1 PCRETEST(1) PCRETEST(1)
5 pcretest - a program for testing Perl-compatible regular expressions.
10 pcretest [options] [source] [destination]
12 pcretest was written as a test program for the PCRE regular expression
13 library itself, but it can also be used for experimenting with regular
14 expressions. This document describes the features of the test program;
15 for details of the regular expressions themselves, see the pcrepattern
16 documentation. For details of the PCRE library function calls and their
17 options, see the pcreapi documentation.
22 -b Behave as if each regex has the /B (show bytecode) modifier;
23 the internal form is output after compilation.
25 -C Output the version number of the PCRE library, and all avail-
26 able information about the optional features that are
27 included, and then exit.
29 -d Behave as if each regex has the /D (debug) modifier; the
30 internal form and information about the compiled pattern is
31 output after compilation; -d is equivalent to -b -i.
33 -dfa Behave as if each data line contains the \D escape sequence;
34 this causes the alternative matching function,
35 pcre_dfa_exec(), to be used instead of the standard
36 pcre_exec() function (more detail is given below).
38 -help Output a brief summary these options and then exit.
40 -i Behave as if each regex has the /I modifier; information
41 about the compiled pattern is given after compilation.
43 -M Behave as if each data line contains the \M escape sequence;
44 this causes PCRE to discover the minimum MATCH_LIMIT and
45 MATCH_LIMIT_RECURSION settings by calling pcre_exec() repeat-
46 edly with different limits.
48 -m Output the size of each compiled pattern after it has been
49 compiled. This is equivalent to adding /M to each regular
50 expression. For compatibility with earlier versions of
51 pcretest, -s is a synonym for -m.
53 -o osize Set the number of elements in the output vector that is used
54 when calling pcre_exec() or pcre_dfa_exec() to be osize. The
55 default value is 45, which is enough for 14 capturing subex-
56 pressions for pcre_exec() or 22 different matches for
57 pcre_dfa_exec(). The vector size can be changed for individ-
58 ual matching calls by including \O in the data line (see
61 -p Behave as if each regex has the /P modifier; the POSIX wrap-
62 per API is used to call PCRE. None of the other options has
63 any effect when -p is set.
65 -q Do not output the version number of pcretest at the start of
68 -S size On Unix-like systems, set the size of the runtime stack to
71 -t Run each compile, study, and match many times with a timer,
72 and output resulting time per compile or match (in millisec-
73 onds). Do not set -m with -t, because you will then get the
74 size output a zillion times, and the timing will be dis-
75 torted. You can control the number of iterations that are
76 used for timing by following -t with a number (as a separate
77 item on the command line). For example, "-t 1000" would iter-
78 ate 1000 times. The default is to iterate 500000 times.
80 -tm This is like -t except that it times only the matching phase,
81 not the compile or study phases.
86 If pcretest is given two filename arguments, it reads from the first
87 and writes to the second. If it is given only one filename argument, it
88 reads from that file and writes to stdout. Otherwise, it reads from
89 stdin and writes to stdout, and prompts for each line of input, using
90 "re>" to prompt for regular expressions, and "data>" to prompt for data
93 When pcretest is built, a configuration option can specify that it
94 should be linked with the libreadline library. When this is done, if
95 the input is from a terminal, it is read using the readline() function.
96 This provides line-editing and history facilities. The output from the
97 -help option states whether or not readline() will be used.
99 The program handles any number of sets of input on a single input file.
100 Each set starts with a regular expression, and continues with any num-
101 ber of data lines to be matched against the pattern.
103 Each data line is matched separately and independently. If you want to
104 do multi-line matches, you have to use the \n escape sequence (or \r or
105 \r\n, etc., depending on the newline setting) in a single line of input
106 to encode the newline sequences. There is no limit on the length of
107 data lines; the input buffer is automatically extended if it is too
110 An empty line signals the end of the data lines, at which point a new
111 regular expression is read. The regular expressions are given enclosed
112 in any non-alphanumeric delimiters other than backslash, for example:
116 White space before the initial delimiter is ignored. A regular expres-
117 sion may be continued over several input lines, in which case the new-
118 line characters are included within it. It is possible to include the
119 delimiter within the pattern by escaping it, for example
123 If you do so, the escape and the delimiter form part of the pattern,
124 but since delimiters are always non-alphanumeric, this does not affect
125 its interpretation. If the terminating delimiter is immediately fol-
126 lowed by a backslash, for example,
130 then a backslash is added to the end of the pattern. This is done to
131 provide a way of testing the error condition that arises if a pattern
132 finishes with a backslash, because
136 is interpreted as the first line of a pattern that starts with "abc/",
137 causing pcretest to read the next line as a continuation of the regular
143 A pattern may be followed by any number of modifiers, which are mostly
144 single characters. Following Perl usage, these are referred to below
145 as, for example, "the /i modifier", even though the delimiter of the
146 pattern need not always be a slash, and no slash is used when writing
147 modifiers. Whitespace may appear between the final pattern delimiter
148 and the first modifier, and between the modifiers themselves.
150 The /i, /m, /s, and /x modifiers set the PCRE_CASELESS, PCRE_MULTILINE,
151 PCRE_DOTALL, or PCRE_EXTENDED options, respectively, when pcre_com-
152 pile() is called. These four modifier letters have the same effect as
153 they do in Perl. For example:
157 The following table shows additional modifiers for setting PCRE com-
158 pile-time options that do not correspond to anything in Perl:
161 /? PCRE_NO_UTF8_CHECK
164 /E PCRE_DOLLAR_ENDONLY
167 /N PCRE_NO_AUTO_CAPTURE
171 /Y PCRE_NO_START_OPTIMIZE
172 /<JS> PCRE_JAVASCRIPT_COMPAT
173 /<cr> PCRE_NEWLINE_CR
174 /<lf> PCRE_NEWLINE_LF
175 /<crlf> PCRE_NEWLINE_CRLF
176 /<anycrlf> PCRE_NEWLINE_ANYCRLF
177 /<any> PCRE_NEWLINE_ANY
178 /<bsr_anycrlf> PCRE_BSR_ANYCRLF
179 /<bsr_unicode> PCRE_BSR_UNICODE
181 The modifiers that are enclosed in angle brackets are literal strings
182 as shown, including the angle brackets, but the letters can be in
183 either case. This example sets multiline matching with CRLF as the line
188 As well as turning on the PCRE_UTF8 option, the /8 modifier also causes
189 any non-printing characters in output strings to be printed using the
190 \x{hh...} notation if they are valid UTF-8 sequences. Full details of
191 the PCRE options are given in the pcreapi documentation.
193 Finding all matches in a string
195 Searching for all possible matches within each subject string can be
196 requested by the /g or /G modifier. After finding a match, PCRE is
197 called again to search the remainder of the subject string. The differ-
198 ence between /g and /G is that the former uses the startoffset argument
199 to pcre_exec() to start searching at a new point within the entire
200 string (which is in effect what Perl does), whereas the latter passes
201 over a shortened substring. This makes a difference to the matching
202 process if the pattern begins with a lookbehind assertion (including \b
205 If any call to pcre_exec() in a /g or /G sequence matches an empty
206 string, the next call is done with the PCRE_NOTEMPTY_ATSTART and
207 PCRE_ANCHORED flags set in order to search for another, non-empty,
208 match at the same point. If this second match fails, the start offset
209 is advanced, and the normal match is retried. This imitates the way
210 Perl handles such cases when using the /g modifier or the split() func-
211 tion. Normally, the start offset is advanced by one character, but if
212 the newline convention recognizes CRLF as a newline, and the current
213 character is CR followed by LF, an advance of two is used.
217 There are yet more modifiers for controlling the way pcretest operates.
219 The /+ modifier requests that as well as outputting the substring that
220 matched the entire pattern, pcretest should in addition output the
221 remainder of the subject string. This is useful for tests where the
222 subject contains multiple copies of the same substring.
224 The /B modifier is a debugging feature. It requests that pcretest out-
225 put a representation of the compiled byte code after compilation. Nor-
226 mally this information contains length and offset values; however, if
227 /Z is also present, this data is replaced by spaces. This is a special
228 feature for use in the automatic test scripts; it ensures that the same
229 output is generated for different internal link sizes.
231 The /D modifier is a PCRE debugging feature, and is equivalent to /BI,
232 that is, both the /B and the /I modifiers.
234 The /F modifier causes pcretest to flip the byte order of the fields in
235 the compiled pattern that contain 2-byte and 4-byte numbers. This
236 facility is for testing the feature in PCRE that allows it to execute
237 patterns that were compiled on a host with a different endianness. This
238 feature is not available when the POSIX interface to PCRE is being
239 used, that is, when the /P pattern modifier is specified. See also the
240 section about saving and reloading compiled patterns below.
242 The /I modifier requests that pcretest output information about the
243 compiled pattern (whether it is anchored, has a fixed first character,
244 and so on). It does this by calling pcre_fullinfo() after compiling a
245 pattern. If the pattern is studied, the results of that are also out-
248 The /K modifier requests pcretest to show names from backtracking con-
249 trol verbs that are returned from calls to pcre_exec(). It causes
250 pcretest to create a pcre_extra block if one has not already been cre-
251 ated by a call to pcre_study(), and to set the PCRE_EXTRA_MARK flag and
252 the mark field within it, every time that pcre_exec() is called. If the
253 variable that the mark field points to is non-NULL for a match, non-
254 match, or partial match, pcretest prints the string to which it points.
255 For a match, this is shown on a line by itself, tagged with "MK:". For
256 a non-match it is added to the message.
258 The /L modifier must be followed directly by the name of a locale, for
263 For this reason, it must be the last modifier. The given locale is set,
264 pcre_maketables() is called to build a set of character tables for the
265 locale, and this is then passed to pcre_compile() when compiling the
266 regular expression. Without an /L (or /T) modifier, NULL is passed as
267 the tables pointer; that is, /L applies only to the expression on which
270 The /M modifier causes the size of memory block used to hold the com-
271 piled pattern to be output.
273 The /S modifier causes pcre_study() to be called after the expression
274 has been compiled, and the results used when the expression is matched.
276 The /T modifier must be followed by a single digit. It causes a spe-
277 cific set of built-in character tables to be passed to pcre_compile().
278 It is used in the standard PCRE tests to check behaviour with different
279 character tables. The digit specifies the tables as follows:
281 0 the default ASCII tables, as distributed in
282 pcre_chartables.c.dist
283 1 a set of tables defining ISO 8859 characters
285 In table 1, some characters whose codes are greater than 128 are iden-
286 tified as letters, digits, spaces, etc.
288 Using the POSIX wrapper API
290 The /P modifier causes pcretest to call PCRE via the POSIX wrapper API
291 rather than its native API. When /P is set, the following modifiers set
292 options for the regcomp() function:
298 /U REG_UNGREEDY ) These options are not part of
299 /W REG_UCP ) the POSIX standard
302 The /+ modifier works as described above. All other modifiers are
308 Before each data line is passed to pcre_exec(), leading and trailing
309 whitespace is removed, and it is then scanned for \ escapes. Some of
310 these are pretty esoteric features, intended for checking out some of
311 the more complicated features of PCRE. If you are just testing "ordi-
312 nary" regular expressions, you probably don't need any of these. The
313 following escapes are recognized:
320 \qdd set the PCRE_MATCH_LIMIT limit to dd
321 (any number of digits)
322 \r carriage return (\x0d)
324 \v vertical tab (\x0b)
325 \nnn octal character (up to 3 octal digits)
326 always a byte unless > 255 in UTF-8 mode
327 \xhh hexadecimal byte (up to 2 hex digits)
328 \x{hh...} hexadecimal character, any number of digits
330 \A pass the PCRE_ANCHORED option to pcre_exec()
332 \B pass the PCRE_NOTBOL option to pcre_exec()
334 \Cdd call pcre_copy_substring() for substring dd
335 after a successful match (number less than 32)
336 \Cname call pcre_copy_named_substring() for substring
337 "name" after a successful match (name termin-
338 ated by next non alphanumeric character)
339 \C+ show the current captured substrings at callout
341 \C- do not supply a callout function
342 \C!n return 1 instead of 0 when callout number n is
344 \C!n!m return 1 instead of 0 when callout number n is
345 reached for the nth time
346 \C*n pass the number n (may be negative) as callout
347 data; this is used as the callout return value
348 \D use the pcre_dfa_exec() match function
349 \F only shortest match for pcre_dfa_exec()
350 \Gdd call pcre_get_substring() for substring dd
351 after a successful match (number less than 32)
352 \Gname call pcre_get_named_substring() for substring
353 "name" after a successful match (name termin-
354 ated by next non-alphanumeric character)
355 \L call pcre_get_substringlist() after a
357 \M discover the minimum MATCH_LIMIT and
358 MATCH_LIMIT_RECURSION settings
359 \N pass the PCRE_NOTEMPTY option to pcre_exec()
360 or pcre_dfa_exec(); if used twice, pass the
361 PCRE_NOTEMPTY_ATSTART option
362 \Odd set the size of the output vector passed to
363 pcre_exec() to dd (any number of digits)
364 \P pass the PCRE_PARTIAL_SOFT option to pcre_exec()
365 or pcre_dfa_exec(); if used twice, pass the
366 PCRE_PARTIAL_HARD option
367 \Qdd set the PCRE_MATCH_LIMIT_RECURSION limit to dd
368 (any number of digits)
369 \R pass the PCRE_DFA_RESTART option to pcre_dfa_exec()
370 \S output details of memory get/free calls during matching
371 \Y pass the PCRE_NO_START_OPTIMIZE option to pcre_exec()
373 \Z pass the PCRE_NOTEOL option to pcre_exec()
375 \? pass the PCRE_NO_UTF8_CHECK option to
376 pcre_exec() or pcre_dfa_exec()
377 \>dd start the match at offset dd (optional "-"; then
378 any number of digits); this sets the startoffset
379 argument for pcre_exec() or pcre_dfa_exec()
380 \<cr> pass the PCRE_NEWLINE_CR option to pcre_exec()
382 \<lf> pass the PCRE_NEWLINE_LF option to pcre_exec()
384 \<crlf> pass the PCRE_NEWLINE_CRLF option to pcre_exec()
386 \<anycrlf> pass the PCRE_NEWLINE_ANYCRLF option to pcre_exec()
388 \<any> pass the PCRE_NEWLINE_ANY option to pcre_exec()
391 Note that \xhh always specifies one byte, even in UTF-8 mode; this
392 makes it possible to construct invalid UTF-8 sequences for testing pur-
393 poses. On the other hand, \x{hh} is interpreted as a UTF-8 character in
394 UTF-8 mode, generating more than one byte if the value is greater than
395 127. When not in UTF-8 mode, it generates one byte for values less than
396 256, and causes an error for greater values.
398 The escapes that specify line ending sequences are literal strings,
399 exactly as shown. No more than one newline setting should be present in
402 A backslash followed by anything else just escapes the anything else.
403 If the very last character is a backslash, it is ignored. This gives a
404 way of passing an empty line as data, since a real empty line termi-
405 nates the data input.
407 If \M is present, pcretest calls pcre_exec() several times, with dif-
408 ferent values in the match_limit and match_limit_recursion fields of
409 the pcre_extra data structure, until it finds the minimum numbers for
410 each parameter that allow pcre_exec() to complete. The match_limit num-
411 ber is a measure of the amount of backtracking that takes place, and
412 checking it out can be instructive. For most simple matches, the number
413 is quite small, but for patterns with very large numbers of matching
414 possibilities, it can become large very quickly with increasing length
415 of subject string. The match_limit_recursion number is a measure of how
416 much stack (or, if PCRE is compiled with NO_RECURSE, how much heap)
417 memory is needed to complete the match attempt.
419 When \O is used, the value specified may be higher or lower than the
420 size set by the -O command line option (or defaulted to 45); \O applies
421 only to the call of pcre_exec() for the line in which it appears.
423 If the /P modifier was present on the pattern, causing the POSIX wrap-
424 per API to be used, the only option-setting sequences that have any
425 effect are \B, \N, and \Z, causing REG_NOTBOL, REG_NOTEMPTY, and
426 REG_NOTEOL, respectively, to be passed to regexec().
428 The use of \x{hh...} to represent UTF-8 characters is not dependent on
429 the use of the /8 modifier on the pattern. It is recognized always.
430 There may be any number of hexadecimal digits inside the braces. The
431 result is from one to six bytes, encoded according to the original
432 UTF-8 rules of RFC 2279. This allows for values in the range 0 to
433 0x7FFFFFFF. Note that not all of those are valid Unicode code points,
434 or indeed valid UTF-8 characters according to the later rules in RFC
438 THE ALTERNATIVE MATCHING FUNCTION
440 By default, pcretest uses the standard PCRE matching function,
441 pcre_exec() to match each data line. From release 6.0, PCRE supports an
442 alternative matching function, pcre_dfa_test(), which operates in a
443 different way, and has some restrictions. The differences between the
444 two functions are described in the pcrematching documentation.
446 If a data line contains the \D escape sequence, or if the command line
447 contains the -dfa option, the alternative matching function is called.
448 This function finds all possible matches at a given point. If, however,
449 the \F escape sequence is present in the data line, it stops after the
450 first match is found. This is always the shortest possible match.
453 DEFAULT OUTPUT FROM PCRETEST
455 This section describes the output when the normal matching function,
456 pcre_exec(), is being used.
458 When a match succeeds, pcretest outputs the list of captured substrings
459 that pcre_exec() returns, starting with number 0 for the string that
460 matched the whole pattern. Otherwise, it outputs "No match" when the
461 return is PCRE_ERROR_NOMATCH, and "Partial match:" followed by the par-
462 tially matching substring when pcre_exec() returns PCRE_ERROR_PARTIAL.
463 (Note that this is the entire substring that was inspected during the
464 partial match; it may include characters before the actual match start
465 if a lookbehind assertion, \K, \b, or \B was involved.) For any other
466 returns, it outputs the PCRE negative error number. Here is an example
467 of an interactive pcretest run.
470 PCRE version 7.0 30-Nov-2006
479 Note that unset capturing substrings that are not followed by one that
480 is set are not returned by pcre_exec(), and are not shown by pcretest.
481 In the following example, there are two capturing substrings, but when
482 the first data line is matched, the second, unset substring is not
483 shown. An "internal" unset substring is shown as "<unset>", as for the
495 If the strings contain any non-printing characters, they are output as
496 \0x escapes, or as \x{...} escapes if the /8 modifier was present on
497 the pattern. See below for the definition of non-printing characters.
498 If the pattern has the /+ modifier, the output for substring 0 is fol-
499 lowed by the the rest of the subject string, identified by "0+" like
507 If the pattern has the /g or /G modifier, the results of successive
508 matching attempts are output in sequence, like this:
519 "No match" is output only if the first match attempt fails.
521 If any of the sequences \C, \G, or \L are present in a data line that
522 is successfully matched, the substrings extracted by the convenience
523 functions are output with C, G, or L after the string number instead of
524 a colon. This is in addition to the normal full list. The string length
525 (that is, the return from the extraction function) is given in paren-
526 theses after each string for \C and \G.
528 Note that whereas patterns can be continued over several lines (a plain
529 ">" prompt is used for continuations), data lines may not. However new-
530 lines can be included in data by means of the \n escape (or \r, \r\n,
531 etc., depending on the newline sequence setting).
534 OUTPUT FROM THE ALTERNATIVE MATCHING FUNCTION
536 When the alternative matching function, pcre_dfa_exec(), is used (by
537 means of the \D escape sequence or the -dfa command line option), the
538 output consists of a list of all the matches that start at the first
539 point in the subject where there is at least one match. For example:
541 re> /(tang|tangerine|tan)/
542 data> yellow tangerine\D
547 (Using the normal matching function on this data finds only "tang".)
548 The longest matching string is always given first (and numbered zero).
549 After a PCRE_ERROR_PARTIAL return, the output is "Partial match:", fol-
550 lowed by the partially matching substring. (Note that this is the
551 entire substring that was inspected during the partial match; it may
552 include characters before the actual match start if a lookbehind asser-
553 tion, \K, \b, or \B was involved.)
555 If /g is present on the pattern, the search for further matches resumes
556 at the end of the longest match. For example:
558 re> /(tang|tangerine|tan)/g
559 data> yellow tangerine and tangy sultana\D
567 Since the matching function does not support substring capture, the
568 escape sequences that are concerned with captured substrings are not
572 RESTARTING AFTER A PARTIAL MATCH
574 When the alternative matching function has given the PCRE_ERROR_PARTIAL
575 return, indicating that the subject partially matched the pattern, you
576 can restart the match with additional subject data by means of the \R
577 escape sequence. For example:
579 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
585 For further information about partial matching, see the pcrepartial
591 If the pattern contains any callout requests, pcretest's callout func-
592 tion is called during matching. This works with both matching func-
593 tions. By default, the called function displays the callout number, the
594 start and current positions in the text at the callout time, and the
595 next pattern item to be tested. For example, the output
600 indicates that callout number 0 occurred for a match attempt starting
601 at the fourth character of the subject string, when the pointer was at
602 the seventh character of the data, and when the next pattern item was
603 \d. Just one circumflex is output if the start and current positions
606 Callouts numbered 255 are assumed to be automatic callouts, inserted as
607 a result of the /C pattern modifier. In this case, instead of showing
608 the callout number, the offset in the pattern, preceded by a plus, is
620 The callout function in pcretest returns zero (carry on matching) by
621 default, but you can use a \C item in a data line (as described above)
624 Inserting callouts can be helpful when using pcretest to check compli-
625 cated regular expressions. For further information about callouts, see
626 the pcrecallout documentation.
629 NON-PRINTING CHARACTERS
631 When pcretest is outputting text in the compiled version of a pattern,
632 bytes other than 32-126 are always treated as non-printing characters
633 are are therefore shown as hex escapes.
635 When pcretest is outputting text that is a matched part of a subject
636 string, it behaves in the same way, unless a different locale has been
637 set for the pattern (using the /L modifier). In this case, the
638 isprint() function to distinguish printing and non-printing characters.
641 SAVING AND RELOADING COMPILED PATTERNS
643 The facilities described in this section are not available when the
644 POSIX inteface to PCRE is being used, that is, when the /P pattern mod-
647 When the POSIX interface is not in use, you can cause pcretest to write
648 a compiled pattern to a file, by following the modifiers with > and a
649 file name. For example:
651 /pattern/im >/some/file
653 See the pcreprecompile documentation for a discussion about saving and
654 re-using compiled patterns.
656 The data that is written is binary. The first eight bytes are the
657 length of the compiled pattern data followed by the length of the
658 optional study data, each written as four bytes in big-endian order
659 (most significant byte first). If there is no study data (either the
660 pattern was not studied, or studying did not return any data), the sec-
661 ond length is zero. The lengths are followed by an exact copy of the
662 compiled pattern. If there is additional study data, this follows imme-
663 diately after the compiled pattern. After writing the file, pcretest
664 expects to read a new pattern.
666 A saved pattern can be reloaded into pcretest by specifing < and a file
667 name instead of a pattern. The name of the file must not contain a <
668 character, as otherwise pcretest will interpret the line as a pattern
669 delimited by < characters. For example:
672 Compiled regex loaded from /some/file
675 When the pattern has been loaded, pcretest proceeds to read data lines
678 You can copy a file written by pcretest to a different host and reload
679 it there, even if the new host has opposite endianness to the one on
680 which the pattern was compiled. For example, you can compile on an i86
681 machine and run on a SPARC machine.
683 File names for saving and reloading can be absolute or relative, but
684 note that the shell facility of expanding a file name that starts with
685 a tilde (~) is not available.
687 The ability to save and reload files in pcretest is intended for test-
688 ing and experimentation. It is not intended for production use because
689 only a single pattern can be written to a file. Furthermore, there is
690 no facility for supplying custom character tables for use with a
691 reloaded pattern. If the original pattern was compiled with custom
692 tables, an attempt to match a subject string using a reloaded pattern
693 is likely to cause pcretest to crash. Finally, if you attempt to load
694 a file that is not in the correct format, the result is undefined.
699 pcre(3), pcreapi(3), pcrecallout(3), pcrematching(3), pcrepartial(d),
700 pcrepattern(3), pcreprecompile(3).
706 University Computing Service
707 Cambridge CB2 3QH, England.
712 Last updated: 21 November 2010
713 Copyright (c) 1997-2010 University of Cambridge.