/[pcre]/code/trunk/doc/Tech.Notes
ViewVC logotype

Contents of /code/trunk/doc/Tech.Notes

Parent Directory Parent Directory | Revision Log Revision Log


Revision 75 - (show annotations) (download)
Sat Feb 24 21:40:37 2007 UTC (7 years, 7 months ago) by nigel
File size: 12771 byte(s)
Load pcre-5.0 into code/trunk.

1 Technical Notes about PCRE
2 --------------------------
3
4 Historical note 1
5 -----------------
6
7 Many years ago I implemented some regular expression functions to an algorithm
8 suggested by Martin Richards. These were not Unix-like in form, and were quite
9 restricted in what they could do by comparison with Perl. The interesting part
10 about the algorithm was that the amount of space required to hold the compiled
11 form of an expression was known in advance. The code to apply an expression did
12 not operate by backtracking, as the original Henry Spencer code and current
13 Perl code does, but instead checked all possibilities simultaneously by keeping
14 a list of current states and checking all of them as it advanced through the
15 subject string. In the terminology of Jeffrey Friedl's book, it was a "DFA
16 algorithm". When the pattern was all used up, all remaining states were
17 possible matches, and the one matching the longest subset of the subject string
18 was chosen. This did not necessarily maximize the individual wild portions of
19 the pattern, as is expected in Unix and Perl-style regular expressions.
20
21 Historical note 2
22 -----------------
23
24 By contrast, the code originally written by Henry Spencer and subsequently
25 heavily modified for Perl actually compiles the expression twice: once in a
26 dummy mode in order to find out how much store will be needed, and then for
27 real. The execution function operates by backtracking and maximizing (or,
28 optionally, minimizing in Perl) the amount of the subject that matches
29 individual wild portions of the pattern. This is an "NFA algorithm" in Friedl's
30 terminology.
31
32 OK, here's the real stuff
33 -------------------------
34
35 For the set of functions that forms PCRE (which are unrelated to those
36 mentioned above), I tried at first to invent an algorithm that used an amount
37 of store bounded by a multiple of the number of characters in the pattern, to
38 save on compiling time. However, because of the greater complexity in Perl
39 regular expressions, I couldn't do this. In any case, a first pass through the
40 pattern is needed, for a number of reasons. PCRE works by running a very
41 degenerate first pass to calculate a maximum store size, and then a second pass
42 to do the real compile - which may use a bit less than the predicted amount of
43 store. The idea is that this is going to turn out faster because the first pass
44 is degenerate and the second pass can just store stuff straight into the
45 vector. It does make the compiling functions bigger, of course, but they have
46 got quite big anyway to handle all the Perl stuff.
47
48 The compiled form of a pattern is a vector of bytes, containing items of
49 variable length. The first byte in an item is an opcode, and the length of the
50 item is either implicit in the opcode or contained in the data bytes that
51 follow it.
52
53 In many cases below "two-byte" data values are specified. This is in fact just
54 a default. PCRE can be compiled to use 3-byte or 4-byte values (impairing the
55 performance). This is necessary only when patterns whose compiled length is
56 greater than 64K are going to be processed. In this description, we assume the
57 "normal" compilation options.
58
59 A list of all the opcodes follows:
60
61 Opcodes with no following data
62 ------------------------------
63
64 These items are all just one byte long
65
66 OP_END end of pattern
67 OP_ANY match any character
68 OP_ANYBYTE match any single byte, even in UTF-8 mode
69 OP_SOD match start of data: \A
70 OP_SOM, start of match (subject + offset): \G
71 OP_CIRC ^ (start of data, or after \n in multiline)
72 OP_NOT_WORD_BOUNDARY \W
73 OP_WORD_BOUNDARY \w
74 OP_NOT_DIGIT \D
75 OP_DIGIT \d
76 OP_NOT_WHITESPACE \S
77 OP_WHITESPACE \s
78 OP_NOT_WORDCHAR \W
79 OP_WORDCHAR \w
80 OP_EODN match end of data or \n at end: \Z
81 OP_EOD match end of data: \z
82 OP_DOLL $ (end of data, or before \n in multiline)
83 OP_EXTUNI match an extended Unicode character
84
85
86 Repeating single characters
87 ---------------------------
88
89 The common repeats (*, +, ?) when applied to a single character use the
90 following opcodes:
91
92 OP_STAR
93 OP_MINSTAR
94 OP_PLUS
95 OP_MINPLUS
96 OP_QUERY
97 OP_MINQUERY
98
99 In ASCII mode, these are two-byte items; in UTF-8 mode, the length is variable.
100 Those with "MIN" in their name are the minimizing versions. Each is followed by
101 the character that is to be repeated. Other repeats make use of
102
103 OP_UPTO
104 OP_MINUPTO
105 OP_EXACT
106
107 which are followed by a two-byte count (most significant first) and the
108 repeated character. OP_UPTO matches from 0 to the given number. A repeat with a
109 non-zero minimum and a fixed maximum is coded as an OP_EXACT followed by an
110 OP_UPTO (or OP_MINUPTO).
111
112
113 Repeating character types
114 -------------------------
115
116 Repeats of things like \d are done exactly as for single characters, except
117 that instead of a character, the opcode for the type is stored in the data
118 byte. The opcodes are:
119
120 OP_TYPESTAR
121 OP_TYPEMINSTAR
122 OP_TYPEPLUS
123 OP_TYPEMINPLUS
124 OP_TYPEQUERY
125 OP_TYPEMINQUERY
126 OP_TYPEUPTO
127 OP_TYPEMINUPTO
128 OP_TYPEEXACT
129
130
131 Match by Unicode property
132 -------------------------
133
134 OP_PROP and OP_NOTPROP are used for positive and negative matches of a
135 character by testing its Unicode property (the \p and \P escape sequences).
136 Each is followed by a single byte that encodes the desired property value.
137
138 Repeats of these items use the OP_TYPESTAR etc. set of opcodes, followed by two
139 bytes: OP_PROP or OP_NOTPROP and then the desired property value.
140
141
142 Matching literal characters
143 ---------------------------
144
145 The OP_CHAR opcode is followed by a single character that is to be matched
146 casefully. For caseless matching, OP_CHARNC is used. In UTF-8 mode, the
147 character may be more than one byte long. (Earlier versions of PCRE used
148 multi-character strings, but this was changed to allow some new features to be
149 added.)
150
151
152 Character classes
153 -----------------
154
155 If there is only one character, OP_CHAR or OP_CHARNC is used for a positive
156 class, and OP_NOT for a negative one (that is, for something like [^a]).
157 However, in UTF-8 mode, the use of OP_NOT applies only to characters with
158 values < 128, because OP_NOT is confined to single bytes.
159
160 Another set of repeating opcodes (OP_NOTSTAR etc.) are used for a repeated,
161 negated, single-character class. The normal ones (OP_STAR etc.) are used for a
162 repeated positive single-character class.
163
164 When there's more than one character in a class and all the characters are less
165 than 256, OP_CLASS is used for a positive class, and OP_NCLASS for a negative
166 one. In either case, the opcode is followed by a 32-byte bit map containing a 1
167 bit for every character that is acceptable. The bits are counted from the least
168 significant end of each byte.
169
170 The reason for having both OP_CLASS and OP_NCLASS is so that, in UTF-8 mode,
171 subject characters with values greater than 256 can be handled correctly. For
172 OP_CLASS they don't match, whereas for OP_NCLASS they do.
173
174 For classes containing characters with values > 255, OP_XCLASS is used. It
175 optionally uses a bit map (if any characters lie within it), followed by a list
176 of pairs and single characters. There is a flag character than indicates
177 whether it's a positive or a negative class.
178
179
180 Back references
181 ---------------
182
183 OP_REF is followed by two bytes containing the reference number.
184
185
186 Repeating character classes and back references
187 -----------------------------------------------
188
189 Single-character classes are handled specially (see above). This applies to
190 OP_CLASS and OP_REF. In both cases, the repeat information follows the base
191 item. The matching code looks at the following opcode to see if it is one of
192
193 OP_CRSTAR
194 OP_CRMINSTAR
195 OP_CRPLUS
196 OP_CRMINPLUS
197 OP_CRQUERY
198 OP_CRMINQUERY
199 OP_CRRANGE
200 OP_CRMINRANGE
201
202 All but the last two are just single-byte items. The others are followed by
203 four bytes of data, comprising the minimum and maximum repeat counts.
204
205
206 Brackets and alternation
207 ------------------------
208
209 A pair of non-capturing (round) brackets is wrapped round each expression at
210 compile time, so alternation always happens in the context of brackets.
211
212 Non-capturing brackets use the opcode OP_BRA, while capturing brackets use
213 OP_BRA+1, OP_BRA+2, etc. [Note for North Americans: "bracket" to some English
214 speakers, including myself, can be round, square, curly, or pointy. Hence this
215 usage.]
216
217 Originally PCRE was limited to 99 capturing brackets (so as not to use up all
218 the opcodes). From release 3.5, there is no limit. What happens is that the
219 first ones, up to EXTRACT_BASIC_MAX are handled with separate opcodes, as
220 above. If there are more, the opcode is set to EXTRACT_BASIC_MAX+1, and the
221 first operation in the bracket is OP_BRANUMBER, followed by a 2-byte bracket
222 number. This opcode is ignored while matching, but is fished out when handling
223 the bracket itself. (They could have all been done like this, but I was making
224 minimal changes.)
225
226 A bracket opcode is followed by two bytes which give the offset to the next
227 alternative OP_ALT or, if there aren't any branches, to the matching OP_KET
228 opcode. Each OP_ALT is followed by two bytes giving the offset to the next one,
229 or to the OP_KET opcode.
230
231 OP_KET is used for subpatterns that do not repeat indefinitely, while
232 OP_KETRMIN and OP_KETRMAX are used for indefinite repetitions, minimally or
233 maximally respectively. All three are followed by two bytes giving (as a
234 positive number) the offset back to the matching OP_BRA opcode.
235
236 If a subpattern is quantified such that it is permitted to match zero times, it
237 is preceded by one of OP_BRAZERO or OP_BRAMINZERO. These are single-byte
238 opcodes which tell the matcher that skipping this subpattern entirely is a
239 valid branch.
240
241 A subpattern with an indefinite maximum repetition is replicated in the
242 compiled data its minimum number of times (or once with OP_BRAZERO if the
243 minimum is zero), with the final copy terminating with OP_KETRMIN or OP_KETRMAX
244 as appropriate.
245
246 A subpattern with a bounded maximum repetition is replicated in a nested
247 fashion up to the maximum number of times, with OP_BRAZERO or OP_BRAMINZERO
248 before each replication after the minimum, so that, for example, (abc){2,5} is
249 compiled as (abc)(abc)((abc)((abc)(abc)?)?)?.
250
251
252 Assertions
253 ----------
254
255 Forward assertions are just like other subpatterns, but starting with one of
256 the opcodes OP_ASSERT or OP_ASSERT_NOT. Backward assertions use the opcodes
257 OP_ASSERTBACK and OP_ASSERTBACK_NOT, and the first opcode inside the assertion
258 is OP_REVERSE, followed by a two byte count of the number of characters to move
259 back the pointer in the subject string. When operating in UTF-8 mode, the count
260 is a character count rather than a byte count. A separate count is present in
261 each alternative of a lookbehind assertion, allowing them to have different
262 fixed lengths.
263
264
265 Once-only subpatterns
266 ---------------------
267
268 These are also just like other subpatterns, but they start with the opcode
269 OP_ONCE.
270
271
272 Conditional subpatterns
273 -----------------------
274
275 These are like other subpatterns, but they start with the opcode OP_COND. If
276 the condition is a back reference, this is stored at the start of the
277 subpattern using the opcode OP_CREF followed by two bytes containing the
278 reference number. If the condition is "in recursion" (coded as "(?(R)"), the
279 same scheme is used, with a "reference number" of 0xffff. Otherwise, a
280 conditional subpattern always starts with one of the assertions.
281
282
283 Recursion
284 ---------
285
286 Recursion either matches the current regex, or some subexpression. The opcode
287 OP_RECURSE is followed by an value which is the offset to the starting bracket
288 from the start of the whole pattern.
289
290
291 Callout
292 -------
293
294 OP_CALLOUT is followed by one byte of data that holds a callout number in the
295 range 0 to 254 for manual callouts, or 255 for an automatic callout. In both
296 cases there follows a two-byte value giving the offset in the pattern to the
297 start of the following item, and another two-byte item giving the length of the
298 next item.
299
300
301 Changing options
302 ----------------
303
304 If any of the /i, /m, or /s options are changed within a pattern, an OP_OPT
305 opcode is compiled, followed by one byte containing the new settings of these
306 flags. If there are several alternatives, there is an occurrence of OP_OPT at
307 the start of all those following the first options change, to set appropriate
308 options for the start of the alternative. Immediately after the end of the
309 group there is another such item to reset the flags to their previous values. A
310 change of flag right at the very start of the pattern can be handled entirely
311 at compile time, and so does not cause anything to be put into the compiled
312 data.
313
314 Philip Hazel
315 September 2004

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12