/[pcre]/code/trunk/HACKING
ViewVC logotype

Contents of /code/trunk/HACKING

Parent Directory Parent Directory | Revision Log Revision Log


Revision 604 - (show annotations)
Thu Jun 2 19:04:54 2011 UTC (4 years ago) by ph10
File size: 19558 byte(s)
Error occurred while calculating annotation data.
Refactoring to reduce stack usage for possessively quantified subpatterns. Also 
fixed a number of bugs related to repeated subpatterns. Some further tidies 
consequent on the removal of OP_OPT are also in this patch.
1 Technical Notes about PCRE
2 --------------------------
3
4 These are very rough technical notes that record potentially useful information
5 about PCRE internals.
6
7
8 Historical note 1
9 -----------------
10
11 Many years ago I implemented some regular expression functions to an algorithm
12 suggested by Martin Richards. These were not Unix-like in form, and were quite
13 restricted in what they could do by comparison with Perl. The interesting part
14 about the algorithm was that the amount of space required to hold the compiled
15 form of an expression was known in advance. The code to apply an expression did
16 not operate by backtracking, as the original Henry Spencer code and current
17 Perl code does, but instead checked all possibilities simultaneously by keeping
18 a list of current states and checking all of them as it advanced through the
19 subject string. In the terminology of Jeffrey Friedl's book, it was a "DFA
20 algorithm", though it was not a traditional Finite State Machine (FSM). When
21 the pattern was all used up, all remaining states were possible matches, and
22 the one matching the longest subset of the subject string was chosen. This did
23 not necessarily maximize the individual wild portions of the pattern, as is
24 expected in Unix and Perl-style regular expressions.
25
26
27 Historical note 2
28 -----------------
29
30 By contrast, the code originally written by Henry Spencer (which was
31 subsequently heavily modified for Perl) compiles the expression twice: once in
32 a dummy mode in order to find out how much store will be needed, and then for
33 real. (The Perl version probably doesn't do this any more; I'm talking about
34 the original library.) The execution function operates by backtracking and
35 maximizing (or, optionally, minimizing in Perl) the amount of the subject that
36 matches individual wild portions of the pattern. This is an "NFA algorithm" in
37 Friedl's terminology.
38
39
40 OK, here's the real stuff
41 -------------------------
42
43 For the set of functions that form the "basic" PCRE library (which are
44 unrelated to those mentioned above), I tried at first to invent an algorithm
45 that used an amount of store bounded by a multiple of the number of characters
46 in the pattern, to save on compiling time. However, because of the greater
47 complexity in Perl regular expressions, I couldn't do this. In any case, a
48 first pass through the pattern is helpful for other reasons.
49
50
51 Computing the memory requirement: how it was
52 --------------------------------------------
53
54 Up to and including release 6.7, PCRE worked by running a very degenerate first
55 pass to calculate a maximum store size, and then a second pass to do the real
56 compile - which might use a bit less than the predicted amount of memory. The
57 idea was that this would turn out faster than the Henry Spencer code because
58 the first pass is degenerate and the second pass can just store stuff straight
59 into the vector, which it knows is big enough.
60
61
62 Computing the memory requirement: how it is
63 -------------------------------------------
64
65 By the time I was working on a potential 6.8 release, the degenerate first pass
66 had become very complicated and hard to maintain. Indeed one of the early
67 things I did for 6.8 was to fix Yet Another Bug in the memory computation. Then
68 I had a flash of inspiration as to how I could run the real compile function in
69 a "fake" mode that enables it to compute how much memory it would need, while
70 actually only ever using a few hundred bytes of working memory, and without too
71 many tests of the mode that might slow it down. So I refactored the compiling
72 functions to work this way. This got rid of about 600 lines of source. It
73 should make future maintenance and development easier. As this was such a major
74 change, I never released 6.8, instead upping the number to 7.0 (other quite
75 major changes were also present in the 7.0 release).
76
77 A side effect of this work was that the previous limit of 200 on the nesting
78 depth of parentheses was removed. However, there is a downside: pcre_compile()
79 runs more slowly than before (30% or more, depending on the pattern) because it
80 is doing a full analysis of the pattern. My hope was that this would not be a
81 big issue, and in the event, nobody has commented on it.
82
83
84 Traditional matching function
85 -----------------------------
86
87 The "traditional", and original, matching function is called pcre_exec(), and
88 it implements an NFA algorithm, similar to the original Henry Spencer algorithm
89 and the way that Perl works. This is not surprising, since it is intended to be
90 as compatible with Perl as possible. This is the function most users of PCRE
91 will use most of the time.
92
93
94 Supplementary matching function
95 -------------------------------
96
97 From PCRE 6.0, there is also a supplementary matching function called
98 pcre_dfa_exec(). This implements a DFA matching algorithm that searches
99 simultaneously for all possible matches that start at one point in the subject
100 string. (Going back to my roots: see Historical Note 1 above.) This function
101 intreprets the same compiled pattern data as pcre_exec(); however, not all the
102 facilities are available, and those that are do not always work in quite the
103 same way. See the user documentation for details.
104
105 The algorithm that is used for pcre_dfa_exec() is not a traditional FSM,
106 because it may have a number of states active at one time. More work would be
107 needed at compile time to produce a traditional FSM where only one state is
108 ever active at once. I believe some other regex matchers work this way.
109
110
111 Changeable options
112 ------------------
113
114 The /i, /m, or /s options (PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL) may
115 change in the middle of patterns. From PCRE 8.13, their processing is handled
116 entirely at compile time by generating different opcodes for the different
117 settings. The runtime functions do not need to keep track of an options state
118 any more.
119
120
121 Format of compiled patterns
122 ---------------------------
123
124 The compiled form of a pattern is a vector of bytes, containing items of
125 variable length. The first byte in an item is an opcode, and the length of the
126 item is either implicit in the opcode or contained in the data bytes that
127 follow it.
128
129 In many cases below LINK_SIZE data values are specified for offsets within the
130 compiled pattern. The default value for LINK_SIZE is 2, but PCRE can be
131 compiled to use 3-byte or 4-byte values for these offsets (impairing the
132 performance). This is necessary only when patterns whose compiled length is
133 greater than 64K are going to be processed. In this description, we assume the
134 "normal" compilation options. Data values that are counts (e.g. for
135 quantifiers) are always just two bytes long.
136
137 Opcodes with no following data
138 ------------------------------
139
140 These items are all just one byte long
141
142 OP_END end of pattern
143 OP_ANY match any one character other than newline
144 OP_ALLANY match any one character, including newline
145 OP_ANYBYTE match any single byte, even in UTF-8 mode
146 OP_SOD match start of data: \A
147 OP_SOM, start of match (subject + offset): \G
148 OP_SET_SOM, set start of match (\K)
149 OP_CIRC ^ (start of data)
150 OP_CIRCM ^ multiline mode (start of data or after newline)
151 OP_NOT_WORD_BOUNDARY \W
152 OP_WORD_BOUNDARY \w
153 OP_NOT_DIGIT \D
154 OP_DIGIT \d
155 OP_NOT_HSPACE \H
156 OP_HSPACE \h
157 OP_NOT_WHITESPACE \S
158 OP_WHITESPACE \s
159 OP_NOT_VSPACE \V
160 OP_VSPACE \v
161 OP_NOT_WORDCHAR \W
162 OP_WORDCHAR \w
163 OP_EODN match end of data or \n at end: \Z
164 OP_EOD match end of data: \z
165 OP_DOLL $ (end of data, or before final newline)
166 OP_DOLLM $ multiline mode (end of data or before newline)
167 OP_EXTUNI match an extended Unicode character
168 OP_ANYNL match any Unicode newline sequence
169
170 OP_ACCEPT ) These are Perl 5.10's "backtracking control
171 OP_COMMIT ) verbs". If OP_ACCEPT is inside capturing
172 OP_FAIL ) parentheses, it may be preceded by one or more
173 OP_PRUNE ) OP_CLOSE, followed by a 2-byte number,
174 OP_SKIP ) indicating which parentheses must be closed.
175
176
177 Backtracking control verbs with data
178 ------------------------------------
179
180 OP_THEN is followed by a LINK_SIZE offset, which is the distance back to the
181 start of the current branch.
182
183 OP_MARK is followed by the mark name, preceded by a one-byte length, and
184 followed by a binary zero. For (*PRUNE), (*SKIP), and (*THEN) with arguments,
185 the opcodes OP_PRUNE_ARG, OP_SKIP_ARG, and OP_THEN_ARG are used. For the first
186 two, the name follows immediately; for OP_THEN_ARG, it follows the LINK_SIZE
187 offset value.
188
189
190 Matching literal characters
191 ---------------------------
192
193 The OP_CHAR opcode is followed by a single character that is to be matched
194 casefully. For caseless matching, OP_CHARI is used. In UTF-8 mode, the
195 character may be more than one byte long. (Earlier versions of PCRE used
196 multi-character strings, but this was changed to allow some new features to be
197 added.)
198
199
200 Repeating single characters
201 ---------------------------
202
203 The common repeats (*, +, ?) when applied to a single character use the
204 following opcodes, which come in caseful and caseless versions:
205
206 Caseful Caseless
207 OP_STAR OP_STARI
208 OP_MINSTAR OP_MINSTARI
209 OP_POSSTAR OP_POSSTARI
210 OP_PLUS OP_PLUSI
211 OP_MINPLUS OP_MINPLUSI
212 OP_POSPLUS OP_POSPLUSI
213 OP_QUERY OP_QUERYI
214 OP_MINQUERY OP_MINQUERYI
215 OP_POSQUERY OP_POSQUERYI
216
217 In ASCII mode, these are two-byte items; in UTF-8 mode, the length is variable.
218 Those with "MIN" in their name are the minimizing versions. Those with "POS" in
219 their names are possessive versions. Each is followed by the character that is
220 to be repeated. Other repeats make use of these opcodes:
221
222 Caseful Caseless
223 OP_UPTO OP_UPTOI
224 OP_MINUPTO OP_MINUPTOI
225 OP_POSUPTO OP_POSUPTOI
226 OP_EXACT OP_EXACTI
227
228 Each of these is followed by a two-byte count (most significant first) and the
229 repeated character. OP_UPTO matches from 0 to the given number. A repeat with a
230 non-zero minimum and a fixed maximum is coded as an OP_EXACT followed by an
231 OP_UPTO (or OP_MINUPTO or OPT_POSUPTO).
232
233
234 Repeating character types
235 -------------------------
236
237 Repeats of things like \d are done exactly as for single characters, except
238 that instead of a character, the opcode for the type is stored in the data
239 byte. The opcodes are:
240
241 OP_TYPESTAR
242 OP_TYPEMINSTAR
243 OP_TYPEPOSSTAR
244 OP_TYPEPLUS
245 OP_TYPEMINPLUS
246 OP_TYPEPOSPLUS
247 OP_TYPEQUERY
248 OP_TYPEMINQUERY
249 OP_TYPEPOSQUERY
250 OP_TYPEUPTO
251 OP_TYPEMINUPTO
252 OP_TYPEPOSUPTO
253 OP_TYPEEXACT
254
255
256 Match by Unicode property
257 -------------------------
258
259 OP_PROP and OP_NOTPROP are used for positive and negative matches of a
260 character by testing its Unicode property (the \p and \P escape sequences).
261 Each is followed by two bytes that encode the desired property as a type and a
262 value.
263
264 Repeats of these items use the OP_TYPESTAR etc. set of opcodes, followed by
265 three bytes: OP_PROP or OP_NOTPROP and then the desired property type and
266 value.
267
268
269 Character classes
270 -----------------
271
272 If there is only one character, OP_CHAR or OP_CHARI is used for a positive
273 class, and OP_NOT or OP_NOTI for a negative one (that is, for something like
274 [^a]). However, in UTF-8 mode, the use of OP_NOT[I] applies only to characters
275 with values < 128, because OP_NOT[I] is confined to single bytes.
276
277 Another set of 13 repeating opcodes (called OP_NOTSTAR etc.) are used for a
278 repeated, negated, single-character class. The normal single-character opcodes
279 (OP_STAR, etc.) are used for a repeated positive single-character class.
280
281 When there is more than one character in a class and all the characters are
282 less than 256, OP_CLASS is used for a positive class, and OP_NCLASS for a
283 negative one. In either case, the opcode is followed by a 32-byte bit map
284 containing a 1 bit for every character that is acceptable. The bits are counted
285 from the least significant end of each byte. In caseless mode, bits for both
286 cases are set.
287
288 The reason for having both OP_CLASS and OP_NCLASS is so that, in UTF-8 mode,
289 subject characters with values greater than 256 can be handled correctly. For
290 OP_CLASS they do not match, whereas for OP_NCLASS they do.
291
292 For classes containing characters with values > 255, OP_XCLASS is used. It
293 optionally uses a bit map (if any characters lie within it), followed by a list
294 of pairs (for a range) and single characters. In caseless mode, both cases are
295 explicitly listed. There is a flag character than indicates whether it is a
296 positive or a negative class.
297
298
299 Back references
300 ---------------
301
302 OP_REF (caseful) or OP_REFI (caseless) is followed by two bytes containing the
303 reference number.
304
305
306 Repeating character classes and back references
307 -----------------------------------------------
308
309 Single-character classes are handled specially (see above). This section
310 applies to OP_CLASS and OP_REF[I]. In both cases, the repeat information
311 follows the base item. The matching code looks at the following opcode to see
312 if it is one of
313
314 OP_CRSTAR
315 OP_CRMINSTAR
316 OP_CRPLUS
317 OP_CRMINPLUS
318 OP_CRQUERY
319 OP_CRMINQUERY
320 OP_CRRANGE
321 OP_CRMINRANGE
322
323 All but the last two are just single-byte items. The others are followed by
324 four bytes of data, comprising the minimum and maximum repeat counts. There are
325 no special possessive opcodes for these repeats; a possessive repeat is
326 compiled into an atomic group.
327
328
329 Brackets and alternation
330 ------------------------
331
332 A pair of non-capturing (round) brackets is wrapped round each expression at
333 compile time, so alternation always happens in the context of brackets.
334
335 [Note for North Americans: "bracket" to some English speakers, including
336 myself, can be round, square, curly, or pointy. Hence this usage.]
337
338 Non-capturing brackets use the opcode OP_BRA. Originally PCRE was limited to 99
339 capturing brackets and it used a different opcode for each one. From release
340 3.5, the limit was removed by putting the bracket number into the data for
341 higher-numbered brackets. From release 7.0 all capturing brackets are handled
342 this way, using the single opcode OP_CBRA.
343
344 A bracket opcode is followed by LINK_SIZE bytes which give the offset to the
345 next alternative OP_ALT or, if there aren't any branches, to the matching
346 OP_KET opcode. Each OP_ALT is followed by LINK_SIZE bytes giving the offset to
347 the next one, or to the OP_KET opcode. For capturing brackets, the bracket
348 number immediately follows the offset, always as a 2-byte item.
349
350 OP_KET is used for subpatterns that do not repeat indefinitely, while
351 OP_KETRMIN and OP_KETRMAX are used for indefinite repetitions, minimally or
352 maximally respectively (see below for possessive repetitions). All three are
353 followed by LINK_SIZE bytes giving (as a positive number) the offset back to
354 the matching bracket opcode.
355
356 If a subpattern is quantified such that it is permitted to match zero times, it
357 is preceded by one of OP_BRAZERO, OP_BRAMINZERO, or OP_SKIPZERO. These are
358 single-byte opcodes that tell the matcher that skipping the following
359 subpattern entirely is a valid branch. In the case of the first two, not
360 skipping the pattern is also valid (greedy and non-greedy). The third is used
361 when a pattern has the quantifier {0,0}. It cannot be entirely discarded,
362 because it may be called as a subroutine from elsewhere in the regex.
363
364 A subpattern with an indefinite maximum repetition is replicated in the
365 compiled data its minimum number of times (or once with OP_BRAZERO if the
366 minimum is zero), with the final copy terminating with OP_KETRMIN or OP_KETRMAX
367 as appropriate.
368
369 A subpattern with a bounded maximum repetition is replicated in a nested
370 fashion up to the maximum number of times, with OP_BRAZERO or OP_BRAMINZERO
371 before each replication after the minimum, so that, for example, (abc){2,5} is
372 compiled as (abc)(abc)((abc)((abc)(abc)?)?)?, except that each bracketed group
373 has the same number.
374
375 When a repeated subpattern has an unbounded upper limit, it is checked to see
376 whether it could match an empty string. If this is the case, the opcode in the
377 final replication is changed to OP_SBRA or OP_SCBRA. This tells the matcher
378 that it needs to check for matching an empty string when it hits OP_KETRMIN or
379 OP_KETRMAX, and if so, to break the loop.
380
381 Possessive brackets
382 -------------------
383
384 When a repeated group (capturing or non-capturing) is marked as possessive by
385 the "+" notation, e.g. (abc)++, different opcodes are used. Their names all
386 have POS on the end, e.g. OP_BRAPOS instead of OP_BRA and OP_SCPBRPOS instead
387 of OP_SCBRA. The end of such a group is marked by OP_KETRPOS. If the minimum
388 repetition is zero, the group is preceded by OP_BRAPOSZERO.
389
390
391 Assertions
392 ----------
393
394 Forward assertions are just like other subpatterns, but starting with one of
395 the opcodes OP_ASSERT or OP_ASSERT_NOT. Backward assertions use the opcodes
396 OP_ASSERTBACK and OP_ASSERTBACK_NOT, and the first opcode inside the assertion
397 is OP_REVERSE, followed by a two byte count of the number of characters to move
398 back the pointer in the subject string. When operating in UTF-8 mode, the count
399 is a character count rather than a byte count. A separate count is present in
400 each alternative of a lookbehind assertion, allowing them to have different
401 fixed lengths.
402
403
404 Once-only (atomic) subpatterns
405 ------------------------------
406
407 These are also just like other subpatterns, but they start with the opcode
408 OP_ONCE. The check for matching an empty string in an unbounded repeat is
409 handled entirely at runtime, so there is just this one opcode.
410
411
412 Conditional subpatterns
413 -----------------------
414
415 These are like other subpatterns, but they start with the opcode OP_COND, or
416 OP_SCOND for one that might match an empty string in an unbounded repeat. If
417 the condition is a back reference, this is stored at the start of the
418 subpattern using the opcode OP_CREF followed by two bytes containing the
419 reference number. OP_NCREF is used instead if the reference was generated by
420 name (so that the runtime code knows to check for duplicate names).
421
422 If the condition is "in recursion" (coded as "(?(R)"), or "in recursion of
423 group x" (coded as "(?(Rx)"), the group number is stored at the start of the
424 subpattern using the opcode OP_RREF or OP_NRREF (cf OP_NCREF), and a value of
425 zero for "the whole pattern". For a DEFINE condition, just the single byte
426 OP_DEF is used (it has no associated data). Otherwise, a conditional subpattern
427 always starts with one of the assertions.
428
429
430 Recursion
431 ---------
432
433 Recursion either matches the current regex, or some subexpression. The opcode
434 OP_RECURSE is followed by an value which is the offset to the starting bracket
435 from the start of the whole pattern. From release 6.5, OP_RECURSE is
436 automatically wrapped inside OP_ONCE brackets (because otherwise some patterns
437 broke it). OP_RECURSE is also used for "subroutine" calls, even though they
438 are not strictly a recursion.
439
440
441 Callout
442 -------
443
444 OP_CALLOUT is followed by one byte of data that holds a callout number in the
445 range 0 to 254 for manual callouts, or 255 for an automatic callout. In both
446 cases there follows a two-byte value giving the offset in the pattern to the
447 start of the following item, and another two-byte item giving the length of the
448 next item.
449
450
451 Philip Hazel
452 May 2011

  ViewVC Help
Powered by ViewVC 1.1.5