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

Diff of /code/trunk/maint/Tech.Notes

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 41 by nigel, Sat Feb 24 21:39:17 2007 UTC revision 75 by nigel, Sat Feb 24 21:40:37 2007 UTC
# Line 1  Line 1 
1  Technical Notes about PCRE  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  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  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  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  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  form of an expression was known in advance. The code to apply an expression did
12  not operate by backtracking, as the Henry Spencer and Perl code does, but  not operate by backtracking, as the original Henry Spencer code and current
13  instead checked all possibilities simultaneously by keeping a list of current  Perl code does, but instead checked all possibilities simultaneously by keeping
14  states and checking all of them as it advanced through the subject string. (In  a list of current states and checking all of them as it advanced through the
15  the terminology of Jeffrey Friedl's book, it was a "DFA algorithm".) When the  subject string. In the terminology of Jeffrey Friedl's book, it was a "DFA
16  pattern was all used up, all remaining states were possible matches, and the  algorithm". When the pattern was all used up, all remaining states were
17  one matching the longest subset of the subject string was chosen. This did not  possible matches, and the one matching the longest subset of the subject string
18  necessarily maximize the individual wild portions of the pattern, as is  was chosen. This did not necessarily maximize the individual wild portions of
19  expected in Unix and Perl-style regular expressions.  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  By contrast, the code originally written by Henry Spencer and subsequently
25  heavily modified for Perl actually compiles the expression twice: once in a  heavily modified for Perl actually compiles the expression twice: once in a
# Line 23  optionally, minimizing in Perl) the amou Line 29  optionally, minimizing in Perl) the amou
29  individual wild portions of the pattern. This is an "NFA algorithm" in Friedl's  individual wild portions of the pattern. This is an "NFA algorithm" in Friedl's
30  terminology.  terminology.
31    
32  For this set of functions that forms PCRE, I tried at first to invent an  OK, here's the real stuff
33  algorithm that used an amount of store bounded by a multiple of the number of  -------------------------
34  characters in the pattern, to save on compiling time. However, because of the  
35  greater complexity in Perl regular expressions, I couldn't do this. In any  For the set of functions that forms PCRE (which are unrelated to those
36  case, a first pass through the pattern is needed, in order to find internal  mentioned above), I tried at first to invent an algorithm that used an amount
37  flag settings like (?i) at top level. So it works by running a very degenerate  of store bounded by a multiple of the number of characters in the pattern, to
38  first pass to calculate a maximum store size, and then a second pass to do the  save on compiling time. However, because of the greater complexity in Perl
39  real compile - which may use a bit less than the predicted amount of store. The  regular expressions, I couldn't do this. In any case, a first pass through the
40  idea is that this is going to turn out faster because the first pass is  pattern is needed, for a number of reasons. PCRE works by running a very
41  degenerate and the second can just store stuff straight into the vector. It  degenerate first pass to calculate a maximum store size, and then a second pass
42  does make the compiling functions bigger, of course, but they have got quite  to do the real compile - which may use a bit less than the predicted amount of
43  big anyway to handle all the Perl stuff.  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  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  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 which  item is either implicit in the opcode or contained in the data bytes that
51  follow it. A list of all the opcodes follows:  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  Opcodes with no following data
62  ------------------------------  ------------------------------
# Line 48  These items are all just one byte long Line 65  These items are all just one byte long
65    
66    OP_END                 end of pattern    OP_END                 end of pattern
67    OP_ANY                 match any character    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    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)    OP_CIRC                ^ (start of data, or after \n in multiline)
72    OP_NOT_WORD_BOUNDARY   \W    OP_NOT_WORD_BOUNDARY   \W
73    OP_WORD_BOUNDARY       \w    OP_WORD_BOUNDARY       \w
# Line 61  These items are all just one byte long Line 80  These items are all just one byte long
80    OP_EODN                match end of data or \n at end: \Z    OP_EODN                match end of data or \n at end: \Z
81    OP_EOD                 match end of data: \z    OP_EOD                 match end of data: \z
82    OP_DOLL                $ (end of data, or before \n in multiline)    OP_DOLL                $ (end of data, or before \n in multiline)
83      OP_EXTUNI              match an extended Unicode character
84    
85    
86  Repeating single characters  Repeating single characters
87  ---------------------------  ---------------------------
88    
89  The common repeats (*, +, ?) when applied to a single character appear as  The common repeats (*, +, ?) when applied to a single character use the
90  two-byte items using the following opcodes:  following opcodes:
91    
92    OP_STAR    OP_STAR
93    OP_MINSTAR    OP_MINSTAR
# Line 76  two-byte items using the following opcod Line 96  two-byte items using the following opcod
96    OP_QUERY    OP_QUERY
97    OP_MINQUERY    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  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  the character that is to be repeated. Other repeats make use of
102    
# Line 107  byte. The opcodes are: Line 128  byte. The opcodes are:
128    OP_TYPEEXACT    OP_TYPEEXACT
129    
130    
131  Matching a character string  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_CHARS opcode is followed by a one-byte count and then that number of  The OP_CHAR opcode is followed by a single character that is to be matched
146  characters. If there are more than 255 characters in sequence, successive  casefully. For caseless matching, OP_CHARNC is used. In UTF-8 mode, the
147  instances of OP_CHARS are used.  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  Character classes
153  -----------------  -----------------
154    
155  OP_CLASS is used for a character class, provided there are at least two  If there is only one character, OP_CHAR or OP_CHARNC is used for a positive
156  characters in the class. If there is only one character, OP_CHARS is used for a  class, and OP_NOT for a negative one (that is, for something like [^a]).
157  positive class, and OP_NOT for a negative one (that is, for something like  However, in UTF-8 mode, the use of OP_NOT applies only to characters with
158  [^a]). Another set of repeating opcodes (OP_NOTSTAR etc.) are used for a  values < 128, because OP_NOT is confined to single bytes.
159  repeated, negated, single-character class. The normal ones (OP_STAR etc.) are  
160  used for a repeated positive single-character class.  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  OP_CLASS is followed by a 32-byte bit map containing a 1  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  bit for every character that is acceptable. The bits are counted from the least
168  significant end of each byte.  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  Back references
181  ---------------  ---------------
182    
183  OP_REF is followed by a single byte containing the reference number.  OP_REF is followed by two bytes containing the reference number.
184    
185    
186  Repeating character classes and back references  Repeating character classes and back references
# Line 159  four bytes of data, comprising the minim Line 206  four bytes of data, comprising the minim
206  Brackets and alternation  Brackets and alternation
207  ------------------------  ------------------------
208    
209  A pair of non-identifying (round) brackets is wrapped round each expression at  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.  compile time, so alternation always happens in the context of brackets.
211  Non-identifying brackets use the opcode OP_BRA, while identifying brackets use  
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  OP_BRA+1, OP_BRA+2, etc. [Note for North Americans: "bracket" to some English
214  speakers, including myself, can be round, square, or curly. Hence this usage.]  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  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 KET  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,  opcode. Each OP_ALT is followed by two bytes giving the offset to the next one,
229  or to the KET opcode.  or to the OP_KET opcode.
230    
231  OP_KET is used for subpatterns that do not repeat indefinitely, while  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  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  maximally respectively. All three are followed by two bytes giving (as a
234  positive number) the offset back to the matching BRA opcode.  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  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  is preceded by one of OP_BRAZERO or OP_BRAMINZERO. These are single-byte
# Line 181  opcodes which tell the matcher that skip Line 239  opcodes which tell the matcher that skip
239  valid branch.  valid branch.
240    
241  A subpattern with an indefinite maximum repetition is replicated in the  A subpattern with an indefinite maximum repetition is replicated in the
242  compiled data its minimum number of times (or once with a BRAZERO if the  compiled data its minimum number of times (or once with OP_BRAZERO if the
243  minimum is zero), with the final copy terminating with a KETRMIN or KETRMAX as  minimum is zero), with the final copy terminating with OP_KETRMIN or OP_KETRMAX
244  appropriate.  as appropriate.
245    
246  A subpattern with a bounded maximum repetition is replicated in a nested  A subpattern with a bounded maximum repetition is replicated in a nested
247  fashion up to the maximum number of times, with BRAZERO or BRAMINZERO before  fashion up to the maximum number of times, with OP_BRAZERO or OP_BRAMINZERO
248  each replication after the minimum, so that, for example, (abc){2,5} is  before each replication after the minimum, so that, for example, (abc){2,5} is
249  compiled as (abc)(abc)((abc)((abc)(abc)?)?)?. The 200-bracket limit does not  compiled as (abc)(abc)((abc)((abc)(abc)?)?)?.
 apply to these internally generated brackets.  
250    
251    
252  Assertions  Assertions
# Line 199  Forward assertions are just like other s Line 256  Forward assertions are just like other s
256  the opcodes OP_ASSERT or OP_ASSERT_NOT. Backward assertions use the opcodes  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  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  is OP_REVERSE, followed by a two byte count of the number of characters to move
259  back the pointer in the subject string. A separate count is present in each  back the pointer in the subject string. When operating in UTF-8 mode, the count
260  alternative of a lookbehind assertion, allowing them to have different fixed  is a character count rather than a byte count. A separate count is present in
261  lengths.  each alternative of a lookbehind assertion, allowing them to have different
262    fixed lengths.
263    
264    
265  Once-only subpatterns  Once-only subpatterns
# Line 216  Conditional subpatterns Line 274  Conditional subpatterns
274    
275  These are like other subpatterns, but they start with the opcode OP_COND. If  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  the condition is a back reference, this is stored at the start of the
277  subpattern using the opcode OP_CREF followed by one byte containing the  subpattern using the opcode OP_CREF followed by two bytes containing the
278  reference number. Otherwise, a conditional subpattern will always start with  reference number. If the condition is "in recursion" (coded as "(?(R)"), the
279  one of the assertions.  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  Changing options
302  ----------------  ----------------
303    
304  If any of the /i, /m, or /s options are changed within a parenthesized group,  If any of the /i, /m, or /s options are changed within a pattern, an OP_OPT
305  an OP_OPT opcode is compiled, followed by one byte containing the new settings  opcode is compiled, followed by one byte containing the new settings of these
306  of these flags. If there are several alternatives in a group, there is an  flags. If there are several alternatives, there is an occurrence of OP_OPT at
307  occurrence of OP_OPT at the start of all those following the first options  the start of all those following the first options change, to set appropriate
308  change, to set appropriate options for the start of the alternative.  options for the start of the alternative. Immediately after the end of the
309  Immediately after the end of the group there is another such item to reset the  group there is another such item to reset the flags to their previous values. A
310  flags to their previous values. Other changes of flag within the pattern can be  change of flag right at the very start of the pattern can be handled entirely
311  handled entirely at compile time, and so do not cause anything to be put into  at compile time, and so does not cause anything to be put into the compiled
312  the compiled data.  data.
   
313    
314  Philip Hazel  Philip Hazel
315  January 1999  September 2004

Legend:
Removed from v.41  
changed lines
  Added in v.75

  ViewVC Help
Powered by ViewVC 1.1.5