This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regcomp.sym: Add node type ANYOF_POSIXL
[perl5.git] / pod / perldebguts.pod
CommitLineData
055fd3a9
GS
1=head1 NAME
2
3perldebguts - Guts of Perl debugging
4
5=head1 DESCRIPTION
6
ba555bf5 7This is not L<perldebug>, which tells you how to use
74410c12
JM
8the debugger. This manpage describes low-level details concerning
9the debugger's internals, which range from difficult to impossible
10to understand for anyone who isn't incredibly intimate with Perl's guts.
11Caveat lector.
055fd3a9
GS
12
13=head1 Debugger Internals
14
15Perl has special debugging hooks at compile-time and run-time used
16to create debugging environments. These hooks are not to be confused
4375e838
GS
17with the I<perl -Dxxx> command described in L<perlrun>, which is
18usable only if a special Perl is built per the instructions in the
055fd3a9
GS
19F<INSTALL> podpage in the Perl source tree.
20
21For example, whenever you call Perl's built-in C<caller> function
74410c12
JM
22from the package C<DB>, the arguments that the corresponding stack
23frame was called with are copied to the C<@DB::args> array. These
24mechanisms are enabled by calling Perl with the B<-d> switch.
25Specifically, the following additional features are enabled
26(cf. L<perlvar/$^P>):
055fd3a9 27
13a2d996 28=over 4
055fd3a9
GS
29
30=item *
31
32Perl inserts the contents of C<$ENV{PERL5DB}> (or C<BEGIN {require
33'perl5db.pl'}> if not present) before the first line of your program.
34
35=item *
36
aa0b556f 37Each array C<@{"_<$filename"}> holds the lines of $filename for a
74410c12
JM
38file compiled by Perl. The same is also true for C<eval>ed strings
39that contain subroutines, or which are currently being executed.
40The $filename for C<eval>ed strings looks like C<(eval 34)>.
8894c26d
MJD
41
42Values in this array are magical in numeric context: they compare
43equal to zero only if the line is not breakable.
055fd3a9
GS
44
45=item *
46
aa0b556f 47Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed
055fd3a9
GS
48by line number. Individual entries (as opposed to the whole hash)
49are settable. Perl only cares about Boolean true here, although
50the values used by F<perl5db.pl> have the form
8894c26d 51C<"$break_condition\0$action">.
055fd3a9
GS
52
53The same holds for evaluated strings that contain subroutines, or
54which are currently being executed. The $filename for C<eval>ed strings
d24ca0c5 55looks like C<(eval 34)>.
055fd3a9
GS
56
57=item *
58
aa0b556f 59Each scalar C<${"_<$filename"}> contains C<"_<$filename">. This is
055fd3a9
GS
60also the case for evaluated strings that contain subroutines, or
61which are currently being executed. The $filename for C<eval>ed
d24ca0c5 62strings looks like C<(eval 34)>.
055fd3a9
GS
63
64=item *
65
66After each C<require>d file is compiled, but before it is executed,
67C<DB::postponed(*{"_<$filename"})> is called if the subroutine
68C<DB::postponed> exists. Here, the $filename is the expanded name of
69the C<require>d file, as found in the values of %INC.
70
71=item *
72
73After each subroutine C<subname> is compiled, the existence of
74C<$DB::postponed{subname}> is checked. If this key exists,
75C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine
76also exists.
77
78=item *
79
80A hash C<%DB::sub> is maintained, whose keys are subroutine names
81and whose values have the form C<filename:startline-endline>.
82C<filename> has the form C<(eval 34)> for subroutines defined inside
d24ca0c5 83C<eval>s.
055fd3a9
GS
84
85=item *
86
87When the execution of your program reaches a point that can hold a
74410c12
JM
88breakpoint, the C<DB::DB()> subroutine is called if any of the variables
89C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true. These variables
055fd3a9
GS
90are not C<local>izable. This feature is disabled when executing
91inside C<DB::DB()>, including functions called from it
92unless C<< $^D & (1<<30) >> is true.
93
94=item *
95
96When execution of the program reaches a subroutine call, a call to
5109e49f
Z
97C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> set to identify
98the called subroutine. (This doesn't happen if the calling subroutine
99was compiled in the C<DB> package.) C<$DB::sub> normally holds the name
100of the called subroutine, if it has a name by which it can be looked up.
101Failing that, C<$DB::sub> will hold a reference to the called subroutine.
102Either way, the C<&DB::sub> subroutine can use C<$DB::sub> as a reference
103by which to call the called subroutine, which it will normally want to do.
055fd3a9 104
77e42cd2
TC
105X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub>
106is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling
107back to C<&DB::sub>(I<args>).
108
261cbad1
TC
109=item *
110
5109e49f
Z
111When execution of the program uses C<goto> to enter a non-XS subroutine
112and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto> is made, with
113C<$DB::sub> set to identify the subroutine being entered. The call to
114C<&DB::goto> does not replace the C<goto>; the requested subroutine will
115still be entered once C<&DB::goto> has returned. C<$DB::sub> normally
116holds the name of the subroutine being entered, if it has one. Failing
117that, C<$DB::sub> will hold a reference to the subroutine being entered.
118Unlike when C<&DB::sub> is called, it is not guaranteed that C<$DB::sub>
119can be used as a reference to operate on the subroutine being entered.
261cbad1 120
055fd3a9
GS
121=back
122
123Note that if C<&DB::sub> needs external data for it to work, no
74410c12
JM
124subroutine call is possible without it. As an example, the standard
125debugger's C<&DB::sub> depends on the C<$DB::deep> variable
126(it defines how many levels of recursion deep into the debugger you can go
127before a mandatory break). If C<$DB::deep> is not defined, subroutine
128calls are not possible, even though C<&DB::sub> exists.
055fd3a9
GS
129
130=head2 Writing Your Own Debugger
131
74410c12 132=head3 Environment Variables
666f95b9 133
74410c12
JM
134The C<PERL5DB> environment variable can be used to define a debugger.
135For example, the minimal "working" debugger (it actually doesn't do anything)
136consists of one line:
666f95b9 137
055fd3a9
GS
138 sub DB::DB {}
139
74410c12 140It can easily be defined like this:
666f95b9 141
055fd3a9
GS
142 $ PERL5DB="sub DB::DB {}" perl -d your-script
143
74410c12 144Another brief debugger, slightly more useful, can be created
055fd3a9
GS
145with only the line:
146
147 sub DB::DB {print ++$i; scalar <STDIN>}
148
74410c12
JM
149This debugger prints a number which increments for each statement
150encountered and waits for you to hit a newline before continuing
151to the next statement.
666f95b9 152
74410c12 153The following debugger is actually useful:
666f95b9 154
055fd3a9
GS
155 {
156 package DB;
157 sub DB {}
158 sub sub {print ++$i, " $sub\n"; &$sub}
159 }
160
74410c12
JM
161It prints the sequence number of each subroutine call and the name of the
162called subroutine. Note that C<&DB::sub> is being compiled into the
163package C<DB> through the use of the C<package> directive.
055fd3a9 164
74410c12
JM
165When it starts, the debugger reads your rc file (F<./.perldb> or
166F<~/.perldb> under Unix), which can set important options.
167(A subroutine (C<&afterinit>) can be defined here as well; it is executed
168after the debugger completes its own initialization.)
055fd3a9
GS
169
170After the rc file is read, the debugger reads the PERLDB_OPTS
74410c12
JM
171environment variable and uses it to set debugger options. The
172contents of this variable are treated as if they were the argument
96090e4f 173of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
74410c12 174
7b406369 175=head3 Debugger Internal Variables
25cf7dea 176
74410c12
JM
177In addition to the file and subroutine-related variables mentioned above,
178the debugger also maintains various magical internal variables.
179
180=over 4
181
182=item *
055fd3a9 183
74410c12
JM
184C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
185holds the lines of the currently-selected file (compiled by Perl), either
186explicitly chosen with the debugger's C<f> command, or implicitly by flow
187of execution.
188
189Values in this array are magical in numeric context: they compare
190equal to zero only if the line is not breakable.
191
192=item *
193
7b406369 194C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
74410c12
JM
195contains breakpoints and actions keyed by line number in
196the currently-selected file, either explicitly chosen with the
055fd3a9
GS
197debugger's C<f> command, or implicitly by flow of execution.
198
74410c12
JM
199As previously noted, individual entries (as opposed to the whole hash)
200are settable. Perl only cares about Boolean true here, although
201the values used by F<perl5db.pl> have the form
202C<"$break_condition\0$action">.
203
204=back
205
7b406369 206=head3 Debugger Customization Functions
74410c12
JM
207
208Some functions are provided to simplify customization.
209
210=over 4
211
212=item *
213
71110851
RGS
214See L<perldebug/"Configurable Options"> for a description of options parsed by
215C<DB::parse_options(string)>.
74410c12
JM
216
217=item *
218
219C<DB::dump_trace(skip[,count])> skips the specified number of frames
220and returns a list containing information about the calling frames (all
221of them, if C<count> is missing). Each entry is reference to a hash
222with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
055fd3a9
GS
223name, or info about C<eval>), C<args> (C<undef> or a reference to
224an array), C<file>, and C<line>.
225
74410c12
JM
226=item *
227
228C<DB::print_trace(FH, skip[, count[, short]])> prints
055fd3a9
GS
229formatted info about caller frames. The last two functions may be
230convenient as arguments to C<< < >>, C<< << >> commands.
231
74410c12
JM
232=back
233
055fd3a9
GS
234Note that any variables and functions that are not documented in
235this manpages (or in L<perldebug>) are considered for internal
236use only, and as such are subject to change without notice.
237
238=head1 Frame Listing Output Examples
239
240The C<frame> option can be used to control the output of frame
241information. For example, contrast this expression trace:
242
243 $ perl -de 42
244 Stack dump during die enabled outside of evals.
245
246 Loading DB routines from perl5db.pl patch level 0.94
247 Emacs support available.
248
ccf3535a 249 Enter h or 'h h' for help.
055fd3a9
GS
250
251 main::(-e:1): 0
252 DB<1> sub foo { 14 }
253
254 DB<2> sub bar { 3 }
255
256 DB<3> t print foo() * bar()
257 main::((eval 172):3): print foo() + bar();
258 main::foo((eval 168):2):
259 main::bar((eval 170):2):
260 42
261
492652be 262with this one, once the C<o>ption C<frame=2> has been set:
055fd3a9 263
492652be 264 DB<4> o f=2
055fd3a9
GS
265 frame = '2'
266 DB<5> t print foo() * bar()
267 3: foo() * bar()
268 entering main::foo
269 2: sub foo { 14 };
270 exited main::foo
271 entering main::bar
272 2: sub bar { 3 };
273 exited main::bar
274 42
275
276By way of demonstration, we present below a laborious listing
277resulting from setting your C<PERLDB_OPTS> environment variable to
278the value C<f=n N>, and running I<perl -d -V> from the command line.
7b406369
FC
279Examples using various values of C<n> are shown to give you a feel
280for the difference between settings. Long though it may be, this
055fd3a9
GS
281is not a complete listing, but only excerpts.
282
283=over 4
284
285=item 1
286
f185f654
KW
287 entering main::BEGIN
288 entering Config::BEGIN
289 Package lib/Exporter.pm.
290 Package lib/Carp.pm.
291 Package lib/Config.pm.
292 entering Config::TIEHASH
293 entering Exporter::import
294 entering Exporter::export
295 entering Config::myconfig
296 entering Config::FETCH
297 entering Config::FETCH
298 entering Config::FETCH
299 entering Config::FETCH
055fd3a9
GS
300
301=item 2
302
f185f654
KW
303 entering main::BEGIN
304 entering Config::BEGIN
305 Package lib/Exporter.pm.
306 Package lib/Carp.pm.
307 exited Config::BEGIN
308 Package lib/Config.pm.
309 entering Config::TIEHASH
310 exited Config::TIEHASH
311 entering Exporter::import
312 entering Exporter::export
313 exited Exporter::export
314 exited Exporter::import
315 exited main::BEGIN
316 entering Config::myconfig
317 entering Config::FETCH
318 exited Config::FETCH
319 entering Config::FETCH
320 exited Config::FETCH
321 entering Config::FETCH
055fd3a9 322
d5e42f17 323=item 3
055fd3a9 324
f185f654
KW
325 in $=main::BEGIN() from /dev/null:0
326 in $=Config::BEGIN() from lib/Config.pm:2
327 Package lib/Exporter.pm.
328 Package lib/Carp.pm.
329 Package lib/Config.pm.
330 in $=Config::TIEHASH('Config') from lib/Config.pm:644
331 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
332 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
333 in @=Config::myconfig() from /dev/null:0
334 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
335 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
336 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
337 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
338 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
339 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
055fd3a9 340
d5e42f17 341=item 4
055fd3a9 342
f185f654
KW
343 in $=main::BEGIN() from /dev/null:0
344 in $=Config::BEGIN() from lib/Config.pm:2
345 Package lib/Exporter.pm.
346 Package lib/Carp.pm.
347 out $=Config::BEGIN() from lib/Config.pm:0
348 Package lib/Config.pm.
349 in $=Config::TIEHASH('Config') from lib/Config.pm:644
350 out $=Config::TIEHASH('Config') from lib/Config.pm:644
351 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
352 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
353 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
354 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
355 out $=main::BEGIN() from /dev/null:0
356 in @=Config::myconfig() from /dev/null:0
357 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
358 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
359 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
360 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
361 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
362 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
363 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
055fd3a9 364
d5e42f17 365=item 5
055fd3a9 366
f185f654
KW
367 in $=main::BEGIN() from /dev/null:0
368 in $=Config::BEGIN() from lib/Config.pm:2
369 Package lib/Exporter.pm.
370 Package lib/Carp.pm.
371 out $=Config::BEGIN() from lib/Config.pm:0
372 Package lib/Config.pm.
373 in $=Config::TIEHASH('Config') from lib/Config.pm:644
374 out $=Config::TIEHASH('Config') from lib/Config.pm:644
375 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
376 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
377 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
378 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
379 out $=main::BEGIN() from /dev/null:0
380 in @=Config::myconfig() from /dev/null:0
381 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
382 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
383 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
384 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
055fd3a9 385
d5e42f17 386=item 6
055fd3a9 387
f185f654
KW
388 in $=CODE(0x15eca4)() from /dev/null:0
389 in $=CODE(0x182528)() from lib/Config.pm:2
390 Package lib/Exporter.pm.
391 out $=CODE(0x182528)() from lib/Config.pm:0
392 scalar context return from CODE(0x182528): undef
393 Package lib/Config.pm.
394 in $=Config::TIEHASH('Config') from lib/Config.pm:628
395 out $=Config::TIEHASH('Config') from lib/Config.pm:628
396 scalar context return from Config::TIEHASH: empty hash
397 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
398 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
399 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
400 scalar context return from Exporter::export: ''
401 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
402 scalar context return from Exporter::import: ''
055fd3a9
GS
403
404=back
405
406In all cases shown above, the line indentation shows the call tree.
407If bit 2 of C<frame> is set, a line is printed on exit from a
408subroutine as well. If bit 4 is set, the arguments are printed
409along with the caller info. If bit 8 is set, the arguments are
410printed even if they are tied or references. If bit 16 is set, the
411return value is printed, too.
412
413When a package is compiled, a line like this
414
415 Package lib/Carp.pm.
416
417is printed with proper indentation.
418
7b406369 419=head1 Debugging Regular Expressions
055fd3a9
GS
420
421There are two ways to enable debugging output for regular expressions.
422
423If your perl is compiled with C<-DDEBUGGING>, you may use the
424B<-Dr> flag on the command line.
425
426Otherwise, one can C<use re 'debug'>, which has effects at
3d71525d
NJ
427compile time and run time. Since Perl 5.9.5, this pragma is lexically
428scoped.
055fd3a9 429
7b406369 430=head2 Compile-time Output
055fd3a9
GS
431
432The debugging output at compile time looks like this:
433
ccf3535a 434 Compiling REx '[bc]d(ef*g)+h[ij]k$'
1c102323
MJD
435 size 45 Got 364 bytes for offset annotations.
436 first at 1
437 rarest char g at 0
438 rarest char d at 0
439 1: ANYOF[bc](12)
440 12: EXACT <d>(14)
441 14: CURLYX[0] {1,32767}(28)
442 16: OPEN1(18)
443 18: EXACT <e>(20)
444 20: STAR(23)
445 21: EXACT <f>(0)
446 23: EXACT <g>(25)
447 25: CLOSE1(27)
448 27: WHILEM[1/1](0)
449 28: NOTHING(29)
450 29: EXACT <h>(31)
451 31: ANYOF[ij](42)
452 42: EXACT <k>(44)
453 44: EOL(45)
454 45: END(0)
ccf3535a
JK
455 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
456 stclass 'ANYOF[bc]' minlen 7
1c102323
MJD
457 Offsets: [45]
458 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
459 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
460 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
461 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
462 Omitting $` $& $' support.
055fd3a9
GS
463
464The first line shows the pre-compiled form of the regex. The second
465shows the size of the compiled form (in arbitrary units, usually
1c102323
MJD
4664-byte words) and the total number of bytes allocated for the
467offset/length table, usually 4+C<size>*8. The next line shows the
468label I<id> of the first node that does a match.
055fd3a9 469
1c102323
MJD
470The
471
ccf3535a
JK
472 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
473 stclass 'ANYOF[bc]' minlen 7
1c102323
MJD
474
475line (split into two lines above) contains optimizer
055fd3a9
GS
476information. In the example shown, the optimizer found that the match
477should contain a substring C<de> at offset 1, plus substring C<gh>
478at some offset between 3 and infinity. Moreover, when checking for
479these substrings (to abandon impossible matches quickly), Perl will check
480for the substring C<gh> before checking for the substring C<de>. The
481optimizer may also use the knowledge that the match starts (at the
1c102323
MJD
482C<first> I<id>) with a character class, and no string
483shorter than 7 characters can possibly match.
055fd3a9 484
1c102323 485The fields of interest which may appear in this line are
055fd3a9 486
13a2d996 487=over 4
055fd3a9
GS
488
489=item C<anchored> I<STRING> C<at> I<POS>
490
491=item C<floating> I<STRING> C<at> I<POS1..POS2>
492
493See above.
494
495=item C<matching floating/anchored>
496
497Which substring to check first.
498
499=item C<minlen>
500
501The minimal length of the match.
502
503=item C<stclass> I<TYPE>
504
505Type of first matching node.
506
507=item C<noscan>
508
509Don't scan for the found substrings.
510
511=item C<isall>
512
1c102323 513Means that the optimizer information is all that the regular
055fd3a9
GS
514expression contains, and thus one does not need to enter the regex engine at
515all.
516
517=item C<GPOS>
518
519Set if the pattern contains C<\G>.
520
521=item C<plus>
522
523Set if the pattern starts with a repeated char (as in C<x+y>).
524
525=item C<implicit>
526
527Set if the pattern starts with C<.*>.
528
529=item C<with eval>
530
531Set if the pattern contain eval-groups, such as C<(?{ code })> and
532C<(??{ code })>.
533
534=item C<anchored(TYPE)>
535
7b406369 536If the pattern may match only at a handful of places, with C<TYPE>
d3d47aac 537being C<SBOL>, C<MBOL>, or C<GPOS>. See the table below.
055fd3a9
GS
538
539=back
540
541If a substring is known to match at end-of-line only, it may be
ccf3535a 542followed by C<$>, as in C<floating 'k'$>.
055fd3a9 543
1c102323
MJD
544The optimizer-specific information is used to avoid entering (a slow) regex
545engine on strings that will not definitely match. If the C<isall> flag
055fd3a9
GS
546is set, a call to the regex engine may be avoided even when the optimizer
547found an appropriate place for the match.
548
1c102323 549Above the optimizer section is the list of I<nodes> of the compiled
055fd3a9
GS
550form of the regex. Each line has format
551
552C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
553
7b406369 554=head2 Types of Nodes
055fd3a9 555
78465a4b 556Here are the current possible types, with short descriptions:
055fd3a9 557
65aa4ca7
FC
558=for comment
559This table is generated by regen/regcomp.pl. Any changes made here
560will be lost.
561
562=for regcomp.pl begin
563
5da6b59a
KW
564 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
565
566 # Exit points
65aa4ca7 567
89829bb5
KW
568 END no End of program.
569 SUCCEED no Return from a subroutine, basically.
5da6b59a 570
d3d47aac 571 # Line Start Anchors:
89829bb5
KW
572 SBOL no Match "" at beginning of line: /^/, /\A/
573 MBOL no Same, assuming multiline: /^/m
5da6b59a 574
d3d47aac 575 # Line End Anchors:
89829bb5
KW
576 SEOL no Match "" at end of line: /$/
577 MEOL no Same, assuming multiline: /$/m
578 EOS no Match "" at end of string: /\z/
d3d47aac
YO
579
580 # Match Start Anchors:
89829bb5 581 GPOS no Matches where last m//g left off.
d3d47aac
YO
582
583 # Word Boundary Opcodes:
89829bb5
KW
584 BOUND no Like BOUNDA for non-utf8, otherwise match
585 "" between any Unicode \w\W or \W\w
586 BOUNDL no Like BOUND/BOUNDU, but \w and \W are
587 defined by current locale
588 BOUNDU no Match "" at any boundary of a given type
589 using Unicode rules
590 BOUNDA no Match "" at any boundary between \w\W or
591 \W\w, where \w is [_a-zA-Z0-9]
592 NBOUND no Like NBOUNDA for non-utf8, otherwise match
593 "" between any Unicode \w\w or \W\W
594 NBOUNDL no Like NBOUND/NBOUNDU, but \w and \W are
595 defined by current locale
596 NBOUNDU no Match "" at any non-boundary of a given
597 type using using Unicode rules
598 NBOUNDA no Match "" betweeen any \w\w or \W\W, where
599 \w is [_a-zA-Z0-9]
5da6b59a
KW
600
601 # [Special] alternatives:
89829bb5
KW
602 REG_ANY no Match any one character (except newline).
603 SANY no Match any one character.
604 ANYOF sv 1 Match character in (or not in) this class,
605 single char match only
606 ANYOFD sv 1 Like ANYOF, but /d is in effect
607 ANYOFL sv 1 Like ANYOF, but /l is in effect
3edce4f5
KW
608 ANYOFPOSIXL sv 1 Like ANYOFL, but matches [[:posix:]]
609 classes
89829bb5
KW
610 ANYOFM byte 1 Like ANYOF, but matches an invariant byte
611 as determined by the mask and arg
7bc66b18 612
d3d47aac 613 # POSIX Character Classes:
89829bb5
KW
614 POSIXD none Some [[:class:]] under /d; the FLAGS field
615 gives which one
616 POSIXL none Some [[:class:]] under /l; the FLAGS field
617 gives which one
618 POSIXU none Some [[:class:]] under /u; the FLAGS field
619 gives which one
620 POSIXA none Some [[:class:]] under /a; the FLAGS field
621 gives which one
622 NPOSIXD none complement of POSIXD, [[:^class:]]
623 NPOSIXL none complement of POSIXL, [[:^class:]]
624 NPOSIXU none complement of POSIXU, [[:^class:]]
625 NPOSIXA none complement of POSIXA, [[:^class:]]
626
627 ASCII none [[:ascii:]]
628 NASCII none [[:^ascii:]]
629
630 CLUMP no Match any extended grapheme cluster
631 sequence
5da6b59a
KW
632
633 # Alternation
634
65aa4ca7
FC
635 # BRANCH The set of branches constituting a single choice are
636 # hooked together with their "next" pointers, since
637 # precedence prevents anything being concatenated to
638 # any individual branch. The "next" pointer of the last
639 # BRANCH in a choice points to the thing following the
640 # whole choice. This is also where the final "next"
641 # pointer of each individual branch points; each branch
642 # starts with the operand node of a BRANCH node.
5da6b59a 643 #
89829bb5 644 BRANCH node Match this alternative, or the next...
5da6b59a 645
5da6b59a
KW
646 # Literals
647
89829bb5
KW
648 EXACT str Match this string (preceded by length).
649 EXACTL str Like EXACT, but /l is in effect (used so
650 locale-related warnings can be checked
651 for).
652 EXACTF str Match this non-UTF-8 string (not guaranteed
653 to be folded) using /id rules (w/len).
654 EXACTFL str Match this string (not guaranteed to be
655 folded) using /il rules (w/len).
656 EXACTFU str Match this string (folded iff in UTF-8,
657 length in folding doesn't change if not in
658 UTF-8) using /iu rules (w/len).
659 EXACTFAA str Match this string (not guaranteed to be
660 folded) using /iaa rules (w/len).
661
662 EXACTFU_SS str Match this string (folded iff in UTF-8,
663 length in folding may change even if not in
664 UTF-8) using /iu rules (w/len).
eb08fbc1 665 EXACTFLU8 str Rare circumstances: like EXACTFU, but is
89829bb5
KW
666 under /l, UTF-8, folded, and everything in
667 it is above 255.
668 EXACTFAA_NO_TRIE str Match this string (which is not trie-able;
669 not guaranteed to be folded) using /iaa
670 rules (w/len).
5da6b59a
KW
671
672 # Do nothing types
673
89829bb5 674 NOTHING no Match empty string.
5da6b59a 675 # A variant of above which delimits a group, thus stops optimizations
89829bb5
KW
676 TAIL no Match empty string. Can jump here from
677 outside.
5da6b59a
KW
678
679 # Loops
680
65aa4ca7 681 # STAR,PLUS '?', and complex '*' and '+', are implemented as
62e6ef33 682 # circular BRANCH structures. Simple cases
65aa4ca7
FC
683 # (one character per match) are implemented with STAR
684 # and PLUS for speed and to minimize recursive plunges.
5da6b59a 685 #
89829bb5
KW
686 STAR node Match this (simple) thing 0 or more times.
687 PLUS node Match this (simple) thing 1 or more times.
7bc66b18 688
89829bb5
KW
689 CURLY sv 2 Match this simple thing {n,m} times.
690 CURLYN no 2 Capture next-after-this simple thing
691 CURLYM no 2 Capture this medium-complex thing {n,m}
692 times.
693 CURLYX sv 2 Match this complex thing {n,m} times.
5da6b59a
KW
694
695 # This terminator creates a loop structure for CURLYX
89829bb5
KW
696 WHILEM no Do curly processing and see if rest
697 matches.
5da6b59a
KW
698
699 # Buffer related
700
701 # OPEN,CLOSE,GROUPP ...are numbered at compile time.
89829bb5
KW
702 OPEN num 1 Mark this point in input as start of #n.
703 CLOSE num 1 Close corresponding OPEN of #n.
704 SROPEN none Same as OPEN, but for script run
705 SRCLOSE none Close preceding SROPEN
706
707 REF num 1 Match some already matched string
708 REFF num 1 Match already matched string, folded using
709 native charset rules for non-utf8
710 REFFL num 1 Match already matched string, folded in
711 loc.
712 REFFU num 1 Match already matched string, folded using
713 unicode rules for non-utf8
714 REFFA num 1 Match already matched string, folded using
715 unicode rules for non-utf8, no mixing
716 ASCII, non-ASCII
65aa4ca7
FC
717
718 # Named references. Code in regcomp.c assumes that these all are after
719 # the numbered references
89829bb5
KW
720 NREF no-sv 1 Match some already matched string
721 NREFF no-sv 1 Match already matched string, folded using
722 native charset rules for non-utf8
723 NREFFL no-sv 1 Match already matched string, folded in
724 loc.
725 NREFFU num 1 Match already matched string, folded using
726 unicode rules for non-utf8
727 NREFFA num 1 Match already matched string, folded using
728 unicode rules for non-utf8, no mixing
729 ASCII, non-ASCII
7bc66b18 730
d3d47aac 731 # Support for long RE
89829bb5
KW
732 LONGJMP off 1 1 Jump far away.
733 BRANCHJ off 1 1 BRANCH with long offset.
d3d47aac
YO
734
735 # Special Case Regops
89829bb5
KW
736 IFMATCH off 1 2 Succeeds if the following matches.
737 UNLESSM off 1 2 Fails if the following matches.
738 SUSPEND off 1 1 "Independent" sub-RE.
739 IFTHEN off 1 1 Switch, should be preceded by switcher.
740 GROUPP num 1 Whether the group matched.
5da6b59a 741
5da6b59a
KW
742 # The heavy worker
743
89829bb5
KW
744 EVAL evl/flags Execute some Perl code.
745 2L
5da6b59a
KW
746
747 # Modifiers
748
89829bb5
KW
749 MINMOD no Next operator is not greedy.
750 LOGICAL no Next opcode should set the flag only.
5da6b59a
KW
751
752 # This is not used yet
89829bb5 753 RENUM off 1 1 Group with independently numbered parens.
5da6b59a
KW
754
755 # Trie Related
756
65aa4ca7
FC
757 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
758 # have inline charclass data (ascii only), the 'C' store it in the
759 # structure.
5da6b59a 760
89829bb5
KW
761 TRIE trie 1 Match many EXACT(F[ALU]?)? at once.
762 flags==type
763 TRIEC trie Same as TRIE, but with embedded charclass
764 charclass data
5da6b59a 765
89829bb5
KW
766 AHOCORASICK trie 1 Aho Corasick stclass. flags==type
767 AHOCORASICKC trie Same as AHOCORASICK, but with embedded
768 charclass charclass data
5da6b59a
KW
769
770 # Regex Subroutines
89829bb5 771 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
5da6b59a
KW
772
773 # Special conditionals
89829bb5
KW
774 NGROUPP no-sv 1 Whether the group matched.
775 INSUBP num 1 Whether we are in a specific recurse.
776 DEFINEP none 1 Never execute directly.
5da6b59a
KW
777
778 # Backtracking Verbs
89829bb5
KW
779 ENDLIKE none Used only for the type field of verbs
780 OPFAIL no-sv 1 Same as (?!), but with verb arg
781 ACCEPT no-sv/num Accepts the current matched string, with
782 2L verbar
5da6b59a
KW
783
784 # Verbs With Arguments
89829bb5
KW
785 VERB no-sv 1 Used only for the type field of verbs
786 PRUNE no-sv 1 Pattern fails at this startpoint if no-
787 backtracking through this
788 MARKPOINT no-sv 1 Push the current location for rollback by
789 cut.
790 SKIP no-sv 1 On failure skip forward (to the mark)
791 before retrying
792 COMMIT no-sv 1 Pattern fails outright if backtracking
793 through this
794 CUTGROUP no-sv 1 On failure go to the next alternation in
795 the group
5da6b59a
KW
796
797 # Control what to keep in $&.
89829bb5 798 KEEPS no $& begins here.
5da6b59a
KW
799
800 # New charclass like patterns
89829bb5 801 LNBREAK none generic newline pattern
5da6b59a
KW
802
803 # SPECIAL REGOPS
804
65aa4ca7
FC
805 # This is not really a node, but an optimized away piece of a "long"
806 # node. To simplify debugging output, we mark it as if it were a node
89829bb5 807 OPTIMIZED off Placeholder for dump.
5da6b59a
KW
808
809 # Special opcode with the property that no opcode in a compiled program
810 # will ever be of this type. Thus it can be used as a flag value that
811 # no other opcode has been seen. END is used similarly, in that an END
65aa4ca7
FC
812 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
813 # mean "not seen anything to optimize yet".
89829bb5 814 PSEUDO off Pseudo opcode for internal use.
65aa4ca7
FC
815
816=for regcomp.pl end
055fd3a9 817
1c102323
MJD
818=for unprinted-credits
819Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
820
821Following the optimizer information is a dump of the offset/length
822table, here split across several lines:
823
824 Offsets: [45]
825 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
826 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
827 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
828 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
829
830The first line here indicates that the offset/length table contains 45
831entries. Each entry is a pair of integers, denoted by C<offset[length]>.
17c338f3 832Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
1c102323
MJD
833entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:>
834(the C<1: ANYOF[bc]>) begins at character position 1 in the
835pre-compiled form of the regex, and has a length of 4 characters.
836C<5[1]> in position 12
837indicates that the node labeled C<12:>
838(the C<< 12: EXACT <d> >>) begins at character position 5 in the
839pre-compiled form of the regex, and has a length of 1 character.
840C<12[1]> in position 14
841indicates that the node labeled C<14:>
842(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
843pre-compiled form of the regex, and has a length of 1 character---that
844is, it corresponds to the C<+> symbol in the precompiled regex.
845
846C<0[0]> items indicate that there is no corresponding node.
847
7b406369 848=head2 Run-time Output
055fd3a9
GS
849
850First of all, when doing a match, one may get no run-time output even
851if debugging is enabled. This means that the regex engine was never
852entered and that all of the job was therefore done by the optimizer.
853
854If the regex engine was entered, the output may look like this:
855
ccf3535a 856 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
055fd3a9
GS
857 Setting an EVAL scope, savestack=3
858 2 <ab> <cdefg__gh_> | 1: ANYOF
859 3 <abc> <defg__gh_> | 11: EXACT <d>
860 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
861 4 <abcd> <efg__gh_> | 26: WHILEM
862 0 out of 1..32767 cc=effff31c
863 4 <abcd> <efg__gh_> | 15: OPEN1
864 4 <abcd> <efg__gh_> | 17: EXACT <e>
865 5 <abcde> <fg__gh_> | 19: STAR
866 EXACT <f> can match 1 times out of 32767...
867 Setting an EVAL scope, savestack=3
868 6 <bcdef> <g__gh__> | 22: EXACT <g>
869 7 <bcdefg> <__gh__> | 24: CLOSE1
870 7 <bcdefg> <__gh__> | 26: WHILEM
871 1 out of 1..32767 cc=effff31c
872 Setting an EVAL scope, savestack=12
873 7 <bcdefg> <__gh__> | 15: OPEN1
874 7 <bcdefg> <__gh__> | 17: EXACT <e>
875 restoring \1 to 4(4)..7
876 failed, try continuation...
877 7 <bcdefg> <__gh__> | 27: NOTHING
878 7 <bcdefg> <__gh__> | 28: EXACT <h>
879 failed...
880 failed...
881
882The most significant information in the output is about the particular I<node>
883of the compiled regex that is currently being tested against the target string.
884The format of these lines is
885
886C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE>
887
888The I<TYPE> info is indented with respect to the backtracking level.
889Other incidental information appears interspersed within.
890
7b406369 891=head1 Debugging Perl Memory Usage
055fd3a9
GS
892
893Perl is a profligate wastrel when it comes to memory use. There
894is a saying that to estimate memory usage of Perl, assume a reasonable
895algorithm for memory allocation, multiply that estimate by 10, and
896while you still may miss the mark, at least you won't be quite so
4375e838 897astonished. This is not absolutely true, but may provide a good
055fd3a9
GS
898grasp of what happens.
899
900Assume that an integer cannot take less than 20 bytes of memory, a
901float cannot take less than 24 bytes, a string cannot take less
902than 32 bytes (all these examples assume 32-bit architectures, the
903result are quite a bit worse on 64-bit architectures). If a variable
904is accessed in two of three different ways (which require an integer,
905a float, or a string), the memory footprint may increase yet another
b9449ee0 90620 bytes. A sloppy malloc(3) implementation can inflate these
055fd3a9
GS
907numbers dramatically.
908
909On the opposite end of the scale, a declaration like
910
911 sub foo;
912
913may take up to 500 bytes of memory, depending on which release of Perl
914you're running.
915
916Anecdotal estimates of source-to-compiled code bloat suggest an
917eightfold increase. This means that the compiled form of reasonable
918(normally commented, properly indented etc.) code will take
919about eight times more space in memory than the code took
920on disk.
921
b30f304a
JH
922The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
923(it was available only if Perl was built with C<-DDEBUGGING>).
924The switch was used to track Perl's memory allocations and possible
925memory leaks. These days the use of malloc debugging tools like
5b6a3331 926F<Purify> or F<valgrind> is suggested instead. See also
7b406369 927L<perlhacktips/PERL_MEM_LOG>.
b30f304a
JH
928
929One way to find out how much memory is being used by Perl data
930structures is to install the Devel::Size module from CPAN: it gives
931you the minimum number of bytes required to store a particular data
932structure. Please be mindful of the difference between the size()
933and total_size().
934
935If Perl has been compiled using Perl's malloc you can analyze Perl
7b406369 936memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
055fd3a9
GS
937
938=head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
939
940If your perl is using Perl's malloc() and was compiled with the
941necessary switches (this is the default), then it will print memory
4375e838 942usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
055fd3a9
GS
943> 1 >>, and before termination of the program when C<<
944$ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to
945the following example:
946
f185f654
KW
947 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
948 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
949 14216 free: 130 117 28 7 9 0 2 2 1 0 0
055fd3a9 950 437 61 36 0 5
f185f654 951 60924 used: 125 137 161 55 7 8 6 16 2 0 1
055fd3a9 952 74 109 304 84 20
f185f654
KW
953 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
954 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
955 30888 free: 245 78 85 13 6 2 1 3 2 0 1
055fd3a9 956 315 162 39 42 11
f185f654 957 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
055fd3a9 958 196 178 1066 798 39
f185f654 959 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
055fd3a9
GS
960
961It is possible to ask for such a statistic at arbitrary points in
b9449ee0 962your execution using the mstat() function out of the standard
055fd3a9
GS
963Devel::Peek module.
964
965Here is some explanation of that format:
966
13a2d996 967=over 4
055fd3a9
GS
968
969=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
970
971Perl's malloc() uses bucketed allocations. Every request is rounded
972up to the closest bucket size available, and a bucket is taken from
973the pool of buckets of that size.
974
975The line above describes the limits of buckets currently in use.
976Each bucket has two sizes: memory footprint and the maximal size
977of user data that can fit into this bucket. Suppose in the above
978example that the smallest bucket were size 4. The biggest bucket
979would have usable size 8188, and the memory footprint would be 8192.
980
981In a Perl built for debugging, some buckets may have negative usable
982size. This means that these buckets cannot (and will not) be used.
983For larger buckets, the memory footprint may be one page greater
7b406369 984than a power of 2. If so, the corresponding power of two is
055fd3a9
GS
985printed in the C<APPROX> field above.
986
987=item Free/Used
988
989The 1 or 2 rows of numbers following that correspond to the number
990of buckets of each size between C<SMALLEST> and C<GREATEST>. In
991the first row, the sizes (memory footprints) of buckets are powers
992of two--or possibly one page greater. In the second row, if present,
993the memory footprints of the buckets are between the memory footprints
994of two buckets "above".
995
4375e838 996For example, suppose under the previous example, the memory footprints
055fd3a9
GS
997were
998
f185f654 999 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
055fd3a9
GS
1000 4 12 24 48 80
1001
7b406369 1002With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
d1be9408 1003a 4-byte overhead, and thus an 8192-long bucket may take up to
055fd3a9
GS
10048188-byte allocations.
1005
1006=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
1007
1008The first two fields give the total amount of memory perl sbrk(2)ed
1009(ess-broken? :-) and number of sbrk(2)s used. The third number is
1010what perl thinks about continuity of returned chunks. So long as
1011this number is positive, malloc() will assume that it is probable
1012that sbrk(2) will provide continuous memory.
1013
1014Memory allocated by external libraries is not counted.
1015
1016=item C<pad: 0>
1017
1018The amount of sbrk(2)ed memory needed to keep buckets aligned.
1019
1020=item C<heads: 2192>
1021
1022Although memory overhead of bigger buckets is kept inside the bucket, for
1023smaller buckets, it is kept in separate areas. This field gives the
1024total size of these areas.
1025
1026=item C<chain: 0>
1027
1028malloc() may want to subdivide a bigger bucket into smaller buckets.
1029If only a part of the deceased bucket is left unsubdivided, the rest
1030is kept as an element of a linked list. This field gives the total
1031size of these chunks.
1032
1033=item C<tail: 6144>
1034
1035To minimize the number of sbrk(2)s, malloc() asks for more memory. This
1036field gives the size of the yet unused part, which is sbrk(2)ed, but
1037never touched.
1038
1039=back
1040
055fd3a9
GS
1041=head1 SEE ALSO
1042
1043L<perldebug>,
1044L<perlguts>,
1045L<perlrun>
1046L<re>,
1047and
fe854a6f 1048L<Devel::DProf>.