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

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

Parent Directory Parent Directory | Revision Log Revision Log


Revision 91 - (hide annotations) (download)
Sat Feb 24 21:41:34 2007 UTC (7 years, 1 month ago) by nigel
File size: 14310 byte(s)
Load pcre-6.7 into code/trunk.

1 nigel 41 Technical Notes about PCRE
2     --------------------------
3    
4 nigel 91 These are very rough technical notes that record potentially useful information
5     about PCRE internals.
6    
7 nigel 75 Historical note 1
8     -----------------
9    
10 nigel 41 Many years ago I implemented some regular expression functions to an algorithm
11     suggested by Martin Richards. These were not Unix-like in form, and were quite
12     restricted in what they could do by comparison with Perl. The interesting part
13     about the algorithm was that the amount of space required to hold the compiled
14     form of an expression was known in advance. The code to apply an expression did
15 nigel 63 not operate by backtracking, as the original Henry Spencer code and current
16     Perl code does, but instead checked all possibilities simultaneously by keeping
17     a list of current states and checking all of them as it advanced through the
18 nigel 75 subject string. In the terminology of Jeffrey Friedl's book, it was a "DFA
19     algorithm". When the pattern was all used up, all remaining states were
20 nigel 63 possible matches, and the one matching the longest subset of the subject string
21     was chosen. This did not necessarily maximize the individual wild portions of
22     the pattern, as is expected in Unix and Perl-style regular expressions.
23 nigel 41
24 nigel 75 Historical note 2
25     -----------------
26    
27 nigel 91 By contrast, the code originally written by Henry Spencer (which was
28     subsequently heavily modified for Perl) compiles the expression twice: once in
29     a dummy mode in order to find out how much store will be needed, and then for
30     real. (The Perl version probably doesn't do this any more; I'm talking about
31     the original library.) The execution function operates by backtracking and
32     maximizing (or, optionally, minimizing in Perl) the amount of the subject that
33     matches individual wild portions of the pattern. This is an "NFA algorithm" in
34     Friedl's terminology.
35 nigel 41
36 nigel 75 OK, here's the real stuff
37     -------------------------
38    
39 nigel 77 For the set of functions that form the "basic" PCRE library (which are
40     unrelated to those mentioned above), I tried at first to invent an algorithm
41     that used an amount of store bounded by a multiple of the number of characters
42     in the pattern, to save on compiling time. However, because of the greater
43     complexity in Perl regular expressions, I couldn't do this. In any case, a
44     first pass through the pattern is needed, for a number of reasons. PCRE works
45     by running a very degenerate first pass to calculate a maximum store size, and
46     then a second pass to do the real compile - which may use a bit less than the
47     predicted amount of store. The idea is that this is going to turn out faster
48     because the first pass is degenerate and the second pass can just store stuff
49     straight into the vector, which it knows is big enough. It does make the
50 nigel 91 compiling functions bigger, of course, but they have become quite big anyway to
51 nigel 77 handle all the Perl stuff.
52 nigel 41
53 nigel 77 Traditional matching function
54     -----------------------------
55    
56     The "traditional", and original, matching function is called pcre_exec(), and
57     it implements an NFA algorithm, similar to the original Henry Spencer algorithm
58     and the way that Perl works. Not surprising, since it is intended to be as
59     compatible with Perl as possible. This is the function most users of PCRE will
60     use most of the time.
61    
62     Supplementary matching function
63     -------------------------------
64    
65     From PCRE 6.0, there is also a supplementary matching function called
66     pcre_dfa_exec(). This implements a DFA matching algorithm that searches
67     simultaneously for all possible matches that start at one point in the subject
68     string. (Going back to my roots: see Historical Note 1 above.) This function
69     intreprets the same compiled pattern data as pcre_exec(); however, not all the
70 nigel 91 facilities are available, and those that are do not always work in quite the
71 nigel 77 same way. See the user documentation for details.
72    
73     Format of compiled patterns
74     ---------------------------
75    
76 nigel 41 The compiled form of a pattern is a vector of bytes, containing items of
77     variable length. The first byte in an item is an opcode, and the length of the
78 nigel 75 item is either implicit in the opcode or contained in the data bytes that
79     follow it.
80 nigel 41
81 nigel 75 In many cases below "two-byte" data values are specified. This is in fact just
82     a default. PCRE can be compiled to use 3-byte or 4-byte values (impairing the
83     performance). This is necessary only when patterns whose compiled length is
84     greater than 64K are going to be processed. In this description, we assume the
85     "normal" compilation options.
86    
87     A list of all the opcodes follows:
88    
89 nigel 41 Opcodes with no following data
90     ------------------------------
91    
92     These items are all just one byte long
93    
94     OP_END end of pattern
95     OP_ANY match any character
96 nigel 75 OP_ANYBYTE match any single byte, even in UTF-8 mode
97 nigel 41 OP_SOD match start of data: \A
98 nigel 71 OP_SOM, start of match (subject + offset): \G
99 nigel 41 OP_CIRC ^ (start of data, or after \n in multiline)
100     OP_NOT_WORD_BOUNDARY \W
101     OP_WORD_BOUNDARY \w
102     OP_NOT_DIGIT \D
103     OP_DIGIT \d
104     OP_NOT_WHITESPACE \S
105     OP_WHITESPACE \s
106     OP_NOT_WORDCHAR \W
107     OP_WORDCHAR \w
108     OP_EODN match end of data or \n at end: \Z
109     OP_EOD match end of data: \z
110     OP_DOLL $ (end of data, or before \n in multiline)
111 nigel 75 OP_EXTUNI match an extended Unicode character
112    
113 nigel 41
114     Repeating single characters
115     ---------------------------
116    
117 nigel 75 The common repeats (*, +, ?) when applied to a single character use the
118     following opcodes:
119 nigel 41
120     OP_STAR
121     OP_MINSTAR
122     OP_PLUS
123     OP_MINPLUS
124     OP_QUERY
125     OP_MINQUERY
126    
127 nigel 75 In ASCII mode, these are two-byte items; in UTF-8 mode, the length is variable.
128 nigel 41 Those with "MIN" in their name are the minimizing versions. Each is followed by
129     the character that is to be repeated. Other repeats make use of
130    
131     OP_UPTO
132     OP_MINUPTO
133     OP_EXACT
134    
135     which are followed by a two-byte count (most significant first) and the
136     repeated character. OP_UPTO matches from 0 to the given number. A repeat with a
137     non-zero minimum and a fixed maximum is coded as an OP_EXACT followed by an
138     OP_UPTO (or OP_MINUPTO).
139    
140    
141     Repeating character types
142     -------------------------
143    
144     Repeats of things like \d are done exactly as for single characters, except
145     that instead of a character, the opcode for the type is stored in the data
146     byte. The opcodes are:
147    
148     OP_TYPESTAR
149     OP_TYPEMINSTAR
150     OP_TYPEPLUS
151     OP_TYPEMINPLUS
152     OP_TYPEQUERY
153     OP_TYPEMINQUERY
154     OP_TYPEUPTO
155     OP_TYPEMINUPTO
156     OP_TYPEEXACT
157    
158    
159 nigel 75 Match by Unicode property
160     -------------------------
161    
162     OP_PROP and OP_NOTPROP are used for positive and negative matches of a
163     character by testing its Unicode property (the \p and \P escape sequences).
164 nigel 91 Each is followed by two bytes that encode the desired property as a type and a
165     value.
166 nigel 75
167 nigel 91 Repeats of these items use the OP_TYPESTAR etc. set of opcodes, followed by
168     three bytes: OP_PROP or OP_NOTPROP and then the desired property type and
169     value.
170 nigel 75
171    
172     Matching literal characters
173 nigel 41 ---------------------------
174    
175 nigel 75 The OP_CHAR opcode is followed by a single character that is to be matched
176     casefully. For caseless matching, OP_CHARNC is used. In UTF-8 mode, the
177     character may be more than one byte long. (Earlier versions of PCRE used
178     multi-character strings, but this was changed to allow some new features to be
179     added.)
180 nigel 41
181    
182     Character classes
183     -----------------
184    
185 nigel 75 If there is only one character, OP_CHAR or OP_CHARNC is used for a positive
186     class, and OP_NOT for a negative one (that is, for something like [^a]).
187     However, in UTF-8 mode, the use of OP_NOT applies only to characters with
188     values < 128, because OP_NOT is confined to single bytes.
189 nigel 41
190 nigel 63 Another set of repeating opcodes (OP_NOTSTAR etc.) are used for a repeated,
191     negated, single-character class. The normal ones (OP_STAR etc.) are used for a
192     repeated positive single-character class.
193    
194 nigel 71 When there's more than one character in a class and all the characters are less
195 nigel 75 than 256, OP_CLASS is used for a positive class, and OP_NCLASS for a negative
196 nigel 71 one. In either case, the opcode is followed by a 32-byte bit map containing a 1
197     bit for every character that is acceptable. The bits are counted from the least
198     significant end of each byte.
199 nigel 41
200 nigel 75 The reason for having both OP_CLASS and OP_NCLASS is so that, in UTF-8 mode,
201     subject characters with values greater than 256 can be handled correctly. For
202 nigel 71 OP_CLASS they don't match, whereas for OP_NCLASS they do.
203    
204 nigel 63 For classes containing characters with values > 255, OP_XCLASS is used. It
205     optionally uses a bit map (if any characters lie within it), followed by a list
206 nigel 75 of pairs and single characters. There is a flag character than indicates
207 nigel 63 whether it's a positive or a negative class.
208 nigel 41
209 nigel 63
210 nigel 41 Back references
211     ---------------
212    
213 nigel 53 OP_REF is followed by two bytes containing the reference number.
214 nigel 41
215    
216     Repeating character classes and back references
217     -----------------------------------------------
218    
219     Single-character classes are handled specially (see above). This applies to
220     OP_CLASS and OP_REF. In both cases, the repeat information follows the base
221     item. The matching code looks at the following opcode to see if it is one of
222    
223     OP_CRSTAR
224     OP_CRMINSTAR
225     OP_CRPLUS
226     OP_CRMINPLUS
227     OP_CRQUERY
228     OP_CRMINQUERY
229     OP_CRRANGE
230     OP_CRMINRANGE
231    
232     All but the last two are just single-byte items. The others are followed by
233     four bytes of data, comprising the minimum and maximum repeat counts.
234    
235    
236     Brackets and alternation
237     ------------------------
238    
239 nigel 43 A pair of non-capturing (round) brackets is wrapped round each expression at
240 nigel 41 compile time, so alternation always happens in the context of brackets.
241 nigel 53
242 nigel 43 Non-capturing brackets use the opcode OP_BRA, while capturing brackets use
243 nigel 41 OP_BRA+1, OP_BRA+2, etc. [Note for North Americans: "bracket" to some English
244 nigel 43 speakers, including myself, can be round, square, curly, or pointy. Hence this
245     usage.]
246 nigel 41
247 nigel 53 Originally PCRE was limited to 99 capturing brackets (so as not to use up all
248     the opcodes). From release 3.5, there is no limit. What happens is that the
249     first ones, up to EXTRACT_BASIC_MAX are handled with separate opcodes, as
250     above. If there are more, the opcode is set to EXTRACT_BASIC_MAX+1, and the
251     first operation in the bracket is OP_BRANUMBER, followed by a 2-byte bracket
252     number. This opcode is ignored while matching, but is fished out when handling
253     the bracket itself. (They could have all been done like this, but I was making
254     minimal changes.)
255    
256 nigel 77 A bracket opcode is followed by LINK_SIZE bytes which give the offset to the
257     next alternative OP_ALT or, if there aren't any branches, to the matching
258     OP_KET opcode. Each OP_ALT is followed by LINK_SIZE bytes giving the offset to
259     the next one, or to the OP_KET opcode.
260 nigel 41
261     OP_KET is used for subpatterns that do not repeat indefinitely, while
262     OP_KETRMIN and OP_KETRMAX are used for indefinite repetitions, minimally or
263 nigel 77 maximally respectively. All three are followed by LINK_SIZE bytes giving (as a
264 nigel 75 positive number) the offset back to the matching OP_BRA opcode.
265 nigel 41
266     If a subpattern is quantified such that it is permitted to match zero times, it
267     is preceded by one of OP_BRAZERO or OP_BRAMINZERO. These are single-byte
268     opcodes which tell the matcher that skipping this subpattern entirely is a
269     valid branch.
270    
271     A subpattern with an indefinite maximum repetition is replicated in the
272 nigel 75 compiled data its minimum number of times (or once with OP_BRAZERO if the
273     minimum is zero), with the final copy terminating with OP_KETRMIN or OP_KETRMAX
274     as appropriate.
275 nigel 41
276     A subpattern with a bounded maximum repetition is replicated in a nested
277 nigel 75 fashion up to the maximum number of times, with OP_BRAZERO or OP_BRAMINZERO
278     before each replication after the minimum, so that, for example, (abc){2,5} is
279     compiled as (abc)(abc)((abc)((abc)(abc)?)?)?.
280 nigel 41
281    
282     Assertions
283     ----------
284    
285     Forward assertions are just like other subpatterns, but starting with one of
286     the opcodes OP_ASSERT or OP_ASSERT_NOT. Backward assertions use the opcodes
287     OP_ASSERTBACK and OP_ASSERTBACK_NOT, and the first opcode inside the assertion
288     is OP_REVERSE, followed by a two byte count of the number of characters to move
289 nigel 49 back the pointer in the subject string. When operating in UTF-8 mode, the count
290     is a character count rather than a byte count. A separate count is present in
291     each alternative of a lookbehind assertion, allowing them to have different
292     fixed lengths.
293 nigel 41
294    
295     Once-only subpatterns
296     ---------------------
297    
298     These are also just like other subpatterns, but they start with the opcode
299     OP_ONCE.
300    
301    
302     Conditional subpatterns
303     -----------------------
304    
305     These are like other subpatterns, but they start with the opcode OP_COND. If
306     the condition is a back reference, this is stored at the start of the
307 nigel 53 subpattern using the opcode OP_CREF followed by two bytes containing the
308 nigel 63 reference number. If the condition is "in recursion" (coded as "(?(R)"), the
309     same scheme is used, with a "reference number" of 0xffff. Otherwise, a
310     conditional subpattern always starts with one of the assertions.
311 nigel 41
312    
313 nigel 71 Recursion
314     ---------
315    
316     Recursion either matches the current regex, or some subexpression. The opcode
317     OP_RECURSE is followed by an value which is the offset to the starting bracket
318 nigel 87 from the start of the whole pattern. From release 6.5, OP_RECURSE is
319     automatically wrapped inside OP_ONCE brackets (because otherwise some patterns
320     broke it). OP_RECURSE is also used for "subroutine" calls, even though they
321     are not strictly a recursion.
322 nigel 71
323    
324     Callout
325     -------
326    
327 nigel 75 OP_CALLOUT is followed by one byte of data that holds a callout number in the
328     range 0 to 254 for manual callouts, or 255 for an automatic callout. In both
329     cases there follows a two-byte value giving the offset in the pattern to the
330     start of the following item, and another two-byte item giving the length of the
331     next item.
332 nigel 71
333    
334 nigel 41 Changing options
335     ----------------
336    
337 nigel 63 If any of the /i, /m, or /s options are changed within a pattern, an OP_OPT
338     opcode is compiled, followed by one byte containing the new settings of these
339     flags. If there are several alternatives, there is an occurrence of OP_OPT at
340     the start of all those following the first options change, to set appropriate
341     options for the start of the alternative. Immediately after the end of the
342     group there is another such item to reset the flags to their previous values. A
343     change of flag right at the very start of the pattern can be handled entirely
344     at compile time, and so does not cause anything to be put into the compiled
345     data.
346 nigel 41
347     Philip Hazel
348 nigel 91 June 2006

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12