3 # pragma for controlling the regex engine
8 our @ISA = qw(Exporter);
9 our @EXPORT_OK = qw(is_regexp regexp_pattern regmust
11 regnames_count regnames_iterinit regnames_iternext);
12 our %EXPORT_OK = map { $_ => 1 } @EXPORT_OK;
14 # *** WARNING *** WARNING *** WARNING *** WARNING *** WARNING ***
16 # If you modify these values see comment below!
19 taint => 0x00100000, # HINT_RE_TAINT
20 eval => 0x00200000, # HINT_RE_EVAL
23 # - File::Basename contains a literal for 'taint' as a fallback. If
24 # taint is changed here, File::Basename must be updated as well.
26 # - ExtUtils::ParseXS uses a hardcoded
27 # BEGIN { $^H |= 0x00200000 }
28 # in it to allow re.xs to be built. So if 'eval' is changed here then
29 # ExtUtils::ParseXS must be changed as well.
31 # *** WARNING *** WARNING *** WARNING *** WARNING *** WARNING ***
34 eval { # Ignore errors
37 my $terminal = Tgetent Term::Cap ({OSPEED => 9600}); # Avoid warning.
38 my $props = $ENV{PERL_RE_TC} || 'md,me,so,se,us,ue';
39 my @props = split /,/, $props;
40 my $colors = join "\t", map {$terminal->Tputs($_,1)} @props;
43 $ENV{PERL_RE_COLORS} = $colors;
46 $ENV{PERL_RE_COLORS} ||= qq'\t\t> <\t> <\t\t';
66 OFFSETSDBG => 0x040000,
68 OPTIMISEM => 0x100000,
72 $flags{All} = $flags{all} = $flags{DUMP} | $flags{EXECUTE};
73 $flags{Extra} = $flags{EXECUTE} | $flags{COMPILE};
74 $flags{More} = $flags{MORE} = $flags{All} | $flags{TRIEC} | $flags{TRIEM} | $flags{STATE};
75 $flags{State} = $flags{DUMP} | $flags{EXECUTE} | $flags{STATE};
76 $flags{TRIE} = $flags{DUMP} | $flags{EXECUTE} | $flags{TRIEC};
82 if ( ! defined($installed) ) {
84 $installed = eval { XSLoader::load('re', $VERSION) } || 0;
85 $installed_error = $@;
94 die "'re' not installed!? ($installed_error)";
96 # We call install() every time, as if we didn't, we wouldn't
97 # "see" any changes to the color environment var since
98 # the last time it was called.
100 # install() returns an integer, which if casted properly
101 # in C resolves to a structure containing the regex
102 # hooks. Setting it to a random integer will guarantee
104 $^H{regcomp} = install();
116 Carp::carp("Useless use of \"re\" pragma");
118 foreach my $idx (0..$#_){
120 if ($s eq 'Debug' or $s eq 'Debugcolor') {
121 setcolor() if $s =~/color/i;
122 ${^RE_DEBUG_FLAGS} = 0 unless defined ${^RE_DEBUG_FLAGS};
123 for my $idx ($idx+1..$#_) {
124 if ($flags{$_[$idx]}) {
126 ${^RE_DEBUG_FLAGS} |= $flags{$_[$idx]};
128 ${^RE_DEBUG_FLAGS} &= ~ $flags{$_[$idx]};
132 Carp::carp("Unknown \"re\" Debug flag '$_[$idx]', possible flags: ",
133 join(", ",sort keys %flags ) );
136 _load_unload($on ? 1 : ${^RE_DEBUG_FLAGS});
138 } elsif ($s eq 'debug' or $s eq 'debugcolor') {
139 setcolor() if $s =~/color/i;
141 } elsif (exists $bitmask{$s}) {
142 $bits |= $bitmask{$s};
143 } elsif ($EXPORT_OK{$s}) {
146 re->export_to_level(2, 're', $s);
149 Carp::carp("Unknown \"re\" subpragma '$s' (known ones are: ",
150 join(', ', map {qq('$_')} 'debug', 'debugcolor', sort keys %bitmask),
164 $^H &= ~ bits(0, @_);
173 re - Perl pragma to alter regular expression behaviour
178 ($x) = ($^X =~ /^(.*)$/s); # $x is tainted here
180 $pat = '(?{ $foo = 1 })';
182 /foo${pat}bar/; # won't fail (when not under -T switch)
185 no re 'taint'; # the default
186 ($x) = ($^X =~ /^(.*)$/s); # $x is not tainted here
188 no re 'eval'; # the default
189 /foo${pat}bar/; # disallowed (with or without -T switch)
192 use re 'debug'; # output debugging info during
193 /^(.*)$/s; # compile and run time
196 use re 'debugcolor'; # same as 'debug', but with colored output
199 use re qw(Debug All); # Finer tuned debugging options.
200 use re qw(Debug More);
201 no re qw(Debug ALL); # Turn of all re debugging in this scope
203 use re qw(is_regexp regexp_pattern); # import utility functions
204 my ($pat,$mods)=regexp_pattern(qr/foo/i);
205 if (is_regexp($obj)) {
206 print "Got regexp: ",
207 scalar regexp_pattern($obj); # just as perl would stringify it
208 } # but no hassle with blessed re's.
210 (We use $^X in these examples because it's tainted by default.)
216 When C<use re 'taint'> is in effect, and a tainted string is the target
217 of a regex, the regex memories (or values returned by the m// operator
218 in list context) are tainted. This feature is useful when regex operations
219 on tainted data aren't meant to extract safe substrings, but to perform
220 other transformations.
224 When C<use re 'eval'> is in effect, a regex is allowed to contain
225 C<(?{ ... })> zero-width assertions even if regular expression contains
226 variable interpolation. That is normally disallowed, since it is a
227 potential security risk. Note that this pragma is ignored when the regular
228 expression is obtained from tainted data, i.e. evaluation is always
229 disallowed with tainted regular expressions. See L<perlre/(?{ code })>.
231 For the purpose of this pragma, interpolation of precompiled regular
232 expressions (i.e., the result of C<qr//>) is I<not> considered variable
237 I<is> allowed if $pat is a precompiled regular expression, even
238 if $pat contains C<(?{ ... })> assertions.
242 When C<use re 'debug'> is in effect, perl emits debugging messages when
243 compiling and using regular expressions. The output is the same as that
244 obtained by running a C<-DDEBUGGING>-enabled perl interpreter with the
245 B<-Dr> switch. It may be quite voluminous depending on the complexity
246 of the match. Using C<debugcolor> instead of C<debug> enables a
247 form of output that can be used to get a colorful display on terminals
248 that understand termcap color sequences. Set C<$ENV{PERL_RE_TC}> to a
249 comma-separated list of C<termcap> properties to use for highlighting
250 strings on/off, pre-point part on/off.
251 See L<perldebug/"Debugging regular expressions"> for additional info.
253 As of 5.9.5 the directive C<use re 'debug'> and its equivalents are
254 lexically scoped, as the other directives are. However they have both
255 compile-time and run-time effects.
257 See L<perlmodlib/Pragmatic Modules>.
261 Similarly C<use re 'Debug'> produces debugging output, the difference
262 being that it allows the fine tuning of what debugging output will be
263 emitted. Options are divided into three groups, those related to
264 compilation, those related to execution and those related to special
265 purposes. The options are as follows:
269 =item Compile related options
275 Turns on all compile related debug options.
279 Turns on debug output related to the process of parsing the pattern.
283 Enables output related to the optimisation phase of compilation.
287 Detailed info about trie compilation.
291 Dump the final program out after it is compiled and optimised.
295 =item Execute related options
301 Turns on all execute related debug options.
305 Turns on debugging of the main matching loop.
309 Extra debugging of how tries execute.
313 Enable debugging of start point optimisations.
317 =item Extra debugging options
323 Turns on all "extra" debugging options.
327 Enable enhanced TRIE debugging. Enhances both TRIEE
332 Enable debugging of states in the engine.
336 Enable debugging of the recursion stack in the engine. Enabling
337 or disabling this option automatically does the same for debugging
338 states as well. This output from this can be quite large.
342 Enable enhanced optimisation debugging and start point optimisations.
343 Probably not useful except when debugging the regex engine itself.
347 Dump offset information. This can be used to see how regops correlate
348 to the pattern. Output format is
350 NODENUM:POSITION[LENGTH]
352 Where 1 is the position of the first char in the string. Note that position
353 can be 0, or larger than the actual length of the pattern, likewise length
358 Enable debugging of offsets information. This emits copious
359 amounts of trace information and doesn't mesh well with other
362 Almost definitely only useful to people hacking
363 on the offsets part of the debug engine.
367 =item Other useful flags
369 These are useful shortcuts to save on the typing.
375 Enable all compile and execute options at once.
379 Enable DUMP and all execute options. Equivalent to:
387 Enable TRIEM and all execute compile and execute options.
393 As of 5.9.5 the directive C<use re 'debug'> and its equivalents are
394 lexically scoped, as the other directives are. However they have both
395 compile-time and run-time effects.
397 =head2 Exportable Functions
399 As of perl 5.9.5 're' debug contains a number of utility functions that
400 may be optionally exported into the caller's namespace. They are listed
405 =item is_regexp($ref)
407 Returns true if the argument is a compiled regular expression as returned
408 by C<qr//>, false if it is not.
410 This function will not be confused by overloading or blessing. In
411 internals terms, this extracts the regexp pointer out of the
412 PERL_MAGIC_qr structure so it it cannot be fooled.
414 =item regexp_pattern($ref)
416 If the argument is a compiled regular expression as returned by C<qr//>,
417 then this function returns the pattern.
419 In list context it returns a two element list, the first element
420 containing the pattern and the second containing the modifiers used when
421 the pattern was compiled.
423 my ($pat, $mods) = regexp_pattern($ref);
425 In scalar context it returns the same as perl would when strigifying a raw
426 C<qr//> with the same pattern inside. If the argument is not a compiled
427 reference then this routine returns false but defined in scalar context,
428 and the empty list in list context. Thus the following
430 if (regexp_pattern($ref) eq '(?i-xsm:foo)')
432 will be warning free regardless of what $ref actually is.
434 Like C<is_regexp> this function will not be confused by overloading
435 or blessing of the object.
439 If the argument is a compiled regular expression as returned by C<qr//>,
440 then this function returns what the optimiser consiers to be the longest
441 anchored fixed string and longest floating fixed string in the pattern.
443 A I<fixed string> is defined as being a substring that must appear for the
444 pattern to match. An I<anchored fixed string> is a fixed string that must
445 appear at a particular offset from the beginning of the match. A I<floating
446 fixed string> is defined as a fixed string that can appear at any point in
447 a range of positions relative to the start of the match. For example,
449 my $qr = qr/here .* there/x;
450 my ($anchored, $floating) = regmust($qr);
451 print "anchored:'$anchored'\nfloating:'$floating'\n";
458 Because the C<here> is before the C<.*> in the pattern, its position
459 can be determined exactly. That's not true, however, for the C<there>;
460 it could appear at any point after where the anchored string appeared.
461 Perl uses both for its optimisations, prefering the longer, or, if they are
464 B<NOTE:> This may not necessarily be the definitive longest anchored and
465 floating string. This will be what the optimiser of the Perl that you
466 are using thinks is the longest. If you believe that the result is wrong
467 please report it via the L<perlbug> utility.
469 =item regname($name,$qr,$all)
471 Returns the contents of a named buffer. If $qr is missing, or is not the
472 result of a qr// then returns the result of the last successful match. If
473 $all is true then returns an array ref containing one entry per buffer,
474 otherwise returns the first defined buffer.
476 =item regnames($qr,$all)
478 Returns a list of all of the named buffers defined in a pattern. If
479 $all is true then it returns all names defined, if not returns only
480 names which were involved in the last successful match. If $qr is omitted
481 or is not the result of a qr// then returns the details for the last
484 =item regnames_iterinit($qr)
486 Initializes the internal hash iterator associated to a regexps named capture
487 buffers. If $qr is omitted resets the iterator associated with the regexp used
488 in the last successful match.
490 =item regnames_iternext($qr,$all)
492 Gets the next key from the hash associated with a regexp. If $qr
493 is omitted resets the iterator associated with the regexp used in the
494 last successful match. If $all is true returns the keys of all of the
495 distinct named buffers in the pattern, if not returns only those names
496 used in the last successful match.
498 =item regnames_count($qr)
500 Returns the number of distinct names defined in the regexp $qr. If
501 $qr is omitted or not a regexp returns the count of names in the
502 last successful match.
504 B<Note:> that this result is always the actual number of distinct
505 named buffers defined, it may not actually match that which is
506 returned by C<regnames()> and related routines when those routines
507 have not been called with the $all parameter set..
513 L<perlmodlib/Pragmatic Modules>.