PCRE2TEST(1)

📦

version 10.47 · 8 versions

💖 PCRE2TEST(1) 💖

General Commands Manual — welcome to the group chat bestie, this is the manual for pcre2test 💅

✨ NAME ✨

pcre2test — a program for testing Perl-compatible regular expressions.

✨ SYNOPSIS ✨

pcre2test [options] [input file [output file]]

✨ DESCRIPTION ✨

Okay so first things first: pcre2test is a test program for the PCRE2 regular expression libraries, but plot twist — it's also genuinely iconic for just messing around with regular expressions. This doc walks you through all the features of the test program itself. But heads up: if you want the tea on the regular expressions themselves, go read the pcre2pattern documentation. And if you wanna know about the PCRE2 library function calls and their options, that's the pcre2api documentation. Different vibes, different docs. 🫶

The input for pcre2test is a sequence of regular expression patterns and subject strings to be matched. There are also command lines for setting defaults and controlling some special actions. The output shows the result of each match attempt. Modifiers on external or internal command lines, the patterns, and the subject lines specify PCRE2 function options, control how the subject is processed, and what output is produced.

There are many obscure modifiers, some of which are specifically designed for use in conjunction with the test script and data files that are distributed as part of PCRE2. All the modifiers are documented here, some without much justification, but many of them are unlikely to be of use except when testing the libraries. Like, no cap, some of these are deep lore. 😭

🦋 PCRE2's 8-BIT, 16-BIT AND 32-BIT LIBRARIES 🦋

So different versions of the PCRE2 library can be built to support character strings that are encoded in 8-bit, 16-bit, or 32-bit code units. One, two, or all three of these libraries may be simultaneously installed. The pcre2test program can be used to test all the libraries. However — and this is the key detail — its own input and output are always in 8-bit format. When testing the 16-bit or 32-bit libraries, patterns and subject strings are converted to 16-bit or 32-bit format before being passed to the library functions. Results are converted back to 8-bit code units for output. It's giving translator energy. 💅

In the rest of this document, the names of library functions and structures are given in generic form, for example, pcre2_compile(). The actual names used in the libraries have a suffix _8, _16, or _32, as appropriate.

🌸 INPUT ENCODING 🌸

Input to pcre2test is processed line by line, either by calling the C library's fgets() function, or via the libreadline or libedit library. In some Windows environments character 26 (hex 1A) causes an immediate end of file, and no further data is read, so this character should be avoided unless you really want that action. Fr fr, watch out for that one on Windows. 💀

The input is processed using C's string functions, so must not contain binary zeros, even though in Unix-like environments, fgets() treats any bytes other than newline as data characters. An error is generated if a binary zero is encountered. By default subject lines are processed for backslash escapes, which makes it possible to include any data value in strings that are passed to the library for matching. For patterns, there is a facility for specifying some or all of the 8-bit input characters as hexadecimal pairs, which makes it possible to include binary zeros. Sneaky but valid. 🎀

💜 Input for the 16-bit and 32-bit libraries 💜

When testing the 16-bit or 32-bit libraries, there is a need to be able to generate character code points greater than 255 in the strings that are passed to the library. For subject lines and some patterns, backslash escapes can be used. In addition, when the utf modifier (see "Setting compilation options" below) is set, the pattern and any following subject lines are interpreted as UTF-8 strings and translated to UTF-16 or UTF-32 as appropriate.

For non-UTF testing of wide characters, the utf8_input modifier can be used. This is mutually exclusive with utf, and is allowed only in 16-bit or 32-bit mode. It causes the pattern and following subject lines to be treated as UTF-8 according to the original definition (RFC 2279), which allows for character values up to 0x7fffffff. Each character is placed in one 16-bit or 32-bit code unit (in the 16-bit case, values greater than 0xffff cause an error to occur). Bestie, 0xffff is the ceiling in 16-bit land. 😭

UTF-8 (in its original definition) is not capable of encoding values greater than 0x7fffffff, but such values can be handled by the 32-bit library. When testing this library in non-UTF mode with utf8_input set, if any character is preceded by the byte 0xff (which is an invalid byte in UTF-8) 0x80000000 is added to the character's value. For subject strings, using an escape sequence is preferable. Lowkey the escape sequence route is the cleaner slay here. ✨

✨ COMMAND LINE OPTIONS ✨

Okay bestie, here's where it gets juicy — every single flag, no skips, no cap. 🎀

-8 — if the 8-bit library got built, this makes it the one in use (and like, that's the default anyway 💅). If the 8-bit library was not built, this option throws an error. Rude but fair.
-16 — if the 16-bit library has been built, this makes it the chosen one. And listen: if the 8-bit library has not been built, this is the default. If the 16-bit library hasn't been built, this option errors out. 💀
-32 — if the 32-bit library has been built, this picks it. If no other library has been built, then this is the default. And if the 32-bit library hasn't been built? Error. Again. 😭
-ac — acts like every pattern is carrying the auto_callout modifier, meaning automatic callouts get inserted into every single pattern that gets compiled. Obsessed with the thoroughness tbh ✨
-AC — same energy as -ac, but extra: it also acts like every subject line has the callout_extra modifier, so you get additional information shown from the callouts. More data, more fun 🦋
-b — pretends every pattern has the fullbincode modifier; the full internal binary form of the pattern gets printed out after compilation. It's giving "show me your soul" 🔥
-C — outputs the version number of the PCRE2 library plus all the available info about optional features that are included, then dips with a zero exit code. Every other option gets ignored. Oh, and if both -C and -LM are present, whichever one shows up first is the one that counts. Main character wins 👑
-C option — outputs information about one specific build-time option, then exits. This whole thing is meant for scripts like RunTest. The following options output the value and set the exit code as indicated:
- linksize — the configured internal link size (2, 3, or 4); the exit code is set to the link size
- newline — the default newline setting: CR, LF, CRLF, ANYCRLF, ANY, or NUL; exit code is always 0
- bsr — the default setting for what \R matches: ANYCRLF or ANY; exit code is always 0

The following options output 1 for true or 0 for false, and set the exit code to that same value:

backslash-C — \C is supported (not locked out)
ebcdic — compiled for an EBCDIC environment
ebcdic-io — if PCRE2 is compiled for EBCDIC, whether pcre2test's input and output is EBCDIC or ASCII
ebcdic-nl25 — if PCRE2 is compiled for EBCDIC, whether NL (= LF) is 0x25 (otherwise it is 0x15, the default)
jit — just-in-time support is available
pcre2-16 — the 16-bit library was built
pcre2-32 — the 32-bit library was built
pcre2-8 — the 8-bit library was built
unicode — Unicode support is available

Heads up bestie ⚠️ — just because JIT support is available in the library doesn't guarantee it can actually be used, because in some environments it can't allocate executable memory. The option jitusable gives you more detailed tea on that. It returns one of these values: - 0 — JIT is available and usable - 1 — JIT is available but cannot allocate executable memory - 2 — JIT is not available - 3 — unexpected return from test call to pcre2_jit_compile()

If you hand it an unknown option, it prints an error message; the exit code is 0. 💀

--colo[u]r[=<always,auto,never>] — by default the output gets coloured if the output file is a terminal (auto). Force or suppress the ANSI colour escapes with always and never respectively. Aesthetic choices are valid 🌈
-d — behaves as if every pattern has the debug modifier; the internal form and information about the compiled pattern is output after compilation. And plot twist: -d is equivalent to -b -i. Two-for-one 🫶
-dfa — behaves as if every subject line has the dfa modifier; matching gets done with the pcre2_dfa_match() function instead of the default pcre2_match(). Different engine, different vibe 🧚
-E — runs in "preprocess only" mode (similar to "gcc -E"). The #if ... #endif commands get processed, and literally every other line is printed verbatim. No notes. ✨
-error number[,number,...] — calls pcre2_get_error_message() for each of the error numbers in the comma-separated list, shows the resulting messages on the standard output, then exits with a zero exit code. The numbers can be positive or negative. It's a convenience facility for PCRE2 maintainers, iykyk 💜
-help — outputs a brief summary of these options and then exits. Your little cheat sheet 🎀
-i — behaves as if every pattern has the info modifier; information about the compiled pattern gets given after compilation.
-jit — behaves as if every pattern line has the jit modifier; after a successful compile, each pattern is passed to the just-in-time compiler, if available. Go little pattern go 🔥
-jitfast — behaves as if every pattern line has the jitfast modifier; after a successful compile, each pattern is passed to the just-in-time compiler, if available, and each subject line gets passed directly to the JIT matcher via its "fast path". Literally zooming ⭐
-jitverify — behaves as if every pattern line has the jitverify modifier; after a successful compile, each pattern gets passed to the just-in-time compiler, if available, and the use of JIT for matching is verified. Trust but verify, bestie 💅
-LM — list modifiers: writes a list of the available pattern and subject modifiers to the standard output, then exits with a zero exit code. All other options are ignored. If both -C and any -Lx options are present, whichever is first gets recognized. 👑
-LP — list properties: writes a list of recognized Unicode properties to the standard output, then exits with a zero exit code. All other options are ignored. Same deal — if both -C and any -Lx options are present, whichever is first is recognized.
-LS — list scripts: writes a list of recognized Unicode script names to the standard output, then exits with a zero exit code. All other options ignored, and again, if both -C and any -Lx options are present, whichever is first wins 🦋
-malloc — exercises malloc() failures by first counting the number of calls made to malloc during pattern compilation and matching, then re-running the compilation and matching that many times, exercising a failure of each malloc() call. Chaos, but make it methodical 😭
-pattern modifier-list — behaves as if every pattern line contains the given modifiers.
-q — does not output the version number of pcre2test at the start of execution. Quiet luxury 💅
-S size — on Unix-like systems, sets the size of the run-time stack to size mebibytes (units of 1024*1024 bytes). Big stack energy 🌸
-subject modifier-list — behaves as if every subject line contains the given modifiers.
-t — runs each compile and match many times with a timer, and outputs the resulting times per compile or match. When JIT is used, you get separate times for the initial compile and the JIT compile. You can control the number of iterations used for timing by following -t with a number (as a separate item on the command line). For example, -t 1000 iterates 1000 times. The default is to iterate 500,000 times. That's a lot of matcha 🍓
-tm — this one is like -t except it times only the matching phase, not the compile phase. Focused queen ⭐
-T, -TM — these behave like -t and -tm, but in addition, at the end of a run, the total times for all compiles and matches get output. The receipts at the end 🧾
-unittest — runs a fixed set of additional tests of the PCRE2 API which are not driven by the test input files, and then exits. Surprise bonus round 🎀
-version — outputs the PCRE2 version number and then exits. Simple, iconic, done 🌟

🌸 DESCRIPTION 🌸

Okay so here's how this actually works, bestie 💖

If pcre2test gets two filename arguments, it reads from the first one and writes to the second. If that first name is -, then the input comes from the standard input instead. If pcre2test gets only one argument, it reads from that file and writes to stdout. And otherwise? It reads from stdin and writes to stdout. Simple little funnel situation 🫶

When pcre2test is built, a configuration option can say it should be linked with the libreadline or libedit library. When that's the case, and the input is from a terminal, it gets read using the readline() function. That's what gives you the line-editing and history facilities — yes, the little arrow-key magic ✨. The output from the -help option tells you whether or not readline() will be used.

The program handles any number of tests, and each test is made up of a set of input lines. Every set starts with a regular expression pattern, followed by any number of subject lines that get matched against that pattern. In between the sets of test data, command lines that begin with # may appear. This file format — with some restrictions — can also be processed by the perltest.sh script that ships with PCRE2, as a way of checking that the behaviour of PCRE2 and Perl is the same. For a spec of perltest.sh, go look at the comments near its beginning. Also see the #perltest command below 👀

When the input is a terminal, pcre2test prompts for each line of input, using re> to prompt for regular expression patterns and data> to prompt for subject lines. Command lines starting with # can only be entered in response to the re> prompt. Rules are rules 🎀

Each subject line is matched separately and independently. If you want to do multi-line matches, you have to use the \n escape sequence (or \r, or \r\n, etc., depending on the newline setting) inside a single line of input to encode the newline sequences. There is no limit on the length of subject lines — the input buffer automatically gets extended if it's too small. And there are replication features that make it possible to generate long repetitive pattern or subject lines without having to type them all out explicitly. Lazy girl win 💅

An empty line or the end of the file signals the end of the subject lines for a test, and at that point a new pattern or command line is expected, if there's still input left to read. Periodt. 🦋

💅 COMMAND LINES 💅

okay bestie, so between sets of test data, a line that begins with # is literally interpreted as a command line. If the first character is followed by white space or an exclamation mark, the line is treated as a comment and ignored. Otherwise, these commands are recognized:

🦋 `#forbid_utf` 🦋

Subsequent patterns automatically have the PCRE2_NEVER_UTF and PCRE2_NEVER_UCP options set, which locks out the use of the PCRE2_UTF and PCRE2_UCP options and the use of (*UTF) and (*UCP) at the start of patterns. This command also forces an error if a subsequent pattern contains any occurrences of \P, \p, or \X, which are still supported when PCRE2_UTF is not set, but which require Unicode property support to be included in the library.

This is a trigger guard that is used in test files to ensure that UTF or Unicode property tests are not accidentally added to files that are used when Unicode support is not included in the library. Setting PCRE2_NEVER_UTF and PCRE2_NEVER_UCP as a default can also be obtained by the use of #pattern; the difference is that #forbid_utf cannot be unset, and the automatic options are not displayed in pattern information, to avoid cluttering up test output.

🎀 `#load <filename>` 🎀

This command is used to load a set of precompiled patterns from a file, as described in the section entitled "Saving and restoring compiled patterns" below.

🌸 `#loadtables <filename>` 🌸

This command is used to load a set of binary character tables that can be accessed by the tables=3 qualifier. Such tables can be created by the pcre2_dftables program with the -b option.

✨ `#newline_default [<newline-list>]` ✨

When PCRE2 is built, a default newline convention can be specified. This determines which characters and/or character pairs are recognized as indicating a newline in a pattern or subject string. The default can be overridden when a pattern is compiled. The standard test files contain tests of various newline conventions, but the majority of the tests expect a single linefeed to be recognized as a newline by default. Without special action the tests would fail when PCRE2 is compiled with either CR or CRLF as the default newline.

The #newline_default command specifies a list of newline types that are acceptable as the default. The types must be one of CR, LF, CRLF, ANYCRLF, ANY, or NUL (in upper or lower case), for example:

#newline_default LF Any anyCRLF

If the default newline is in the list, this command has no effect. Otherwise, except when testing the POSIX API, a newline modifier that specifies the first newline convention in the list (LF in the above example) is added to any pattern that does not already have a newline modifier. If the newline list is empty, the feature is turned off. This command is present in a number of the standard test input files.

When the POSIX API is being tested there is no way to override the default newline convention, though it is possible to set the newline convention from within the pattern. A warning is given if the posix or posix_nosub modifier is used when #newline_default would set a default for the non-POSIX API.

💜 `#pattern <modifier-list>` 💜

This command sets a default modifier list that applies to all subsequent patterns. Modifiers on a pattern can change these settings.

🍓 `#perltest` 🍓

This line is used in test files that can also be processed by perltest.sh to confirm that Perl gives the same results as PCRE2. Subsequent tests are checked for the use of pcre2test features that are incompatible with the perltest.sh script.

Patterns must use / as their delimiter, and only certain modifiers are supported. Comment lines, #pattern commands, and #subject commands that set or unset "mark" are recognized and acted on. The #perltest, #forbid_utf, and #newline_default commands, which are needed in the relevant pcre2test files, are silently ignored. All other command lines are ignored, but give a warning message. The #perltest command helps detect tests that are accidentally put in the wrong file or use the wrong delimiter. For more details of the perltest.sh script see the comments it contains.

🌈 `#pop [<modifiers>]` / `#popcopy [<modifiers>]` 🌈

These commands are used to manipulate the stack of compiled patterns, as described in the section entitled "Saving and restoring compiled patterns" below.

🫶 `#save <filename>` 🫶

This command is used to save a set of compiled patterns to a file, as described in the section entitled "Saving and restoring compiled patterns" below.

💖 `#subject <modifier-list>` 💖

This command sets a default modifier list that applies to all subsequent subject lines. Modifiers on a subject line can change these settings.

🔥 `#if CONDITION` / `#endif` 🔥

#if CONDITION
...
#endif

If CONDITION is true, then the command is printed, and its contents are processed as normal, including printing the commandlines to the output. If CONDITION is false, then all lines between the #if and #endif are skipped and not printed. The CONDITION can be any of the conditions which are tested by the -C commandline option and which set pcre2test's exit code to a boolean value. The CONDITION may also be preceded by !.

🌈 MODIFIER SYNTAX 🌈

Modifier lists are used with both pattern and subject lines. Items in a list are separated by commas followed by optional white space. Trailing white space in a modifier list is ignored. Some modifiers may be given for both patterns and subject lines, whereas others are valid only for one or the other. Each modifier has a long name, for example "anchored", and some of them must be followed by an equals sign and a value, for example, "offset=12". Values cannot contain comma characters, but may contain spaces. Modifiers that do not take values may be preceded by a minus sign to turn off a previous setting.

A few of the more common modifiers can also be specified as single letters, for example "i" for "caseless". In documentation, following the Perl convention, these are written with a slash ("the /i modifier") for clarity. Abbreviated modifiers must all be concatenated in the first item of a modifier list. If the first item is not recognized as a long modifier name, it is interpreted as a sequence of these abbreviations. For example:

/abc/ig,newline=cr,jit=3

This is a pattern line whose modifier list starts with two one-letter modifiers (/i and /g). The lower-case abbreviated modifiers are the same as used in Perl.

🦋 PATTERN SYNTAX 🦋

A pattern line must start with one of the following characters (common symbols, excluding pattern meta-characters):

/ ! " ' ` - = _ : ; , % & @ ~

This is interpreted as the pattern's delimiter. A regular expression may be continued over several input lines, in which case the newline characters are included within it. It is possible to include the delimiter as a literal within the pattern by escaping it with a backslash, for example

/abc\/def/

If you do this, the escape and the delimiter form part of the pattern, but since the delimiters are all non-alphanumeric, the inclusion of the backslash does not affect the pattern's interpretation. Note, however, that this trick does not work within \Q...\E literal bracketing because the backslash will itself be interpreted as a literal. If the terminating delimiter is immediately followed by a backslash, for example,

/abc/\

a backslash is added to the end of the pattern. This is done to provide a way of testing the error condition that arises if a pattern finishes with a backslash, because

/abc\/

is interpreted as the first line of a pattern that starts with "abc/", causing pcre2test to read the next line as a continuation of the regular expression.

A pattern can be followed by a modifier list (details below).

🌸 SUBJECT LINE SYNTAX 🌸

Okay bestie, before each subject line gets passed to pcre2_match(), pcre2_dfa_match(), or pcre2_jit_match(), the leading and trailing white space is removed (bye, extra spaces ✨), and the line is scanned for backslash escapes — unless the subject_literal modifier was set for the pattern. The following provide a means of encoding non-printing characters in a visible way:

\a — alarm (BEL, \x07)
\b — backspace (\x08)
\e — escape (\x27)
\f — form feed (\x0c)
\n — newline (\x0a)
\N{U+hh...} — unicode character (any number of hex digits)
\r — carriage return (\x0d)
\t — tab (\x09)
\v — vertical tab (\x0b)
\ddd — octal number (up to 3 octal digits); represent a single code point unless larger than 255 with the 8-bit library
\o{dd...} — octal number (any number of octal digits} representing a character in UTF mode or a code point
\xhh — hexadecimal byte (up to 2 hex digits)
\x{hh...} — hexadecimal number (up to 8 hex digits) representing a character in UTF mode or a code point

Invoking \N{U+hh...} or \x{hh...} doesn't require the use of the utf modifier on the pattern. It is always recognized. There may be any number of hexadecimal digits inside the braces; invalid values provoke error messages but when using \N{U+hh...} with some invalid unicode characters they will be accepted with a warning instead.

Note that even in UTF-8 mode, \xhh (and depending of how large, \ddd) describe one byte rather than one character; this makes it possible to construct invalid UTF-8 sequences for testing purposes. On the other hand, \x{hh...} is interpreted as a UTF-8 character in UTF-8 mode, only generating more than one byte if the value is greater than 127. To avoid the ambiguity it is preferred to use \N{U+hh...} when describing characters. When testing the 8-bit library not in UTF-8 mode, \x{hh} generates one byte for values that could fit on it, and causes an error for greater values.

When testing the 16-bit library, not in UTF-16 mode, all 4-digit \x{hhhh} values are accepted. This makes it possible to construct invalid UTF-16 sequences for testing purposes.

When testing the 32-bit library, not in UTF-32 mode, all 4 to 8-digit \x{...} values are accepted. This makes it possible to construct invalid UTF-32 sequences for testing purposes.

There is a special backslash sequence that specifies replication of one or more characters:

\[<characters>]{<count>}

This makes it possible to test long strings without having to provide them as part of the file. For example:

\[abc]{4}

is converted to "abcabcabcabc". This feature does not support nesting. To include a closing square bracket in the characters, code it as \x5D.

A backslash followed by an equals sign marks the end of the subject string and the start of a modifier list. For example:

abc\=notbol,notempty

If the subject string is empty and \= is followed by white space, the line is treated as a comment line, and is not used for matching. For example:

\= This is a comment.
abc\= This is an invalid modifier list.

A backslash followed by any other non-alphanumeric character just escapes that character. A backslash followed by anything else causes an error. However, if the very last character in the line is a backslash (and there is no modifier list), it is ignored. This gives a way of passing an empty line as data, since a real empty line terminates the data input.

If the subject_literal modifier is set for a pattern, all subject lines that follow are treated as literals, with no special treatment of backslashes. No replication is possible, and any subject modifiers must be set as defaults by a #subject command.

🎀 PATTERN MODIFIERS 🎀

There are several types of modifier that can appear in pattern lines. Except where noted below, they may also be used in #pattern commands. A pattern's modifier list can add to or override default modifiers that were set by a previous #pattern command.

✨ Setting compilation options ✨

The following modifiers set options for pcre2_compile(). Most of them set bits in the options argument of that function, but those whose names start with PCRE2_EXTRA are additional options that are set in the compile context. Some of these options have single-letter abbreviations. There is special handling for /x: if a second x is present, PCRE2_EXTENDED is converted into PCRE2_EXTENDED_MORE as in Perl. A third appearance adds PCRE2_EXTENDED as well, though this makes no difference to the way pcre2_compile() behaves. See pcre2api for a description of the effects of these options.

allow_empty_class — set PCRE2_ALLOW_EMPTY_CLASS
allow_lookaround_bsk — set PCRE2_EXTRA_ALLOW_LOOKAROUND_BSK
allow_surrogate_escapes — set PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES
alt_bsux — set PCRE2_ALT_BSUX
alt_circumflex — set PCRE2_ALT_CIRCUMFLEX
alt_extended_class — set PCRE2_ALT_EXTENDED_CLASS
alt_verbnames — set PCRE2_ALT_VERBNAMES
anchored — set PCRE2_ANCHORED
/a, ascii_all — set all ASCII options
ascii_bsd — set PCRE2_EXTRA_ASCII_BSD
ascii_bss — set PCRE2_EXTRA_ASCII_BSS
ascii_bsw — set PCRE2_EXTRA_ASCII_BSW
ascii_digit — set PCRE2_EXTRA_ASCII_DIGIT
ascii_posix — set PCRE2_EXTRA_ASCII_POSIX
auto_callout — set PCRE2_AUTO_CALLOUT
bad_escape_is_literal — set PCRE2_EXTRA_BAD_ESCAPE_IS_LITERAL
/i, caseless — set PCRE2_CASELESS
/r, caseless_restrict — set PCRE2_EXTRA_CASELESS_RESTRICT
dollar_endonly — set PCRE2_DOLLAR_ENDONLY
/s, dotall — set PCRE2_DOTALL
dupnames — set PCRE2_DUPNAMES
endanchored — set PCRE2_ENDANCHORED
escaped_cr_is_lf — set PCRE2_EXTRA_ESCAPED_CR_IS_LF
/x, extended — set PCRE2_EXTENDED
/xx, extended_more — set PCRE2_EXTENDED_MORE
extra_alt_bsux — set PCRE2_EXTRA_ALT_BSUX
firstline — set PCRE2_FIRSTLINE
literal — set PCRE2_LITERAL
match_line — set PCRE2_EXTRA_MATCH_LINE
match_invalid_utf — set PCRE2_MATCH_INVALID_UTF
match_unset_backref — set PCRE2_MATCH_UNSET_BACKREF
match_word — set PCRE2_EXTRA_MATCH_WORD
/m, multiline — set PCRE2_MULTILINE
never_backslash_c — set PCRE2_NEVER_BACKSLASH_C
never_callout — set PCRE2_EXTRA_NEVER_CALLOUT
never_ucp — set PCRE2_NEVER_UCP
never_utf — set PCRE2_NEVER_UTF
/n, no_auto_capture — set PCRE2_NO_AUTO_CAPTURE
no_auto_possess — set PCRE2_NO_AUTO_POSSESS
no_bs0 — set PCRE2_EXTRA_NO_BS0
no_dotstar_anchor — set PCRE2_NO_DOTSTAR_ANCHOR
no_start_optimize — set PCRE2_NO_START_OPTIMIZE
no_utf_check — set PCRE2_NO_UTF_CHECK
python_octal — set PCRE2_EXTRA_PYTHON_OCTAL
turkish_casing — set PCRE2_EXTRA_TURKISH_CASING
ucp — set PCRE2_UCP
ungreedy — set PCRE2_UNGREEDY
use_offset_limit — set PCRE2_USE_OFFSET_LIMIT
utf — set PCRE2_UTF

As well as turning on the PCRE2_UTF option, the utf modifier causes all non-printing characters in output strings to be printed using the \x{hh...} notation. Otherwise, those less than 0x100 are output in hex without the curly brackets. Setting utf in 16-bit or 32-bit mode also causes pattern and subject strings to be translated to UTF-16 or UTF-32, respectively, before being passed to library functions.

The following modifiers enable or disable performance optimizations by calling pcre2_set_optimize() before invoking the regex compiler.

optimization_full — enable all optional optimizations
optimization_none — disable all optional optimizations
auto_possess — auto-possessify variable quantifiers
auto_possess_off — don't auto-possessify variable quantifiers
dotstar_anchor — anchor patterns starting with .*
dotstar_anchor_off — don't anchor patterns starting with .*
start_optimize — enable pre-scan of subject string
start_optimize_off — disable pre-scan of subject string

See the pcre2_set_optimize documentation for details on these optimizations.

💅 Setting compilation controls 💅

The following modifiers affect the compilation process or request information about the pattern. There are single-letter abbreviations for some that are heavily used in the test files.

/B, bincode — show the binary code, but without the length values
bsr=[anycrlf|unicode] — tell it how \R should behave
callout_info — show callout info
convert=<options> — ask for a foreign pattern to get converted
convert_glob_escape=c — set the glob escape character
convert_glob_separator=c — set the glob separator character
convert_length — set the convert buffer length
debug — literally the same as info,fullbincode, i.e. give me everything
expand — expand the repetition syntax inside the pattern
framesize — show the size of the matching frame
fullbincode — show the binary code with lengths this time
/I, info — show info about the compiled pattern
hex — unquoted characters are read as hexadecimal
jit[=<number>] — use JIT
jitfast — use the JIT fast path
jitverify — verify that JIT was actually used
locale=<name> — use this locale
max_pattern_compiled_length=<n> — set the maximum compiled pattern length, in bytes
max_pattern_length=<n> — set the maximum pattern length, in code units
max_varlookbehind=<n> — set the maximum variable lookbehind length
memory — show how much memory got used
newline=<type> — set the newline type
null_context — compile with a NULL context
null_pattern — pass the pattern as NULL
parens_nest_limit=<n> — set the maximum parentheses depth
posix — use the POSIX API
posix_nosub — use the POSIX API with REG_NOSUB
push — push the compiled pattern onto the stack
pushcopy — push a copy onto the stack
pushtablescopy — push a copy, with tables, onto the stack
stackguard=<number> — test out the stackguard feature
subject_literal — treat every subject line as literal
tables=[0|1|2|3] — pick which internal tables to use
use_length — don't zero-terminate the pattern
utf8_input — treat the input as UTF-8

The effects of these modifiers are described in the following sections. 💅

🌊 Newline and `\R` handling

The bsr modifier is the one that decides what \R in a pattern actually matches, bestie.

Set it to "anycrlf" and \R matches CR, LF, or CRLF only.
Set it to "unicode" and \R matches any Unicode newline sequence.

The default can be specified when PCRE2 is built; if it isn't, the default is set to Unicode. ✨

The newline modifier says which characters get interpreted as newlines — both in the pattern and in subject lines. The type has to be one of CR, LF, CRLF, ANYCRLF, ANY, or NUL (upper or lower case, she doesn't care).

🔍 Information about a pattern

The debug modifier is a shorthand for info,fullbincode — it requests all available information. All of it. Every crumb. 🍓

The bincode modifier makes a representation of the compiled code get output after compilation. The way this works: the information does not contain length and offset values, which ensures that the same output is generated for different internal link sizes and different code unit widths. By using bincode, the same regression tests can be used in different environments. slay, portable queen 👑

The fullbincode modifier, by contrast, does include length and offset values. This is used in a few special tests that run only for specific code unit widths and link sizes, and is also useful for one-off tests.

The info modifier requests information about the compiled pattern — whether it's anchored, whether it has a fixed first character, and so on. The information is obtained from the pcre2_pattern_info() function. Here are some typical examples:

    re> /(?i)(^a|^b)/m,info
  Capture group count = 1
  Compile options: multiline
  Overall options: caseless multiline
  First code unit at start or follows newline
  Subject length lower bound = 1

    re> /(?i)abc/info
  Capture group count = 0
  Compile options: <none>
  Overall options: caseless
  First code unit = 'a' (caseless)
  Last code unit = 'c' (caseless)
  Subject length lower bound = 3

So: "Compile options" are those specified by modifiers; "overall options" have added options that are taken or deduced from the pattern. If both sets of options are the same, just a single "options" line is output; if there are no options, the line is omitted entirely. "First code unit" is where any match must start; if there's more than one, they get listed as "starting code units". "Last code unit" is the last literal code unit that must be present in any match — and no, that's not necessarily the last character, okurrr. These lines are omitted if no starting or ending code units are recorded. The subject length line is omitted when no_start_optimize is set, because the minimum length is not calculated when it can never be used.

The framesize modifier shows the size, in bytes, of each storage frame used by pcre2_match() for handling backtracking. That size depends on the number of capturing parentheses in the pattern. A vector of these frames is used at matching time; its overall size is shown when the heaframes_size subject modifier is set.

The callout_info modifier requests information about all the callouts in the pattern. A list of them is output at the end of any other information that is requested. For each callout, either its number or its string is given, followed by the item that follows it in the pattern. 🎀

🫥 Passing a NULL context

Normally, pcre2test passes a context block to pcre2_compile(). If the null_context modifier is set, though, NULL is what gets passed. This is for testing that pcre2_compile() behaves correctly in this case — it uses default values. Chaos testing but make it respectful 💀

🕳️ Passing a NULL pattern

The null_pattern modifier is for testing the behaviour of pcre2_compile() when the pattern argument is NULL. The length value passed is the default PCRE2_ZERO_TERMINATED unless use_length is set. Any length other than zero causes an error. periodt.

🔢 Specifying pattern characters in hexadecimal

The hex modifier specifies that the characters of the pattern — except for substrings enclosed in single or double quotes — are to be interpreted as pairs of hexadecimal digits. This feature exists as a way of creating patterns that contain binary zeros and other non-printing characters. White space is permitted between pairs of digits. For example, this pattern contains three characters:

  /ab 32 59/hex

Parts of such a pattern are taken literally if quoted. This pattern contains nine characters, only two of which are specified in hexadecimal:

  /ab "literal" 32/hex

Either single or double quotes may be used. There is no way of including the delimiter within a substring. Also — the hex and expand modifiers are mutually exclusive, they cannot both be on, no cap.

📏 Specifying the pattern's length

By default, patterns are passed to the compiling functions as zero-terminated strings, but they can be passed by length instead of being zero-terminated. The use_length modifier causes this to happen. Using a length happens automatically — whether or not use_length is set — when hex is set, because patterns specified in hexadecimal may contain binary zeros.

If hex or use_length is used with the POSIX wrapper API (see "Using the POSIX wrapper API" below), the REG_PEND extension is used to pass the pattern's length.

🦋 Specifying a maximum for variable lookbehinds

Variable lookbehind assertions are supported only if, for each one, there's a maximum length (in characters) that it can match. There's a limit on this, whose default can be set at build time, with an ultimate default of 255. The max_varlookbehind modifier uses the pcre2_set_max_varlookbehind() function to change the limit. Lookbehinds whose branches each match a fixed length are limited to 65535 characters per branch. Big numbers, big vibes ✨

🌈 Specifying wide characters in 16-bit and 32-bit modes

In 16-bit and 32-bit modes, all input is automatically treated as UTF-8 and translated to UTF-16 or UTF-32 when the utf modifier is set. For testing the 16-bit and 32-bit libraries in non-UTF mode, the utf8_input modifier can be used. It is mutually exclusive with utf. Input lines are interpreted as UTF-8 as a means of specifying wide characters. More details are given in "Input encoding" above. 👑

🔁 Generating long repetitive patterns

Some tests use long patterns that are very repetitive. Instead of creating a very long input line for such a pattern, you can use a special repetition feature, similar to the one described for subject lines above. If the expand modifier is present on a pattern, parts of the pattern that have the form

  \[<characters>]{<count>}

are expanded before the pattern is passed to pcre2_compile(). For example, \[AB]{6000} is expanded to "ABAB..." 6000 times. This construction cannot be nested. An initial "\[" sequence is recognized only if "]{" followed by decimal digits and "}" is found later in the pattern. If not, the characters remain in the pattern unaltered. The expand and hex modifiers are mutually exclusive.

If part of an expanded pattern looks like an expansion, but is really part of the actual pattern, unwanted expansion can be avoided by giving two values in the quantifier. For example, \[AB]{6000,6000} is not recognized as an expansion item. iykyk 🧚

If the info modifier is set on an expanded pattern, the result of the expansion is included in the information that is output.

⚡ JIT compilation

Just-in-time (JIT) compiling is a heavyweight optimization that can greatly speed up pattern matching. See the pcre2jit documentation for details. JIT compiling happens, optionally, after a pattern has been successfully compiled into an internal form. The JIT compiler converts this to optimized machine code. It needs to know whether the match-time options PCRE2_PARTIAL_HARD and PCRE2_PARTIAL_SOFT are going to be used, because different code is generated for the different cases. See the partial modifier in "Subject Modifiers" below for details of how these options are specified for each match attempt.

JIT compilation is requested by the jit pattern modifier, which may optionally be followed by an equals sign and a number in the range 0 to 7. The three bits that make up the number specify which of the three JIT operating modes are to be compiled:

  1  compile JIT code for non-partial matching
  2  compile JIT code for soft partial matching
  4  compile JIT code for hard partial matching

The possible values for the jit modifier are therefore:

  0  disable JIT
  1  normal matching only
  2  soft partial matching only
  3  normal and soft partial matching
  4  hard partial matching only
  6  soft and hard partial matching only
  7  all three modes

If no number is given, 7 is assumed. The phrase "partial matching" means a call to pcre2_match() with either the PCRE2_PARTIAL_SOFT or the PCRE2_PARTIAL_HARD option set. 💅 Heads up though — that call may still return a complete match; the options just enable the possibility of a partial match, they don't require one. Also: if you request JIT compilation only for partial matching (like jit=2) but you don't set the partial modifier on a subject line, that match won't use JIT code, because literally none was compiled for non-partial matching. 😭

If JIT compilation is successful, the compiled JIT code gets used automatically whenever an appropriate type of match is run — except when incompatible run-time options are specified. More deets in the pcre2jit documentation. Also peep the jitstack modifier below, it's the whole vibe for setting the size of the JIT stack. ✨

If the jitfast modifier is specified, matching is done using the JIT "fast path" interface, pcre2_jit_match(), which skips some of the sanity checks that pcre2_match() does — and obvs doesn't work when JIT isn't supported. If jitfast is specified without jit, jit=7 is assumed.

If the jitverify modifier is specified, information about the compiled pattern shows whether JIT compilation was or was not successful. If jitverify is specified without jit, jit=7 is assumed. If JIT compilation is successful when jitverify is set, the text (JIT) is added to the first output line after a match or non match when JIT-compiled code was actually used in the match. 🎀

🌍 Setting a locale (yes bestie, actual locales)

The locale modifier must specify the name of a locale, for example:

/pattern/locale=fr_FR

The given locale is set, pcre2_maketables() is called to build a set of character tables for the locale, and this is then passed to pcre2_compile() when compiling the regular expression. The same tables are used when matching the following subject lines. The locale modifier applies only to the pattern on which it appears, but can be given in a #pattern command if a default is needed. Setting a locale and alternate character tables are mutually exclusive — pick one, you can't have both. 🙅‍♀️

🧠 Showing pattern memory

The memory modifier causes the size in bytes of the memory used to hold the compiled pattern to be output. This does not include the size of the pcre2_code block; it's just the actual compiled data. If the pattern is subsequently passed to the JIT compiler, the size of the JIT compiled code is also output. Iconic little example:

    re> /a(b)c/jit,memory
  Memory allocation (code space): 21
  Memory allocation (JIT code): 1910

🪆 Limiting nested parentheses

The parens_nest_limit modifier sets a limit on the depth of nested parentheses in a pattern. Breaching the limit causes a compilation error. 😤 The default for the library is set when PCRE2 is built, but pcre2test sets its own default of 220, which is required for running the standard test suite.

📏 Limiting the pattern length

The max_pattern_length modifier sets a limit, in code units, to the length of pattern that pcre2_compile() will accept. Breaching the limit causes a compilation error. The default is the largest number a PCRE2_SIZE variable can hold (essentially unlimited). ✨

🧱 Limiting the size of a compiled pattern

The max_pattern_compiled_length modifier sets a limit, in bytes, to the amount of memory used by a compiled pattern. Breaching the limit causes a compilation error. The default is the largest number a PCRE2_SIZE variable can hold (essentially unlimited).

🫶 Using the POSIX wrapper API

The posix and posix_nosub modifiers cause pcre2test to call PCRE2 via the POSIX wrapper API rather than its native API. When posix_nosub is used, the POSIX option REG_NOSUB is passed to regcomp(). The POSIX wrapper supports only the 8-bit library. Note that it does not imply POSIX matching semantics — for more detail see the pcre2posix documentation. The following pattern modifiers set options for the regcomp() function:

caseless           REG_ICASE
multiline          REG_NEWLINE
dotall             REG_DOTALL     )
ungreedy           REG_UNGREEDY   ) These options are not part of
ucp                REG_UCP        )   the POSIX standard
utf                REG_UTF8       )

The regerror_buffsize modifier specifies a size for the error buffer that is passed to regerror() in the event of a compilation error. For example:

/abc/posix,regerror_buffsize=20

This provides a means of testing the behaviour of regerror() when the buffer is too small for the error message. If this modifier has not been set, a large buffer is used.

The aftertext and allaftertext subject modifiers work as described below. All other modifiers are either ignored, with a warning message, or cause an error. 😬

The pattern is passed to regcomp() as a zero-terminated string by default, but if the use_length or hex modifiers are set, the REG_PEND extension is used to pass it by length.

🛡️ Testing the stack guard feature

The stackguard modifier is used to test the use of pcre2_set_compile_recursion_guard(), a function that is provided to enable stack availability to be checked during compilation (see the pcre2api documentation for details). If the number specified by the modifier is greater than zero, pcre2_set_compile_recursion_guard() is called to set up callback from pcre2_compile() to a local function. The argument it receives is the current nesting parenthesis depth; if this is greater than the value given by the modifier, non-zero is returned, causing the compilation to be aborted. 💀

🔤 Using alternative character tables

The value specified for the tables modifier must be one of the digits 0, 1, 2, or 3. It causes a specific set of built-in character tables to be passed to pcre2_compile(). This is used in the PCRE2 tests to check behaviour with different character tables. The digit specifies the tables as follows:

0 — do not pass any special character tables
1 — the default ASCII tables, as distributed in pcre2_chartables.c.dist
2 — a set of tables defining ISO 8859 characters
3 — a set of tables loaded by the #loadtables command

In tables 2, some characters whose codes are greater than 128 are identified as letters, digits, spaces, etc. Tables 3 can be used only after a #loadtables command has loaded them from a binary file. Setting alternate character tables and a locale are mutually exclusive. 🙈

🎛️ Setting certain match controls

The following modifiers are really subject modifiers, and are described under "Subject Modifiers" below. But plot twist — they may be included in a pattern's modifier list, in which case they are applied to every subject line that is processed with that pattern. These modifiers do not affect the compilation process. ✨

aftertext — show text after match
allaftertext — show text after captures
allcaptures — show all captures
allvector — show the entire ovector
allusedtext — show all consulted text
altglobal — alternative global matching
global (or /g) — global matching
heapframes_size — show match data heapframes size
jitstack=<n> — set size of JIT stack
mark — show mark values
null_substitute_match_data — substitute with NULL match data
replace=<str> — specify a replacement string
startchar — show starting character when relevant
substitute_callout — use substitution callouts
substitute_case_callout — use substitution case callouts
substitute_extended — use PCRE2_SUBSTITUTE_EXTENDED
substitute_literal — use PCRE2_SUBSTITUTE_LITERAL
substitute_matched — use PCRE2_SUBSTITUTE_MATCHED
substitute_overflow_length — use PCRE2_SUBSTITUTE_OVERFLOW_LENGTH
substitute_replacement_only — use PCRE2_SUBSTITUTE_REPLACEMENT_ONLY
substitute_skip=<n> — skip substitution <n>
substitute_stop=<n> — skip substitution <n> and following
substitute_unknown_unset — use PCRE2_SUBSTITUTE_UNKNOWN_UNSET
substitute_unset_empty — use PCRE2_SUBSTITUTE_UNSET_EMPTY

These modifiers may not appear in a #pattern command. If you want them as defaults, set them in a #subject command. 🫡

📝 Specifying literal subject lines

If the subject_literal modifier is present on a pattern, all the subject lines that it matches are taken as literal strings, with no interpretation of backslashes. It is not possible to set subject modifiers on such lines, but any that are set as defaults by a #subject command are recognized.

💾 Saving a compiled pattern

When a pattern with the push modifier is successfully compiled, it is pushed onto a stack of compiled patterns, and pcre2test expects the next line to contain a new pattern (or a command) instead of a subject line. This facility is used when saving compiled patterns to a file, as described in the section entitled "Saving and restoring compiled patterns" below. If pushcopy is used instead of push, a copy of the compiled pattern is stacked, leaving the original as current, ready to match the following input lines. This provides a way of testing the pcre2_code_copy() function. The push and pushcopy modifiers are incompatible with compilation modifiers such as global that act at match time. Any that are specified are ignored (for the stacked copy), with a warning message, except for replace, which causes an error. Note that jitverify, which is allowed, does not carry through to any subsequent matching that uses a stacked pattern. 🎀

🧚 Testing foreign pattern conversion

The experimental foreign pattern conversion functions in PCRE2 can be tested by setting the convert modifier. Its argument is a colon-separated list of options, which set the equivalent option for the pcre2_pattern_convert() function:

glob — PCRE2_CONVERT_GLOB
glob_no_starstar — PCRE2_CONVERT_GLOB_NO_STARSTAR
glob_no_wild_separator — PCRE2_CONVERT_GLOB_NO_WILD_SEPARATOR
posix_basic — PCRE2_CONVERT_POSIX_BASIC
posix_extended — PCRE2_CONVERT_POSIX_EXTENDED
unset — Unset all options

The "unset" value is useful for turning off a default that has been set by a #pattern command. When one of these options is set, the input pattern is passed to pcre2_pattern_convert(). If the conversion is successful, the result is reflected in the output and then passed to pcre2_compile(). The normal utf and no_utf_check options, if set, cause the PCRE2_CONVERT_UTF and PCRE2_CONVERT_NO_UTF_CHECK options to be passed to pcre2_pattern_convert(). 🌈

By default, the conversion function is allowed to allocate a buffer for its output. But bestie, if convert_length is set to a value greater than zero, pcre2test passes a buffer of exactly that given length. That's literally the whole point: it makes it possible to test the length check. ✨

The convert_glob_escape and convert_glob_separator modifiers specify the escape and separator characters for glob processing, overriding the defaults, which are operating-system dependent. iykyk.

💅 SUBJECT MODIFIERS 💅

The modifiers that can appear in subject lines and in the #subject command come in two flavors.

🎯 Setting match options 🎯

These modifiers set options for pcre2_match() or pcre2_dfa_match(). See pcre2api for a description of their effects.

anchored — sets PCRE2_ANCHORED
copy_matched_subject — sets PCRE2_COPY_MATCHED_SUBJECT
endanchored — sets PCRE2_ENDANCHORED
dfa_restart — sets PCRE2_DFA_RESTART
dfa_shortest — sets PCRE2_DFA_SHORTEST
disable_recurseloop_check — sets PCRE2_DISABLE_RECURSELOOP_CHECK
no_jit — sets PCRE2_NO_JIT
no_utf_check — sets PCRE2_NO_UTF_CHECK
notbol — sets PCRE2_NOTBOL
notempty — sets PCRE2_NOTEMPTY
notempty_atstart — sets PCRE2_NOTEMPTY_ATSTART
noteol — sets PCRE2_NOTEOL
partial_hard (or ph) — sets PCRE2_PARTIAL_HARD
partial_soft (or ps) — sets PCRE2_PARTIAL_SOFT

The partial matching modifiers get abbreviations because they show up so frequently in tests. Periodt.

If the posix or posix_nosub modifier was present on the pattern, causing the POSIX wrapper API to be used, the only option-setting modifiers that have any effect are notbol, notempty, and noteol, causing REG_NOTBOL, REG_NOTEMPTY, and REG_NOTEOL, respectively, to be passed to regexec(). The other modifiers are ignored, with a warning message. Not them being ignored 💀.

There is one additional modifier that can be used with the POSIX wrapper. It is ignored (with a warning) if used for non-POSIX matching.

posix_startend=<n>[:<m>] — causes the subject string to be passed to regexec() using the REG_STARTEND option, which uses offsets to specify which part of the string is searched. If only one number is given, the end offset is passed as the end of the subject string. For more detail of REG_STARTEND, see the pcre2posix documentation. If the subject string contains binary zeros (coded as escapes such as \x{00} because pcre2test does not support actual binary zeros in its input), you must use posix_startend to specify its length.

🎛️ Setting match controls 🎛️

These modifiers affect the matching process or request additional information. Some of them may also be specified on a pattern line (see above), in which case they apply to every subject line that is matched against that pattern, but they can be overridden by modifiers on the subject. Main character energy: subject-line modifiers win.

aftertext — show text after match
allaftertext — show text after captures
allcaptures — show all captures
allusedtext — show all consulted text (non-JIT only)
allvector — show the entire ovector
altglobal — alternative global matching
callout_capture — show captures at callout time
callout_data=<n> — set a value to pass via callouts
callout_error=<n>[:<m>] — control callout error
callout_extra — show extra callout information
callout_fail=<n>[:<m>] — control callout failure
callout_no_where — do not show position of a callout
callout_none — do not supply a callout function
copy=<number or name> — copy captured substring
depth_limit=<n> — set a depth limit
dfa — use pcre2_dfa_match()
find_limits — find heap, match and depth limits
find_limits_noheap — find match and depth limits
get=<number or name> — extract captured substring
getall — extract all captured substrings
/g, global — global matching
heapframes_size — show match data heapframes size
heap_limit=<n> — set a limit on heap memory (Kbytes)
jitstack=<n> — set size of JIT stack
mark — show mark values
match_limit=<n> — set a match limit
memory — show heap memory usage
null_context — match with a NULL context
null_replacement — substitute with NULL replacement
null_subject — match with NULL subject
null_substitute_match_data — substitute with NULL match data
offset=<n> — set starting offset
offset_limit=<n> — set offset limit
ovector=<n> — set size of output vector
recursion_limit=<n> — obsolete synonym for depth_limit
replace=<str> — specify a replacement string
startchar — show startchar when relevant
startoffset=<n> — same as offset=<n>
substitute_callout — use substitution callouts
substitute_case_callout — use substitution case callouts
substitute_extended — use PCRE2_SUBSTITUTE_EXTENDED
substitute_literal — use PCRE2_SUBSTITUTE_LITERAL
substitute_matched — use PCRE2_SUBSTITUTE_MATCHED
substitute_overflow_length — use PCRE2_SUBSTITUTE_OVERFLOW_LENGTH
substitute_replacement_only — use PCRE2_SUBSTITUTE_REPLACEMENT_ONLY
substitute_skip=<n> — skip substitution number n
substitute_stop=<n> — skip substitution number n and greater
substitute_subject=<str> — specify a different subject for substitution
substitute_unknown_unset — use PCRE2_SUBSTITUTE_UNKNOWN_UNSET
substitute_unset_empty — use PCRE2_SUBSTITUTE_UNSET_EMPTY
zero_terminate — pass the subject as zero-terminated

The effects of these modifiers are described in the following sections. When matching via the POSIX wrapper API, the aftertext, allaftertext, and ovector subject modifiers work as described below. All other modifiers are either ignored, with a warning message, or cause an error. Choose wisely 🫶.

👀 Showing more text 👀

The aftertext modifier requests that, as well as outputting the part of the subject string that matched the entire pattern, pcre2test should in addition output the remainder of the subject string. This is useful for tests where the subject contains multiple copies of the same substring. The allaftertext modifier requests the same action for captured substrings as well as the main matched substring. In each case the remainder is output on the following line with a plus character following the capture number. Cute, right? 💖

The allusedtext modifier requests that all the text that was consulted during a successful pattern match by the interpreter should be shown, for both full and partial matches. This feature is not supported for JIT matching, and if requested with JIT it is ignored (with a warning message). Setting this modifier affects the output if there is a lookbehind at the start of a match, or, for a complete match, a lookahead at the end, or if \K is used in the pattern. Characters that precede or follow the start and end of the actual match are indicated in the output by < or > characters underneath them. Here is an example:

re> /(?<=pqr)abc(?=xyz)/
data> 123pqrabcxyz456\=allusedtext
 0: pqrabcxyz
    <<<   >>>
data> 123pqrabcxy\=ph,allusedtext
Partial match: pqrabcxy
               <<<

The first, complete match shows that the matched string is "abc", with the preceding and following strings "pqr" and "xyz" having been consulted during the match (when processing the assertions). The partial match can indicate only the preceding string. No cap.

The startchar modifier requests that the starting character for the match be indicated, if it is different to the start of the matched string. The only time when this occurs is when \K has been processed as part of the match. In this situation, the output for the matched string is displayed from the starting character instead of from the match point, with circumflex characters under the earlier characters. For example:

re> /abc\Kxyz/
data> abcxyz\=startchar
 0: abcxyz
    ^^^

Unlike allusedtext, the startchar modifier can be used with JIT. However, these two modifiers are mutually exclusive. Can't have both, bestie.

🎀 Showing the value of all capture groups 🎀

The allcaptures modifier requests that the values of all potential captured parentheses be output after a match. By default, only those up to the highest one actually used in the match are output (corresponding to the return code from pcre2_match()). Groups that did not take part in the match are output as "<unset>". This modifier is not relevant for DFA matching (which does no capturing) and does not apply when replace is specified; it is ignored, with a warning message, if present. Bye 👋.

🌈 Showing the entire ovector, for all outcomes 🌈

The allvector modifier requests that the entire ovector be shown, whatever the outcome of the match. Compare allcaptures, which shows only up to the maximum number of capture groups for the pattern, and then only for a successful complete non-DFA match. This modifier, which acts after any match result, and also for DFA matching, provides a means of checking that there are no unexpected modifications to ovector fields. Before each match attempt, the ovector is filled with a special value, and if this is found in both elements of a capturing pair, "<unchanged>" is output. After a successful match, this applies to all groups after the maximum capture group for the pattern. In other cases it applies to the entire ovector. After a partial match, the first two elements are the only ones that should be set. After a DFA match, the amount of ovector that is used depends on the number of matches that were found. It's giving forensic vibes 🔍.

📞 Testing pattern callouts 📞

A callout function is supplied when pcre2test calls the library matching functions, unless callout_none is specified. Its behaviour can be controlled by various modifiers listed above whose names begin with callout_. Details are given in the section entitled "Callouts" below. Testing callouts from pcre2_substitute() is described separately in "Testing the substitution function" below. Callout girlies, this one's for you 💅.

🔎 Finding all matches in a string 🔎

Ok so, the global or altglobal modifier is your girl when you wanna find every possible match in a subject. After a match is found, the matching function gets called again to search the rest of the subject 🎀

The tea on global vs altglobal? global uses the start_offset argument to pcre2_match() or pcre2_dfa_match() to start searching from a new point inside the whole string (literally what Perl does 😌), whereas altglobal just hands over a shortened subject instead. Why does that matter? Because if your pattern starts with a lookbehind assertion (and yes, that includes \b or \B), the matching process can absolutely tell the difference 👀

And if an empty string gets matched, the next match runs with the PCRE2_NOTEMPTY_ATSTART flag set, so it goes hunting for another, non-empty match at that same point in the subject. This is imitating how Perl handles these cases with the /g modifier or the split() function. iykyk 💅

🧪 Testing substring extraction functions

The copy and get modifiers are here to test the pcre2_substring_copy_xxx() and pcre2_substring_get_xxx() functions. You can drop them more than once, and each one can name a capture group by name or number, like this bestie:

abcd\=copy=1,copy=3,get=G1

If the #subject command was used to set default copy and/or get lists, you can unset them by giving a negative number to cancel all numbered groups, and an empty name to cancel all named groups. Clean slate energy ✨

The getall modifier tests pcre2_substring_list_get(), which extracts all the captured substrings at once. She's an overachiever, we love her.

If the subject line matches successfully, the substrings pulled out by the convenience functions are output with C, G, or L after the string number instead of a colon. This is on top of the normal full list, not instead of it. The string length (that is, the return from the extraction function) shows up in parentheses after each substring, followed by the name when the extraction was by name. Receipts everywhere 🧾

🔁 Testing the substitution function

If the replace modifier is set, pcre2_substitute() gets called instead of one of the matching functions (or after one call of pcre2_match() in the case of PCRE2_SUBSTITUTE_MATCHED). Quick heads-up though: replacement strings can't contain commas, because a comma means the end of a modifier. The manpage itself says this isn't thought to be an issue in a test program, so like, it's fine 🌸

Specifying a completely empty replacement string disables this modifier. But plot twist — you can still specify an empty replacement by providing a buffer length, as described below, for an otherwise empty replacement 👑

Unlike subject strings, pcre2test does not process replacement strings for escape sequences. In UTF mode, a replacement string gets checked to see if it's a valid UTF-8 string. If it is, it's correctly converted to a UTF string of the appropriate code unit width. If it's not valid UTF-8, the individual code units are copied directly. Which is honestly a feature, because it gives you a way to pass an invalid UTF-8 string in for testing purposes 💀

The following modifiers set options (in addition to the normal match options) for pcre2_substitute():

global — PCRE2_SUBSTITUTE_GLOBAL
substitute_extended — PCRE2_SUBSTITUTE_EXTENDED
substitute_literal — PCRE2_SUBSTITUTE_LITERAL
substitute_matched — PCRE2_SUBSTITUTE_MATCHED
substitute_overflow_length — PCRE2_SUBSTITUTE_OVERFLOW_LENGTH
substitute_replacement_only — PCRE2_SUBSTITUTE_REPLACEMENT_ONLY
substitute_unknown_unset — PCRE2_SUBSTITUTE_UNKNOWN_UNSET
substitute_unset_empty — PCRE2_SUBSTITUTE_UNSET_EMPTY

See the pcre2api documentation for details of these options. (Go read it, bestie, it's giving essential lore 📚)

After a successful substitution, the modified string is output, preceded by the number of replacements. And yes, that number can be zero if there were no matches — no matches is still a valid outcome, don't come for her. Here's a simple little substitution test:

  /abc/replace=xxx
      =abc=abc=
   1: =xxx=abc=
      =abc=abc=\=global
   2: =xxx=xxx=

Subject and replacement strings should be kept relatively short (fewer than 256 characters) for substitution tests, because fixed-size buffers are used. To make it easy to test for buffer overflow, if the replacement string starts with a number in square brackets, that number is passed to pcre2_substitute() as the size of the output buffer, with the replacement string starting at the next character. Here's an edge-case moment:

  /abc/
      123abc123\=replace=[10]XYZ
   1: 123XYZ123
      123abc123\=replace=[9]XYZ
  Failed: error -48: no more memory

The default action of pcre2_substitute() is to return PCRE2_ERROR_NOMEMORY when the output buffer is too small. But if the PCRE2_SUBSTITUTE_OVERFLOW_LENGTH option is set (via the substitute_overflow_length modifier), pcre2_substitute() keeps going through the motions of matching and substituting (just without doing any callouts) so it can compute how big the buffer actually needs to be. When that happens, pcre2test shows you the required buffer length (which includes space for the trailing zero) as part of the error message. Iconic helpfulness:

  /abc/substitute_overflow_length
      123abc123\=replace=[9]XYZ
  Failed: error -48: no more memory: 10 code units are needed

A replacement string is ignored with POSIX and DFA matching. And specifying partial matching provokes an error return ("bad option value") from pcre2_substitute(). So like, don't do that 🙈

The substitute_subject modifier may be used to test the use of the PCRE2 API, in which a client calls pcre2_match() followed by pcre2_substitute() with PCRE2_SUBSTITUTE_MATCHED, but the client performs an unexpected and unsupported modification of the subject buffer in-place, in between the match and substitution. So it's a chaos test. We respect the chaos 🫶

🎯 Testing substitute callouts

If the substitute_callout modifier is set, a substitution callout function gets set up. The null_context modifier must not be set, because the address of the callout function is passed in a match context. When the callout function is called (after each substitution), details of the input and output strings are output. Like so:

  /abc/g,replace=<$0>,substitute_callout
      abcdefabcpqr
   1(1) Old 0 3 "abc" New 0 5 "<abc>"
   2(1) Old 6 9 "abc" New 8 13 "<abc>"
   2: <abc>def<abc>pqr

How to read this masterpiece: the first number on each callout line is the count of matches. The parenthesized number is the number of pairs that are set in the ovector (that is, one more than the number of capturing groups that were set). Then you get the offsets of the old substring, its contents, and the same for the replacement. Literally a before-and-after photoshoot 📸

By default, the substitution callout function returns zero, which accepts the replacement and causes matching to continue if /g was used. Two further modifiers test other return values. If substitute_skip is set to a value greater than zero, the callout function returns +1 for the match of that number, and similarly substitute_stop returns -1. These cause the replacement to be rejected, and -1 also means no further matching takes place. If either of them are set, substitute_callout is assumed. Example time:

  /abc/g,replace=<$0>,substitute_skip=1
      abcdefabcpqr
   1(1) Old 0 3 "abc" New 0 5 "<abc> SKIPPED"
   2(1) Old 6 9 "abc" New 6 11 "<abc>"
   2: abcdef<abc>pqr
      abcdefabcpqr\=substitute_stop=1
   1(1) Old 0 3 "abc" New 0 5 "<abc> STOPPED"
   1: abcdefabcpqr

If both are set for the same number, stop takes precedence. Only a single skip or stop is supported, which is sufficient for testing that the feature works. One job, done well 💅

🔠 Testing substitute case callouts

If the substitute_case_callout modifier is set, a substitution case callout function gets set up. The callout function is called for each substituted chunk which is to be case-transformed.

The callout function passed is a fixed function with implementation for certain behaviours: inputs which shrink when case-transformed; inputs which grow; inputs with distinct upper/lower/titlecase forms. The characters which are not special-cased for testing purposes are left unmodified, as if they are caseless characters. She's got range ✨

📚 Setting the JIT stack size

The jitstack modifier is your way of setting the maximum stack size used by the just-in-time optimization code. It's ignored if JIT optimization isn't being used, so no stress. The value is a number of kibibytes (units of 1024 bytes). Setting zero reverts to the default of 32KiB. Providing a stack that's larger than the default is necessary only for very complicated patterns. And if jitstack is set non-zero on a subject line, it overrides any value that was set on the pattern. Subject line wins, periodt 👑

🧮 Setting heap, match, and depth limits

The heap_limit, match_limit, and depth_limit modifiers set the appropriate limits in the match context. These values are ignored when the find_limits or find_limits_noheap modifier is specified. Noted 📝

🔍 Finding minimum limits

If the find_limits modifier is present on a subject line, pcre2test calls the relevant matching function several times, setting different values in the match context via pcre2_set_heap_limit(), pcre2_set_match_limit(), or pcre2_set_depth_limit() until it finds the smallest value for each parameter that lets the match complete without a "limit exceeded" error. The match itself may succeed or fail — she's just here for the limits. An alternative modifier, find_limits_noheap, omits the heap limit. This is what's used in the standard tests, because the minimum heap limit varies between systems. If JIT is being used, only the match limit is relevant, and the other two are automatically omitted 🎀

When using this modifier, the pattern should not contain any limit settings such as (*LIMIT_MATCH=...) within it. If such a setting is present and is lower than the minimum matching value, the minimum value cannot be found, because pcre2_set_match_limit() etc. are only able to reduce the value of an in-pattern limit; they cannot increase it. So yeah, that would be a whole mess 😭

For non-DFA matching, the minimum depth_limit number is a measure of how much nested backtracking happens (that is, how deeply the pattern's tree is searched). In the case of DFA matching, depth_limit controls the depth of recursive calls of the internal function that is used for handling pattern recursion, lookaround assertions, and atomic groups.

For non-DFA matching, the match_limit number is a measure of the amount of backtracking that takes place, and learning the minimum value can be instructive — lowkey educational content 🌈. For most simple matches, the number is quite small, but for patterns with very large numbers of matching possibilities, it can become large very quickly with increasing length of subject string. In the case of DFA matching, match_limit controls the total number of calls, both recursive and non-recursive, to the internal matching function, thus controlling the overall amount of computing resource that is used 🔥

For both kinds of matching, the heap_limit number, which is in kibibytes (units of 1024 bytes), limits the amount of heap memory used for matching. Budget your heap, bestie 💜

🏷️ Showing MARK names

The mark modifier makes pcre2test show the names that come back from backtracking control verbs in the calls to pcre2_match(). If a mark gets returned — for a match, a non-match, or a partial match — bestie, you will see it. For a match it sits on its own line, tagged "MK:". Otherwise it just gets tacked onto the non-match message. 💅

✨ Showing memory usage ✨

The memory modifier makes pcre2test log the sizes of all the heap memory allocation and freeing calls that happen during a call to pcre2_match() or pcre2_dfa_match(). In that second case, heap memory only gets used when a match needs more internal workspace than the default allocation on the stack, so ngl a lot of the time you'll get no output at all. And no heap memory is allocated during matching with JIT. For this modifier to actually work, the null_context modifier must not be set on both the pattern and the subject — though it's totally fine to set it on one or the other. 🫶

✨ Showing the heap frame overall vector size ✨

The heapframes_size modifier is relevant for matches using pcre2_match() without JIT. After a match has run (whether it slayed or not), pcre2test shows the size, in bytes, of the allocated heap frames vector that's left attached to the match data block. If the matching action involved several calls to pcre2_match() (like, say, global matching or for timing), only the final value gets shown. ⭐

This modifier is ignored — with a warning — for POSIX or DFA matching. JIT matching doesn't use the heap frames vector at all, so the size is always zero, unless there was a previous non-JIT match. Also note: specifying a size of zero for the output vector (see below) makes pcre2test free its match data block (and the associated heap frames vector) and allocate a new one. 🌀

✨ Setting a starting offset ✨

The offset modifier sets an offset in the subject string where matching starts. Its value is a number of code units, not characters. 👑

✨ Setting an offset limit ✨

The offset_limit modifier sets a limit for unanchored matches. If a match can't be found starting at or before this offset in the subject, you get a "no match" return. The data value is a number of code units, not characters. When you use this modifier, the use_offset_limit modifier must have been set for the pattern — if it wasn't, an error gets generated. 🚨

✨ Setting the size of the output vector ✨

The ovector modifier applies only to the subject line it appears in, though of course you can also use it to set a default in a #subject command. It specifies the number of pairs of offsets available for storing matching information. The default is 15. 🌸

A value of zero is useful when you're testing the POSIX API, because it causes regexec() to be called with a NULL capture vector. When you're not testing the POSIX API, a value of zero is used to make pcre2_match_data_create_from_pattern() get called, so a new match block gets created that's exactly the right size for the pattern. (You literally cannot create a match block with a zero-length ovector — there's always at least one pair of offsets.) The old match data block gets freed. 💀

✨ Passing the subject as zero-terminated ✨

By default, the subject string gets passed to a native API matching function with its correct length. To test the facility for passing a zero-terminated string, there's the zero_terminate modifier. It causes the length to be passed as PCRE2_ZERO_TERMINATED. When matching via the POSIX interface, this modifier is ignored — with a warning. 🐚

When testing pcre2_substitute(), this modifier also has the effect of passing the replacement string as zero-terminated.

✨ Passing a NULL context, subject, or replacement ✨

Normally, pcre2test passes a context block to pcre2_match(), pcre2_dfa_match(), pcre2_jit_match() or pcre2_substitute(). If the null_context modifier is set, though, NULL is passed instead. This exists so you can test that the matching and substitution functions behave correctly in that case (they use default values). This modifier cannot be used with the find_limits, find_limits_noheap, or substitute_callout modifiers. 🙅‍♀️

Similarly, for testing purposes, if the null_subject or null_replacement modifier is set, the subject or replacement string pointers are passed as NULL, respectively, to the relevant functions. 🎀

🦋 THE ALTERNATIVE MATCHING FUNCTION 🦋

By default, pcre2test uses the standard PCRE2 matching function, pcre2_match(), to match each subject line. But PCRE2 also supports an alternative matching function, pcre2_dfa_match(), which operates in a different way and has some restrictions. The differences between the two functions are described in the pcre2matching documentation. 📚

If the dfa modifier is set, the alternative matching function gets used. This function finds all possible matches at a given point in the subject. If, however, the dfa_shortest modifier is set, processing stops after the first match is found — which is always the shortest possible match. ⚡

🔥 DEFAULT OUTPUT FROM pcre2test 🔥

This section describes the output when the normal matching function, pcre2_match(), is being used.

When a match succeeds, pcre2test outputs the list of captured substrings, starting with number 0 for the string that matched the whole pattern. Otherwise, it outputs "No match" when the return is PCRE2_ERROR_NOMATCH, or "Partial match:" followed by the partially matching substring when the return is PCRE2_ERROR_PARTIAL. (Note that this is the entire substring that was inspected during the partial match; it may include characters before the actual match start if a lookbehind assertion, \K, \b, or \B was involved.) 👀

For any other return, pcre2test outputs the PCRE2 negative error number and a short descriptive phrase. If the error is a failed UTF string check, the code unit offset of the start of the failing character is also output. Here's an example of an interactive pcre2test run:

$ pcre2test
PCRE2 version 10.22 2016-07-29

  re> /^abc(\d+)/
data> abc123
 0: abc123
 1: 123
data> xyz
No match

Unset capturing substrings that aren't followed by one that is set are not shown by pcre2test unless the allcaptures modifier is specified. In this next example there are two capturing substrings, but when the first data line is matched, the second, unset substring isn't shown. An "internal" unset substring is shown as <unset>, like in the second data line. 🙈

  re> /(a)|(b)/
data> a
 0: a
 1: a
data> b
 0: b
 1: <unset>
 2: b

If the strings contain any non-printing characters, they get output as \xhh escapes if the value is less than 256 and UTF mode is not set. Otherwise they get output as \x{hh...} escapes. (See below for the definition of non-printing characters.) If the aftertext modifier is set, the output for substring 0 is followed by the rest of the subject string, identified by "0+" like this:

  re> /cat/aftertext
data> cataract
 0: cat
 0+ aract

If global matching is requested, the results of successive matching attempts get output in sequence, like this:

  re> /\Bi(\w\w)/g
data> Mississippi
 0: iss
 1: ss
 0: iss
 1: ss
 0: ipp
 1: pp

"No match" is output only if the first match attempt fails. Here's an example of a failure message (the offset 4 that's specified by the offset modifier is past the end of the subject string):

  re> /xyz/
data> xyz\=offset=4
Error -24 (bad offset value)

Note that whereas patterns can be continued over several lines (a plain ">" prompt is used for continuations), subject lines may not. However, newlines can be included in a subject by means of the \n escape (or \r, \r\n, etc., depending on the newline sequence setting). 📝

🌈 OUTPUT FROM THE ALTERNATIVE MATCHING FUNCTION 🌈

When the alternative matching function, pcre2_dfa_match(), is used, the output consists of a list of all the matches that start at the first point in the subject where there's at least one match. For example:

  re> /(tang|tangerine|tan)/
data> yellow tangerine\=dfa
 0: tangerine
 1: tang
 2: tan

Using the normal matching function on this data finds only "tang". The longest matching string is always given first (and numbered zero). After a PCRE2_ERROR_PARTIAL return, the output is "Partial match:", followed by the partially matching substring. Note that this is the entire substring that was inspected during the partial match; it may include characters before the actual match start if a lookbehind assertion, \b, or \B was involved. (\K is not supported for DFA matching.) 💅

If global matching is requested, the search for further matches resumes at the end of the longest match. For example:

  re> /(tang|tangerine|tan)/g
data> yellow tangerine and tangy sultana\=dfa
 0: tangerine
 1: tang
 2: tan
 0: tang
 1: tan
 0: tan

The alternative matching function does not support substring capture, so the modifiers that are concerned with captured substrings are simply not relevant here. 🚫

🧚 RESTARTING AFTER A PARTIAL MATCH 🧚

When the alternative matching function has given the PCRE2_ERROR_PARTIAL return — indicating that the subject partially matched the pattern — you can restart the match with additional subject data by means of the dfa_restart modifier. For example:

  re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
data> 23ja\=ps,dfa
Partial match: 23ja
data> n05\=dfa,dfa_restart
 0: n05

For further information about partial matching, see the pcre2partial documentation. 🦋

💅 CALLOUTS 💅

Okay bestie, so if your pattern has any callout requests, pcre2test's callout function gets called during matching — unless you specify callout_none, that is. This works with both matching functions and with JIT, though there are some differences in behaviour. The output for callouts with numerical arguments and those with string arguments is slightly different, so let’s break it down. 🫶

🔢 Callouts with numerical arguments

By default, the callout function displays the callout number, the start and current positions in the subject text at the callout time, and the next pattern item to be tested. For example:

         --->pqrabcdef
           0    ^  ^     \d

This output indicates that callout number 0 occurred for a match attempt starting at the fourth character of the subject string, when the pointer was at the seventh character, and when the next pattern item was \d. Just one circumflex is output if the start and current positions are the same, or if the current position precedes the start position — which can happen if the callout is in a lookbehind assertion. 😌

Callouts numbered 255 are assumed to be automatic callouts, inserted as a result of the auto_callout pattern modifier. In this case, instead of showing the callout number, the offset in the pattern, preceded by a plus, is output. For example:

           re> /\d?[A-E]\*/auto_callout
         data> E*
         --->E*
          +0 ^      \d?
          +3 ^      [A-E]
          +8 ^^     \*
         +10 ^ ^
          0: E*

If a pattern contains (*MARK) items, an additional line is output whenever a change of latest mark is passed to the callout function. For example:

           re> /a(*MARK:X)bc/auto_callout
         data> abc
         --->abc
          +0 ^       a
          +1 ^^      (*MARK:X)
         +10 ^^      b
         Latest Mark: X
         +11 ^ ^     c
         +12 ^  ^
          0: abc

The mark changes between matching "a" and "b", but stays the same for the rest of the match, so nothing more is output. If, as a result of backtracking, the mark reverts to being unset, the text <unset> is output. 🎀

🦋 Callouts with string arguments

The output for a callout with a string argument is similar, except that instead of outputting a callout number before the position indicators, the callout string and its offset in the pattern string are output before the reflection of the subject string, and the subject string is reflected for each callout. For example:

           re> /^ab(?C'first')cd(?C"second")ef/
         data> abcdefg
         Callout (7): 'first'
         --->abcdefg
             ^ ^         c
         Callout (20): "second"
         --->abcdefg
             ^   ^       e
          0: abcdef

🌸 Callout modifiers

The callout function in pcre2test returns zero (carry on matching) by default, but you can use a callout_fail modifier in a subject line to change this and other parameters of the callout (see below). ✨

If the callout_capture modifier is set, the current captured groups are output when a callout occurs. This is useful only for non-DFA matching, as pcre2_dfa_match() does not support capturing, so no captures are ever shown.

The normal callout output, showing the callout number or pattern offset (as described above) is suppressed if the callout_no_where modifier is set.

When using the interpretive matching function pcre2_match() without JIT, setting the callout_extra modifier causes additional output from pcre2test's callout function to be generated. For the first callout in a match attempt at a new starting position in the subject, "New match attempt" is output. If there has been a backtrack since the last callout (or start of matching if this is the first callout), "Backtrack" is output, followed by "No other matching paths" if the backtrack ended the previous match attempt. For example:

          re> /(a+)b/auto_callout,no_start_optimize,no_auto_possess
         data> aac\=callout_extra
         New match attempt
         --->aac
          +0 ^       (
          +1 ^       a+
          +3 ^ ^     )
          +4 ^ ^     b
         Backtrack
         --->aac
          +3 ^^      )
          +4 ^^      b
         Backtrack
         No other matching paths
         New match attempt
         --->aac
          +0  ^      (
          +1  ^      a+
          +3  ^^     )
          +4  ^^     b
         Backtrack
         No other matching paths
         New match attempt
         --->aac
          +0   ^     (
          +1   ^     a+
         Backtrack
         No other matching paths
         New match attempt
         --->aac
          +0    ^    (
          +1    ^    a+
         No match

Notice that various optimizations must be turned off if you want all possible matching paths to be scanned. If no_start_optimize is not used, there is an immediate "no match", without any callouts, because the starting optimization fails to find "b" in the subject, which it knows must be present for any match. If no_auto_possess is not used, the "a+" item is turned into "a++", which reduces the number of backtracks. 😭

The callout_extra modifier has no effect if used with the DFA matching function, or with JIT.

👑 Return values from callouts

The default return from the callout function is zero, which allows matching to continue. The callout_fail modifier can be given one or two numbers. If there is only one number, 1 is returned instead of 0 (causing matching to backtrack) when a callout of that number is reached. If two numbers (<n>:<m>) are given, 1 is returned when callout <n> is reached and there have been at least <m> callouts. The callout_error modifier is similar, except that PCRE2_ERROR_CALLOUT is returned, causing the entire matching process to be aborted. If both these modifiers are set for the same callout number, callout_error takes precedence. Note that callouts with string arguments are always given the number zero. 💀

The callout_data modifier can be given an unsigned or a negative number. This is set as the "user data" that is passed to the matching function, and passed back when the callout function is invoked. Any value other than zero is used as a return from pcre2test's callout function.

Inserting callouts can be helpful when using pcre2test to check complicated regular expressions. For further information about callouts, see the pcre2callout documentation.

🌈 NON-PRINTING CHARACTERS 🌈

When pcre2test is outputting text in the compiled version of a pattern, bytes other than 32-126 are always treated as non-printing characters and are therefore shown as hex escapes. 🧚

When pcre2test is outputting text that is a matched part of a subject string, it behaves in the same way, unless a different locale has been set for the pattern (using the locale modifier). In this case, the isprint() function is used to distinguish printing and non-printing characters. 💖

🎀 SAVING AND RESTORING COMPILED PATTERNS 🎀

It is possible to save compiled patterns on disc or elsewhere, and reload them later, subject to a number of restrictions. JIT data cannot be saved. The host on which the patterns are reloaded must be running the same version of PCRE2, with the same code unit width, and must also have the same endianness, pointer width and PCRE2_SIZE type. Before compiled patterns can be saved they must be serialized, that is, converted to a stream of bytes. A single byte stream may contain any number of compiled patterns, but they must all use the same character tables. A single copy of the tables is included in the byte stream (its size is 1088 bytes). ✨

The functions whose names begin with pcre2_serialize_ are used for serializing and de-serializing. They are described in the pcre2serialize documentation. In this section we describe the features of pcre2test that can be used to test these functions.

Note that "serialization" in PCRE2 does not convert compiled patterns to an abstract format like Java or .NET. It just makes a reloadable byte code stream. Hence the restrictions on reloading mentioned above. 😌

In pcre2test, when a pattern with push modifier is successfully compiled, it is pushed onto a stack of compiled patterns, and pcre2test expects the next line to contain a new pattern (or command) instead of a subject line. By contrast, the pushcopy modifier causes a copy of the compiled pattern to be stacked, leaving the original available for immediate matching. By using push and/or pushcopy, a number of patterns can be compiled and retained. These modifiers are incompatible with posix, and control modifiers that act at match time are ignored (with a message) for the stacked patterns. The jitverify modifier applies only at compile time. 🦋

The command

#save <filename>

causes all the stacked patterns to be serialized and the result written to the named file. Afterwards, all the stacked patterns are freed. The command

#load <filename>

reads the data in the file, and then arranges for it to be de-serialized, with the resulting compiled patterns added to the pattern stack. The pattern on the top of the stack can be retrieved by the #pop command, which must be followed by lines of subjects that are to be matched with the pattern, terminated as usual by an empty line or end of file. This command may be followed by a modifier list containing only control modifiers that act after a pattern has been compiled. In particular, hex, posix, posix_nosub, push, and pushcopy are not allowed, nor are any option-setting modifiers. The JIT modifiers are, however permitted. Here is an example that saves and reloads two patterns. 💅

         /abc/push
         /xyz/push
         #save tempfile
         #load tempfile
         #pop info
         xyz

         #pop jit,bincode
         abc

If jitverify is used with #pop, it does not automatically imply jit, which is different behaviour from when it is used on a pattern.

The #popcopy command is analogous to the pushcopy modifier in that it makes current a copy of the topmost stack pattern, leaving the original still on the stack. 🌸

👀 SEE ALSO 👀

pcre2(3), pcre2api(3), pcre2callout(3), pcre2jit, pcre2matching(3), pcre2partial(d), pcre2pattern(3), pcre2serialize(3).

🫶 AUTHOR 🫶

Philip Hazel Retired from University Computing Service Cambridge, England.

⭐ REVISION ⭐

That’s the whole pcre2test(1) page, bestie — PCRE2 10.47, dated 12 October 2025. You absolutely ate that, no cap. 💖✨

📦 where this page comes from

fedora 43: pcre2-tools 10.47-1.fc43 · BSD-3-Clause
fedora 44: pcre2-tools 10.47-1.fc44.1 · BSD-3-Clause
💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash-1017 on 2026-11-06: a language model's rewrite of the original, under its terms