Input_Pos + 1; | |
2648 | end; | |
2649 | ||
2650 | when OPEN => | |
2651 | declare | |
2652 | No : constant Natural := | |
2653 | Character'Pos (Program (Operand (Scan))); | |
2654 | begin | |
2655 | Matches_Tmp (No) := Input_Pos; | |
2656 | end; | |
2657 | ||
2658 | when CLOSE => | |
2659 | declare | |
2660 | No : constant Natural := | |
2661 | Character'Pos (Program (Operand (Scan))); | |
2662 | begin | |
2663 | Matches_Full (No) := (Matches_Tmp (No), Input_Pos - 1); | |
2664 | if Last_Paren < No then | |
2665 | Last_Paren := No; | |
2666 | end if; | |
2667 | end; | |
2668 | ||
2669 | when REFF => | |
2670 | declare | |
2671 | No : constant Natural := | |
2672 | Character'Pos (Program (Operand (Scan))); | |
2673 | Data_Pos : Natural; | |
2674 | ||
2675 | begin | |
2676 | -- If we haven't seen that parenthesis yet | |
2677 | ||
2678 | if Last_Paren < No then | |
2679 | return False; | |
2680 | end if; | |
2681 | ||
2682 | Data_Pos := Matches_Full (No).First; | |
2683 | while Data_Pos <= Matches_Full (No).Last loop | |
2684 | if Input_Pos > Data'Last | |
2685 | or else Data (Input_Pos) /= Data (Data_Pos) | |
2686 | then | |
2687 | return False; | |
2688 | end if; | |
2689 | ||
2690 | Input_Pos := Input_Pos + 1; | |
2691 | Data_Pos := Data_Pos + 1; | |
2692 | end loop; | |
2693 | end; | |
2694 | ||
2695 | when MINMOD => | |
2696 | Greedy := False; | |
2697 | ||
2698 | when STAR | PLUS | CURLY => | |
2699 | declare | |
2700 | Greed : constant Boolean := Greedy; | |
2701 | begin | |
2702 | Greedy := True; | |
2703 | return Match_Simple_Operator (Op, Scan, Next, Greed); | |
2704 | end; | |
2705 | ||
2706 | when CURLYX => | |
2707 | ||
2708 | -- Looking at something like: | |
2709 | -- 1: CURLYX {n,m} (->4) | |
2710 | -- 2: code for complex thing (->3) | |
2711 | -- 3: WHILEM (->0) | |
2712 | -- 4: NOTHING | |
2713 | ||
2714 | declare | |
2715 | Cc : aliased Current_Curly_Record; | |
2716 | Min : Natural := Read_Natural (Program, Scan + 3); | |
2717 | Max : Natural := Read_Natural (Program, Scan + 5); | |
2718 | ||
2719 | Has_Match : Boolean; | |
2720 | ||
2721 | begin | |
2722 | Cc := (Paren_Floor => Last_Paren, | |
2723 | Cur => -1, | |
2724 | Min => Min, | |
2725 | Max => Max, | |
2726 | Greedy => Greedy, | |
2727 | Scan => Scan + 7, | |
2728 | Next => Next, | |
2729 | Lastloc => 0, | |
2730 | Old_Cc => Current_Curly); | |
2731 | Current_Curly := Cc'Unchecked_Access; | |
2732 | ||
2733 | Has_Match := Match (Next - 3); | |
2734 | ||
2735 | -- Start on the WHILEM | |
2736 | ||
2737 | Current_Curly := Cc.Old_Cc; | |
2738 | return Has_Match; | |
2739 | end; | |
2740 | ||
2741 | when WHILEM => | |
2742 | return Match_Whilem (IP); | |
2743 | ||
2744 | when others => | |
2745 | raise Expression_Error; -- Invalid instruction | |
2746 | end case; | |
2747 | ||
2748 | Scan := Next; | |
2749 | end loop State_Machine; | |
2750 | ||
2751 | -- If we get here, there is no match. | |
2752 | -- For successful matches when EOP is the terminating point. | |
2753 | ||
2754 | return False; | |
2755 | end Match; | |
2756 | ||
2757 | --------------------------- | |
2758 | -- Match_Simple_Operator -- | |
2759 | --------------------------- | |
2760 | ||
2761 | function Match_Simple_Operator | |
2762 | (Op : Opcode; | |
2763 | Scan : Pointer; | |
2764 | Next : Pointer; | |
2765 | Greedy : Boolean) | |
2766 | return Boolean | |
2767 | is | |
2768 | Next_Char : Character := ASCII.Nul; | |
2769 | Next_Char_Known : Boolean := False; | |
2770 | No : Integer; -- Can be negative | |
2771 | Min : Natural; | |
2772 | Max : Natural := Natural'Last; | |
2773 | Operand_Code : Pointer; | |
2774 | Old : Natural; | |
2775 | Last_Pos : Natural; | |
2776 | Save : Natural := Input_Pos; | |
2777 | ||
2778 | begin | |
2779 | -- Lookahead to avoid useless match attempts | |
2780 | -- when we know what character comes next. | |
2781 | ||
2782 | if Program (Next) = EXACT then | |
2783 | Next_Char := Program (String_Operand (Next)); | |
2784 | Next_Char_Known := True; | |
2785 | end if; | |
2786 | ||
2787 | -- Find the minimal and maximal values for the operator | |
2788 | ||
2789 | case Op is | |
2790 | when STAR => | |
2791 | Min := 0; | |
2792 | Operand_Code := Operand (Scan); | |
2793 | ||
2794 | when PLUS => | |
2795 | Min := 1; | |
2796 | Operand_Code := Operand (Scan); | |
2797 | ||
2798 | when others => | |
2799 | Min := Read_Natural (Program, Scan + 3); | |
2800 | Max := Read_Natural (Program, Scan + 5); | |
2801 | Operand_Code := Scan + 7; | |
2802 | end case; | |
2803 | ||
2804 | -- Non greedy operators | |
2805 | ||
2806 | if not Greedy then | |
2807 | -- Test the minimal repetitions | |
2808 | ||
2809 | if Min /= 0 | |
2810 | and then Repeat (Operand_Code, Min) < Min | |
2811 | then | |
2812 | return False; | |
2813 | end if; | |
2814 | ||
2815 | Old := Input_Pos; | |
2816 | ||
2817 | -- Find the place where 'next' could work | |
2818 | ||
2819 | if Next_Char_Known then | |
2820 | -- Last position to check | |
2821 | ||
2822 | Last_Pos := Input_Pos + Max; | |
2823 | ||
2824 | if Last_Pos > Data'Last | |
2825 | or else Max = Natural'Last | |
2826 | then | |
2827 | Last_Pos := Data'Last; | |
2828 | end if; | |
2829 | ||
2830 | -- Look for the first possible opportunity | |
2831 | ||
2832 | loop | |
2833 | -- Find the next possible position | |
2834 | ||
2835 | while Input_Pos <= Last_Pos | |
2836 | and then Data (Input_Pos) /= Next_Char | |
2837 | loop | |
2838 | Input_Pos := Input_Pos + 1; | |
2839 | end loop; | |
2840 | ||
2841 | if Input_Pos > Last_Pos then | |
2842 | return False; | |
2843 | end if; | |
2844 | ||
2845 | -- Check that we still match if we stop | |
2846 | -- at the position we just found. | |
2847 | ||
2848 | declare | |
2849 | Num : constant Natural := Input_Pos - Old; | |
2850 | ||
2851 | begin | |
2852 | Input_Pos := Old; | |
2853 | ||
2854 | if Repeat (Operand_Code, Num) < Num then | |
2855 | return False; | |
2856 | end if; | |
2857 | end; | |
2858 | ||
2859 | -- Input_Pos now points to the new position | |
2860 | ||
2861 | if Match (Get_Next (Program, Scan)) then | |
2862 | return True; | |
2863 | end if; | |
2864 | ||
2865 | Old := Input_Pos; | |
2866 | Input_Pos := Input_Pos + 1; | |
2867 | end loop; | |
2868 | ||
2869 | -- We know what the next character is | |
2870 | ||
2871 | else | |
2872 | while Max >= Min loop | |
2873 | ||
2874 | -- If the next character matches | |
2875 | ||
2876 | if Match (Next) then | |
2877 | return True; | |
2878 | end if; | |
2879 | ||
2880 | Input_Pos := Save + Min; | |
2881 | ||
2882 | -- Could not or did not match -- move forward | |
2883 | ||
2884 | if Repeat (Operand_Code, 1) /= 0 then | |
2885 | Min := Min + 1; | |
2886 | else | |
2887 | return False; | |
2888 | end if; | |
2889 | end loop; | |
2890 | end if; | |
2891 | ||
2892 | return False; | |
2893 | ||
2894 | -- Greedy operators | |
2895 | ||
2896 | else | |
2897 | No := Repeat (Operand_Code, Max); | |
2898 | ||
2899 | -- ??? Perl has some special code here in case the | |
2900 | -- next instruction is of type EOL, since $ and \Z | |
2901 | -- can match before *and* after newline at the end. | |
2902 | ||
2903 | -- ??? Perl has some special code here in case (paren) | |
2904 | -- is True. | |
2905 | ||
2906 | -- Else, if we don't have any parenthesis | |
2907 | ||
2908 | while No >= Min loop | |
2909 | if not Next_Char_Known | |
2910 | or else (Input_Pos <= Data'Last | |
2911 | and then Data (Input_Pos) = Next_Char) | |
2912 | then | |
2913 | if Match (Next) then | |
2914 | return True; | |
2915 | end if; | |
2916 | end if; | |
2917 | ||
2918 | -- Could not or did not work, we back up | |
2919 | ||
2920 | No := No - 1; | |
2921 | Input_Pos := Save + No; | |
2922 | end loop; | |
2923 | return False; | |
2924 | end if; | |
2925 | end Match_Simple_Operator; | |
2926 | ||
2927 | ------------------ | |
2928 | -- Match_Whilem -- | |
2929 | ------------------ | |
2930 | ||
2931 | -- This is really hard to understand, because after we match what we're | |
2932 | -- trying to match, we must make sure the rest of the REx is going to | |
2933 | -- match for sure, and to do that we have to go back UP the parse tree | |
2934 | -- by recursing ever deeper. And if it fails, we have to reset our | |
2935 | -- parent's current state that we can try again after backing off. | |
2936 | ||
2937 | function Match_Whilem (IP : Pointer) return Boolean is | |
2938 | Cc : Current_Curly_Access := Current_Curly; | |
2939 | N : Natural := Cc.Cur + 1; | |
2940 | Ln : Natural; | |
2941 | Lastloc : Natural := Cc.Lastloc; | |
2942 | -- Detection of 0-len. | |
2943 | ||
2944 | begin | |
2945 | -- If degenerate scan matches "", assume scan done. | |
2946 | ||
2947 | if Input_Pos = Cc.Lastloc | |
2948 | and then N >= Cc.Min | |
2949 | then | |
2950 | -- Temporarily restore the old context, and check that we | |
2951 | -- match was comes after CURLYX. | |
2952 | ||
2953 | Current_Curly := Cc.Old_Cc; | |
2954 | ||
2955 | if Current_Curly /= null then | |
2956 | Ln := Current_Curly.Cur; | |
2957 | end if; | |
2958 | ||
2959 | if Match (Cc.Next) then | |
2960 | return True; | |
2961 | end if; | |
2962 | ||
2963 | if Current_Curly /= null then | |
2964 | Current_Curly.Cur := Ln; | |
2965 | end if; | |
2966 | ||
2967 | Current_Curly := Cc; | |
2968 | return False; | |
2969 | end if; | |
2970 | ||
2971 | -- First, just match a string of min scans. | |
2972 | ||
2973 | if N < Cc.Min then | |
2974 | Cc.Cur := N; | |
2975 | Cc.Lastloc := Input_Pos; | |
2976 | ||
2977 | if Match (Cc.Scan) then | |
2978 | return True; | |
2979 | end if; | |
2980 | ||
2981 | Cc.Cur := N - 1; | |
2982 | Cc.Lastloc := Lastloc; | |
2983 | return False; | |
2984 | end if; | |
2985 | ||
2986 | -- Prefer next over scan for minimal matching. | |
2987 | ||
2988 | if not Cc.Greedy then | |
2989 | Current_Curly := Cc.Old_Cc; | |
2990 | ||
2991 | if Current_Curly /= null then | |
2992 | Ln := Current_Curly.Cur; | |
2993 | end if; | |
2994 | ||
2995 | if Recurse_Match (Cc.Next, Cc.Paren_Floor) then | |
2996 | return True; | |
2997 | end if; | |
2998 | ||
2999 | if Current_Curly /= null then | |
3000 | Current_Curly.Cur := Ln; | |
3001 | end if; | |
3002 | ||
3003 | Current_Curly := Cc; | |
3004 | ||
3005 | -- Maximum greed exceeded ? | |
3006 | ||
3007 | if N >= Cc.Max then | |
3008 | return False; | |
3009 | end if; | |
3010 | ||
3011 | -- Try scanning more and see if it helps | |
3012 | Cc.Cur := N; | |
3013 | Cc.Lastloc := Input_Pos; | |
3014 | ||
3015 | if Recurse_Match (Cc.Scan, Cc.Paren_Floor) then | |
3016 | return True; | |
3017 | end if; | |
3018 | ||
3019 | Cc.Cur := N - 1; | |
3020 | Cc.Lastloc := Lastloc; | |
3021 | return False; | |
3022 | end if; | |
3023 | ||
3024 | -- Prefer scan over next for maximal matching | |
3025 | ||
3026 | if N < Cc.Max then -- more greed allowed ? | |
3027 | Cc.Cur := N; | |
3028 | Cc.Lastloc := Input_Pos; | |
3029 | ||
3030 | if Recurse_Match (Cc.Scan, Cc.Paren_Floor) then | |
3031 | return True; | |
3032 | end if; | |
3033 | end if; | |
3034 | ||
3035 | -- Failed deeper matches of scan, so see if this one works | |
3036 | ||
3037 | Current_Curly := Cc.Old_Cc; | |
3038 | ||
3039 | if Current_Curly /= null then | |
3040 | Ln := Current_Curly.Cur; | |
3041 | end if; | |
3042 | ||
3043 | if Match (Cc.Next) then | |
3044 | return True; | |
3045 | end if; | |
3046 | ||
3047 | if Current_Curly /= null then | |
3048 | Current_Curly.Cur := Ln; | |
3049 | end if; | |
3050 | ||
3051 | Current_Curly := Cc; | |
3052 | Cc.Cur := N - 1; | |
3053 | Cc.Lastloc := Lastloc; | |
3054 | return False; | |
3055 | end Match_Whilem; | |
3056 | ||
3057 | ------------ | |
3058 | -- Repeat -- | |
3059 | ------------ | |
3060 | ||
3061 | function Repeat | |
3062 | (IP : Pointer; | |
3063 | Max : Natural := Natural'Last) | |
3064 | return Natural | |
3065 | is | |
3066 | Scan : Natural := Input_Pos; | |
3067 | Last : Natural; | |
3068 | Op : constant Opcode := Opcode'Val (Character'Pos (Program (IP))); | |
3069 | Count : Natural; | |
3070 | C : Character; | |
3071 | Is_First : Boolean := True; | |
3072 | Bitmap : Character_Class; | |
3073 | ||
3074 | begin | |
3075 | if Max = Natural'Last or else Scan + Max - 1 > Data'Last then | |
3076 | Last := Data'Last; | |
3077 | else | |
3078 | Last := Scan + Max - 1; | |
3079 | end if; | |
3080 | ||
3081 | case Op is | |
3082 | when ANY => | |
3083 | while Scan <= Last | |
3084 | and then Data (Scan) /= ASCII.LF | |
3085 | loop | |
3086 | Scan := Scan + 1; | |
3087 | end loop; | |
3088 | ||
3089 | when SANY => | |
3090 | Scan := Last + 1; | |
3091 | ||
3092 | when EXACT => | |
3093 | ||
3094 | -- The string has only one character if Repeat was called | |
3095 | ||
3096 | C := Program (String_Operand (IP)); | |
3097 | while Scan <= Last | |
3098 | and then C = Data (Scan) | |
3099 | loop | |
3100 | Scan := Scan + 1; | |
3101 | end loop; | |
3102 | ||
3103 | when EXACTF => | |
3104 | ||
3105 | -- The string has only one character if Repeat was called | |
3106 | ||
3107 | C := Program (String_Operand (IP)); | |
3108 | while Scan <= Last | |
3109 | and then To_Lower (C) = Data (Scan) | |
3110 | loop | |
3111 | Scan := Scan + 1; | |
3112 | end loop; | |
3113 | ||
3114 | when ANYOF => | |
3115 | if Is_First then | |
3116 | Bitmap_Operand (Program, IP, Bitmap); | |
3117 | Is_First := False; | |
3118 | end if; | |
3119 | ||
3120 | while Scan <= Last | |
3121 | and then Get_From_Class (Bitmap, Data (Scan)) | |
3122 | loop | |
3123 | Scan := Scan + 1; | |
3124 | end loop; | |
3125 | ||
3126 | when ALNUM => | |
3127 | while Scan <= Last | |
3128 | and then Is_Alnum (Data (Scan)) | |
3129 | loop | |
3130 | Scan := Scan + 1; | |
3131 | end loop; | |
3132 | ||
3133 | when NALNUM => | |
3134 | while Scan <= Last | |
3135 | and then not Is_Alnum (Data (Scan)) | |
3136 | loop | |
3137 | Scan := Scan + 1; | |
3138 | end loop; | |
3139 | ||
3140 | when SPACE => | |
3141 | while Scan <= Last | |
3142 | and then Is_Space (Data (Scan)) | |
3143 | loop | |
3144 | Scan := Scan + 1; | |
3145 | end loop; | |
3146 | ||
3147 | when NSPACE => | |
3148 | while Scan <= Last | |
3149 | and then not Is_Space (Data (Scan)) | |
3150 | loop | |
3151 | Scan := Scan + 1; | |
3152 | end loop; | |
3153 | ||
3154 | when DIGIT => | |
3155 | while Scan <= Last | |
3156 | and then Is_Digit (Data (Scan)) | |
3157 | loop | |
3158 | Scan := Scan + 1; | |
3159 | end loop; | |
3160 | ||
3161 | when NDIGIT => | |
3162 | while Scan <= Last | |
3163 | and then not Is_Digit (Data (Scan)) | |
3164 | loop | |
3165 | Scan := Scan + 1; | |
3166 | end loop; | |
3167 | ||
3168 | when others => | |
3169 | raise Program_Error; | |
3170 | end case; | |
3171 | ||
3172 | Count := Scan - Input_Pos; | |
3173 | Input_Pos := Scan; | |
3174 | return Count; | |
3175 | end Repeat; | |
3176 | ||
3177 | --------- | |
3178 | -- Try -- | |
3179 | --------- | |
3180 | ||
3181 | function Try (Pos : in Positive) return Boolean is | |
3182 | begin | |
3183 | Input_Pos := Pos; | |
3184 | Last_Paren := 0; | |
3185 | Matches_Full := (others => No_Match); | |
3186 | ||
3187 | if Match (Program_First + 1) then | |
3188 | Matches_Full (0) := (Pos, Input_Pos - 1); | |
3189 | return True; | |
3190 | end if; | |
3191 | ||
3192 | return False; | |
3193 | end Try; | |
3194 | ||
3195 | -- Start of processing for Match | |
3196 | ||
3197 | begin | |
3198 | -- Do we have the regexp Never_Match? | |
3199 | ||
3200 | if Self.Size = 0 then | |
3201 | Matches (0) := No_Match; | |
3202 | return; | |
3203 | end if; | |
3204 | ||
3205 | -- Check validity of program | |
3206 | ||
3207 | pragma Assert | |
3208 | (Program (Program_First) = MAGIC, | |
3209 | "Corrupted Pattern_Matcher"); | |
3210 | ||
3211 | -- If there is a "must appear" string, look for it | |
3212 | ||
3213 | if Self.Must_Have_Length > 0 then | |
3214 | declare | |
3215 | First : constant Character := Program (Self.Must_Have); | |
3216 | Must_First : constant Pointer := Self.Must_Have; | |
3217 | Must_Last : constant Pointer := | |
3218 | Must_First + Pointer (Self.Must_Have_Length - 1); | |
3219 | Next_Try : Natural := Index (Data'First, First); | |
3220 | ||
3221 | begin | |
3222 | while Next_Try /= 0 | |
3223 | and then Data (Next_Try .. Next_Try + Self.Must_Have_Length - 1) | |
3224 | = String (Program (Must_First .. Must_Last)) | |
3225 | loop | |
3226 | Next_Try := Index (Next_Try + 1, First); | |
3227 | end loop; | |
3228 | ||
3229 | if Next_Try = 0 then | |
3230 | Matches_Full := (others => No_Match); | |
3231 | return; -- Not present | |
3232 | end if; | |
3233 | end; | |
3234 | end if; | |
3235 | ||
3236 | -- Mark beginning of line for ^ | |
3237 | ||
3238 | BOL_Pos := Data'First; | |
3239 | ||
3240 | -- Simplest case first: an anchored match need be tried only once | |
3241 | ||
3242 | if Self.Anchored and then (Self.Flags and Multiple_Lines) = 0 then | |
3243 | Matched := Try (Data'First); | |
3244 | ||
3245 | elsif Self.Anchored then | |
3246 | declare | |
3247 | Next_Try : Natural := Data'First; | |
3248 | begin | |
3249 | -- Test the first position in the buffer | |
3250 | Matched := Try (Next_Try); | |
3251 | ||
3252 | -- Else only test after newlines | |
3253 | ||
3254 | if not Matched then | |
3255 | while Next_Try <= Data'Last loop | |
3256 | while Next_Try <= Data'Last | |
3257 | and then Data (Next_Try) /= ASCII.LF | |
3258 | loop | |
3259 | Next_Try := Next_Try + 1; | |
3260 | end loop; | |
3261 | ||
3262 | Next_Try := Next_Try + 1; | |
3263 | ||
3264 | if Next_Try <= Data'Last then | |
3265 | Matched := Try (Next_Try); | |
3266 | exit when Matched; | |
3267 | end if; | |
3268 | end loop; | |
3269 | end if; | |
3270 | end; | |
3271 | ||
3272 | elsif Self.First /= ASCII.NUL then | |
3273 | ||
3274 | -- We know what char it must start with | |
3275 | ||
3276 | declare | |
3277 | Next_Try : Natural := Index (Data'First, Self.First); | |
3278 | ||
3279 | begin | |
3280 | while Next_Try /= 0 loop | |
3281 | Matched := Try (Next_Try); | |
3282 | exit when Matched; | |
3283 | Next_Try := Index (Next_Try + 1, Self.First); | |
3284 | end loop; | |
3285 | end; | |
3286 | ||
3287 | else | |
3288 | -- Messy cases: try all locations (including for the empty string) | |
3289 | ||
3290 | Matched := Try (Data'First); | |
3291 | ||
3292 | if not Matched then | |
3293 | for S in Data'First + 1 .. Data'Last loop | |
3294 | Matched := Try (S); | |
3295 | exit when Matched; | |
3296 | end loop; | |
3297 | end if; | |
3298 | end if; | |
3299 | ||
3300 | -- Matched has its value | |
3301 | ||
3302 | for J in Last_Paren + 1 .. Matches'Last loop | |
3303 | Matches_Full (J) := No_Match; | |
3304 | end loop; | |
3305 | ||
3306 | Matches := Matches_Full (Matches'Range); | |
3307 | return; | |
3308 | end Match; | |
3309 | ||
3310 | function Match | |
3311 | (Self : Pattern_Matcher; | |
3312 | Data : String) | |
3313 | return Natural | |
3314 | is | |
3315 | Matches : Match_Array (0 .. 0); | |
3316 | ||
3317 | begin | |
3318 | Match (Self, Data, Matches); | |
3319 | if Matches (0) = No_Match then | |
3320 | return Data'First - 1; | |
3321 | else | |
3322 | return Matches (0).First; | |
3323 | end if; | |
3324 | end Match; | |
3325 | ||
3326 | procedure Match | |
3327 | (Expression : String; | |
3328 | Data : String; | |
3329 | Matches : out Match_Array; | |
3330 | Size : Program_Size := 0) | |
3331 | is | |
3332 | PM : Pattern_Matcher (Size); | |
3333 | Finalize_Size : Program_Size; | |
3334 | ||
3335 | begin | |
3336 | if Size = 0 then | |
3337 | Match (Compile (Expression), Data, Matches); | |
3338 | else | |
3339 | Compile (PM, Expression, Finalize_Size); | |
3340 | Match (PM, Data, Matches); | |
3341 | end if; | |
3342 | end Match; | |
3343 | ||
3344 | function Match | |
3345 | (Expression : String; | |
3346 | Data : String; | |
3347 | Size : Program_Size := 0) | |
3348 | return Natural | |
3349 | is | |
3350 | PM : Pattern_Matcher (Size); | |
3351 | Final_Size : Program_Size; -- unused | |
3352 | ||
3353 | begin | |
3354 | if Size = 0 then | |
3355 | return Match (Compile (Expression), Data); | |
3356 | else | |
3357 | Compile (PM, Expression, Final_Size); | |
3358 | return Match (PM, Data); | |
3359 | end if; | |
3360 | end Match; | |
3361 | ||
3362 | function Match | |
3363 | (Expression : String; | |
3364 | Data : String; | |
3365 | Size : Program_Size := 0) | |
3366 | return Boolean | |
3367 | is | |
3368 | Matches : Match_Array (0 .. 0); | |
3369 | PM : Pattern_Matcher (Size); | |
3370 | Final_Size : Program_Size; -- unused | |
3371 | ||
3372 | begin | |
3373 | if Size = 0 then | |
3374 | Match (Compile (Expression), Data, Matches); | |
3375 | else | |
3376 | Compile (PM, Expression, Final_Size); | |
3377 | Match (PM, Data, Matches); | |
3378 | end if; | |
3379 | ||
3380 | return Matches (0).First >= Data'First; | |
3381 | end Match; | |
3382 | ||
3383 | ------------- | |
3384 | -- Operand -- | |
3385 | ------------- | |
3386 | ||
3387 | function Operand (P : Pointer) return Pointer is | |
3388 | begin | |
3389 | return P + 3; | |
3390 | end Operand; | |
3391 | ||
3392 | -------------- | |
3393 | -- Optimize -- | |
3394 | -------------- | |
3395 | ||
3396 | procedure Optimize (Self : in out Pattern_Matcher) is | |
3397 | Max_Length : Program_Size; | |
3398 | This_Length : Program_Size; | |
3399 | Longest : Pointer; | |
3400 | Scan : Pointer; | |
3401 | Program : Program_Data renames Self.Program; | |
3402 | ||
3403 | begin | |
3404 | -- Start with safe defaults (no optimization): | |
3405 | -- * No known first character of match | |
3406 | -- * Does not necessarily start at beginning of line | |
3407 | -- * No string known that has to appear in data | |
3408 | ||
3409 | Self.First := ASCII.NUL; | |
3410 | Self.Anchored := False; | |
3411 | Self.Must_Have := Program'Last + 1; | |
3412 | Self.Must_Have_Length := 0; | |
3413 | ||
3414 | Scan := Program_First + 1; -- First instruction (can be anything) | |
3415 | ||
3416 | if Program (Scan) = EXACT then | |
3417 | Self.First := Program (String_Operand (Scan)); | |
3418 | ||
3419 | elsif Program (Scan) = BOL | |
3420 | or else Program (Scan) = SBOL | |
3421 | or else Program (Scan) = MBOL | |
3422 | then | |
3423 | Self.Anchored := True; | |
3424 | end if; | |
3425 | ||
3426 | -- If there's something expensive in the regexp, find the | |
3427 | -- longest literal string that must appear and make it the | |
3428 | -- regmust. Resolve ties in favor of later strings, since | |
3429 | -- the regstart check works with the beginning of the regexp. | |
3430 | -- and avoiding duplication strengthens checking. Not a | |
3431 | -- strong reason, but sufficient in the absence of others. | |
3432 | ||
3433 | if False then -- if Flags.SP_Start then ??? | |
3434 | Longest := 0; | |
3435 | Max_Length := 0; | |
3436 | while Scan /= 0 loop | |
3437 | if Program (Scan) = EXACT or else Program (Scan) = EXACTF then | |
3438 | This_Length := String_Length (Program, Scan); | |
3439 | ||
3440 | if This_Length >= Max_Length then | |
3441 | Longest := String_Operand (Scan); | |
3442 | Max_Length := This_Length; | |
3443 | end if; | |
3444 | end if; | |
3445 | ||
3446 | Scan := Get_Next (Program, Scan); | |
3447 | end loop; | |
3448 | ||
3449 | Self.Must_Have := Longest; | |
3450 | Self.Must_Have_Length := Natural (Max_Length) + 1; | |
3451 | end if; | |
3452 | end Optimize; | |
3453 | ||
3454 | ----------------- | |
3455 | -- Paren_Count -- | |
3456 | ----------------- | |
3457 | ||
3458 | function Paren_Count (Regexp : Pattern_Matcher) return Match_Count is | |
3459 | begin | |
3460 | return Regexp.Paren_Count; | |
3461 | end Paren_Count; | |
3462 | ||
3463 | ----------- | |
3464 | -- Quote -- | |
3465 | ----------- | |
3466 | ||
3467 | function Quote (Str : String) return String is | |
3468 | S : String (1 .. Str'Length * 2); | |
3469 | Last : Natural := 0; | |
3470 | ||
3471 | begin | |
3472 | for J in Str'Range loop | |
3473 | case Str (J) is | |
3474 | when '^' | '$' | '|' | '*' | '+' | '?' | '{' | |
3475 | | '}' | '[' | ']' | '(' | ')' | '\' => | |
3476 | ||
3477 | S (Last + 1) := '\'; | |
3478 | S (Last + 2) := Str (J); | |
3479 | Last := Last + 2; | |
3480 | ||
3481 | when others => | |
3482 | S (Last + 1) := Str (J); | |
3483 | Last := Last + 1; | |
3484 | end case; | |
3485 | end loop; | |
3486 | ||
3487 | return S (1 .. Last); | |
3488 | end Quote; | |
3489 | ||
3490 | ------------------ | |
3491 | -- Read_Natural -- | |
3492 | ------------------ | |
3493 | ||
3494 | function Read_Natural | |
3495 | (Program : Program_Data; | |
3496 | IP : Pointer) | |
3497 | return Natural | |
3498 | is | |
3499 | begin | |
3500 | return Character'Pos (Program (IP)) + | |
3501 | 256 * Character'Pos (Program (IP + 1)); | |
3502 | end Read_Natural; | |
3503 | ||
3504 | ----------------- | |
3505 | -- Reset_Class -- | |
3506 | ----------------- | |
3507 | ||
3508 | procedure Reset_Class (Bitmap : in out Character_Class) is | |
3509 | begin | |
3510 | Bitmap := (others => 0); | |
3511 | end Reset_Class; | |
3512 | ||
3513 | ------------------ | |
3514 | -- Set_In_Class -- | |
3515 | ------------------ | |
3516 | ||
3517 | procedure Set_In_Class | |
3518 | (Bitmap : in out Character_Class; | |
3519 | C : Character) | |
3520 | is | |
3521 | Value : constant Class_Byte := Character'Pos (C); | |
3522 | ||
3523 | begin | |
3524 | Bitmap (Value / 8) := Bitmap (Value / 8) | |
3525 | or Bit_Conversion (Value mod 8); | |
3526 | end Set_In_Class; | |
3527 | ||
3528 | ------------------- | |
3529 | -- String_Length -- | |
3530 | ------------------- | |
3531 | ||
3532 | function String_Length | |
3533 | (Program : Program_Data; | |
3534 | P : Pointer) | |
3535 | return Program_Size | |
3536 | is | |
3537 | begin | |
3538 | pragma Assert (Program (P) = EXACT or else Program (P) = EXACTF); | |
3539 | return Character'Pos (Program (P + 3)); | |
3540 | end String_Length; | |
3541 | ||
3542 | -------------------- | |
3543 | -- String_Operand -- | |
3544 | -------------------- | |
3545 | ||
3546 | function String_Operand (P : Pointer) return Pointer is | |
3547 | begin | |
3548 | return P + 4; | |
3549 | end String_Operand; | |
3550 | ||
3551 | end GNAT.Regpat; |