# Contents of /code/trunk/doc/html/pcrematching.html

Revision 77 - (show annotations)
Sat Feb 24 21:40:45 2007 UTC (13 years ago) by nigel
File MIME type: text/html
File size: 8358 byte(s)
```Load pcre-6.0 into code/trunk.
```
 1 2 3 pcrematching specification 4 5 6

pcrematching man page

7

10

11 This page is part of the PCRE HTML documentation. It was generated automatically 12 from the original man page. If there is any nonsense in it, please consult the 13 man page, in case the conversion went wrong. 14
15

23
PCRE MATCHING ALGORITHMS
24

25 This document describes the two different algorithms that are available in PCRE 26 for matching a compiled regular expression against a given subject string. The 27 "standard" algorithm is the one provided by the pcre_exec() function. 28 This works in the same was as Perl's matching function, and provides a 29 Perl-compatible matching operation. 30

31

32 An alternative algorithm is provided by the pcre_dfa_exec() function; 33 this operates in a different way, and is not Perl-compatible. It has advantages 34 and disadvantages compared with the standard algorithm, and these are described 35 below. 36

37

38 When there is only one possible way in which a given subject string can match a 39 pattern, the two algorithms give the same answer. A difference arises, however, 40 when there are multiple possibilities. For example, if the pattern 41

42  ^<.*>
43
44 is matched against the string 45
46  <something> <something else> <something further>
47
48 there are three possible answers. The standard algorithm finds only one of 49 them, whereas the DFA algorithm finds all three. 50

51
REGULAR EXPRESSIONS AS TREES
52

53 The set of strings that are matched by a regular expression can be represented 54 as a tree structure. An unlimited repetition in the pattern makes the tree of 55 infinite size, but it is still a tree. Matching the pattern to a given subject 56 string (from a given starting point) can be thought of as a search of the tree. 57 There are two standard ways to search a tree: depth-first and breadth-first, 58 and these correspond to the two matching algorithms provided by PCRE. 59

60
THE STANDARD MATCHING ALGORITHM
61

62 In the terminology of Jeffrey Friedl's book \fIMastering Regular 63 Expressions\fP, the standard algorithm is an "NFA algorithm". It conducts a 64 depth-first search of the pattern tree. That is, it proceeds along a single 65 path through the tree, checking that the subject matches what is required. When 66 there is a mismatch, the algorithm tries any alternatives at the current point, 67 and if they all fail, it backs up to the previous branch point in the tree, and 68 tries the next alternative branch at that level. This often involves backing up 69 (moving to the left) in the subject string as well. The order in which 70 repetition branches are tried is controlled by the greedy or ungreedy nature of 71 the quantifier. 72

73

74 If a leaf node is reached, a matching string has been found, and at that point 75 the algorithm stops. Thus, if there is more than one possible match, this 76 algorithm returns the first one that it finds. Whether this is the shortest, 77 the longest, or some intermediate length depends on the way the greedy and 78 ungreedy repetition quantifiers are specified in the pattern. 79

80

81 Because it ends up with a single path through the tree, it is relatively 82 straightforward for this algorithm to keep track of the substrings that are 83 matched by portions of the pattern in parentheses. This provides support for 84 capturing parentheses and back references. 85

86
THE DFA MATCHING ALGORITHM
87

88 DFA stands for "deterministic finite automaton", but you do not need to 89 understand the origins of that name. This algorithm conducts a breadth-first 90 search of the tree. Starting from the first matching point in the subject, it 91 scans the subject string from left to right, once, character by character, and 92 as it does this, it remembers all the paths through the tree that represent 93 valid matches. 94

95

96 The scan continues until either the end of the subject is reached, or there are 97 no more unterminated paths. At this point, terminated paths represent the 98 different matching possibilities (if there are none, the match has failed). 99 Thus, if there is more than one possible match, this algorithm finds all of 100 them, and in particular, it finds the longest. In PCRE, there is an option to 101 stop the algorithm after the first match (which is necessarily the shortest) 102 has been found. 103

104

105 Note that all the matches that are found start at the same point in the 106 subject. If the pattern 107

108  cat(er(pillar)?)
109
110 is matched against the string "the caterpillar catchment", the result will be 111 the three strings "cat", "cater", and "caterpillar" that start at the fourth 112 character of the subject. The algorithm does not automatically move on to find 113 matches that start at later positions. 114

115

116 There are a number of features of PCRE regular expressions that are not 117 supported by the DFA matching algorithm. They are as follows: 118

119

120 1. Because the algorithm finds all possible matches, the greedy or ungreedy 121 nature of repetition quantifiers is not relevant. Greedy and ungreedy 122 quantifiers are treated in exactly the same way. 123

124

125 2. When dealing with multiple paths through the tree simultaneously, it is not 126 straightforward to keep track of captured substrings for the different matching 127 possibilities, and PCRE's implementation of this algorithm does not attempt to 128 do this. This means that no captured substrings are available. 129

130

131 3. Because no substrings are captured, back references within the pattern are 132 not supported, and cause errors if encountered. 133

134

135 4. For the same reason, conditional expressions that use a backreference as the 136 condition are not supported. 137

138

139 5. Callouts are supported, but the value of the capture_top field is 140 always 1, and the value of the capture_last field is always -1. 141

142

143 6. 144 The \C escape sequence, which (in the standard algorithm) matches a single 145 byte, even in UTF-8 mode, is not supported because the DFA algorithm moves 146 through the subject string one character at a time, for all active paths 147 through the tree. 148

149
150

151 Using the DFA matching algorithm provides the following advantages: 152

153

154 1. All possible matches (at a single point in the subject) are automatically 155 found, and in particular, the longest match is found. To find more than one 156 match using the standard algorithm, you have to do kludgy things with 157 callouts. 158

159

160 2. There is much better support for partial matching. The restrictions on the 161 content of the pattern that apply when using the standard algorithm for partial 162 matching do not apply to the DFA algorithm. For non-anchored patterns, the 163 starting position of a partial match is available. 164

165

166 3. Because the DFA algorithm scans the subject string just once, and never 167 needs to backtrack, it is possible to pass very long subject strings to the 168 matching function in several pieces, checking for partial matching each time. 169

170
171

172 The DFA algorithm suffers from a number of disadvantages: 173

174

175 1. It is substantially slower than the standard algorithm. This is partly 176 because it has to search for all possible matches, but is also because it is 177 less susceptible to optimization. 178

179

180 2. Capturing parentheses and back references are not supported. 181

182

183 3. The "atomic group" feature of PCRE regular expressions is supported, but 184 does not provide the advantage that it does for the standard algorithm. 185

186

187 Last updated: 28 February 2005 188