xref: /openbsd-src/gnu/usr.bin/perl/pod/perldebguts.pod (revision f2da64fbbbf1b03f09f390ab01267c93dfd77c4c)
1=head1 NAME
2
3perldebguts - Guts of Perl debugging
4
5=head1 DESCRIPTION
6
7This is not L<perldebug>, which tells you how to use
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.
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
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
19F<INSTALL> podpage in the Perl source tree.
20
21For example, whenever you call Perl's built-in C<caller> function
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/$^P>):
27
28=over 4
29
30=item *
31
32Perl inserts the contents of C<$ENV{PERL5DB}> (or C<BEGIN {require
33'perl5db.pl'}> if not present) before the first line of your program.
34
35=item *
36
37Each array C<@{"_<$filename"}> holds the lines of $filename for a
38file compiled by Perl.  The same is also true for C<eval>ed strings
39that contain subroutines, or which are currently being executed.
40The $filename for C<eval>ed strings looks like C<(eval 34)>.
41
42Values in this array are magical in numeric context: they compare
43equal to zero only if the line is not breakable.
44
45=item *
46
47Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed
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
51C<"$break_condition\0$action">.
52
53The same holds for evaluated strings that contain subroutines, or
54which are currently being executed.  The $filename for C<eval>ed strings
55looks like C<(eval 34)>.
56
57=item *
58
59Each scalar C<${"_<$filename"}> contains C<"_<$filename">.  This is
60also the case for evaluated strings that contain subroutines, or
61which are currently being executed.  The $filename for C<eval>ed
62strings looks like C<(eval 34)>.
63
64=item *
65
66After each C<require>d file is compiled, but before it is executed,
67C<DB::postponed(*{"_<$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
83C<eval>s.
84
85=item *
86
87When the execution of your program reaches a point that can hold a
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
90are not C<local>izable.  This feature is disabled when executing
91inside C<DB::DB()>, including functions called from it
92unless C<< $^D & (1<<30) >> is true.
93
94=item *
95
96When execution of the program reaches a subroutine call, a call to
97C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> holding the
98name of the called subroutine. (This doesn't happen if the subroutine
99was compiled in the C<DB> package.)
100
101X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub>
102is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling
103back to C<&DB::sub>(I<args>).
104
105=item *
106
107When execution of the program uses C<goto> to enter a non-XS
108subroutine and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto>
109is made, with C<$DB::sub> holding the name of the subroutine being
110entered.
111
112=back
113
114Note that if C<&DB::sub> needs external data for it to work, no
115subroutine call is possible without it. As an example, the standard
116debugger's C<&DB::sub> depends on the C<$DB::deep> variable
117(it defines how many levels of recursion deep into the debugger you can go
118before a mandatory break).  If C<$DB::deep> is not defined, subroutine
119calls are not possible, even though C<&DB::sub> exists.
120
121=head2 Writing Your Own Debugger
122
123=head3 Environment Variables
124
125The C<PERL5DB> environment variable can be used to define a debugger.
126For example, the minimal "working" debugger (it actually doesn't do anything)
127consists of one line:
128
129  sub DB::DB {}
130
131It can easily be defined like this:
132
133  $ PERL5DB="sub DB::DB {}" perl -d your-script
134
135Another brief debugger, slightly more useful, can be created
136with only the line:
137
138  sub DB::DB {print ++$i; scalar <STDIN>}
139
140This debugger prints a number which increments for each statement
141encountered and waits for you to hit a newline before continuing
142to the next statement.
143
144The following debugger is actually useful:
145
146  {
147    package DB;
148    sub DB  {}
149    sub sub {print ++$i, " $sub\n"; &$sub}
150  }
151
152It prints the sequence number of each subroutine call and the name of the
153called subroutine.  Note that C<&DB::sub> is being compiled into the
154package C<DB> through the use of the C<package> directive.
155
156When it starts, the debugger reads your rc file (F<./.perldb> or
157F<~/.perldb> under Unix), which can set important options.
158(A subroutine (C<&afterinit>) can be defined here as well; it is executed
159after the debugger completes its own initialization.)
160
161After the rc file is read, the debugger reads the PERLDB_OPTS
162environment variable and uses it to set debugger options. The
163contents of this variable are treated as if they were the argument
164of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
165
166=head3 Debugger Internal Variables
167
168In addition to the file and subroutine-related variables mentioned above,
169the debugger also maintains various magical internal variables.
170
171=over 4
172
173=item *
174
175C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
176holds the lines of the currently-selected file (compiled by Perl), either
177explicitly chosen with the debugger's C<f> command, or implicitly by flow
178of execution.
179
180Values in this array are magical in numeric context: they compare
181equal to zero only if the line is not breakable.
182
183=item *
184
185C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
186contains breakpoints and actions keyed by line number in
187the currently-selected file, either explicitly chosen with the
188debugger's C<f> command, or implicitly by flow of execution.
189
190As previously noted, individual entries (as opposed to the whole hash)
191are settable.  Perl only cares about Boolean true here, although
192the values used by F<perl5db.pl> have the form
193C<"$break_condition\0$action">.
194
195=back
196
197=head3 Debugger Customization Functions
198
199Some functions are provided to simplify customization.
200
201=over 4
202
203=item *
204
205See L<perldebug/"Configurable Options"> for a description of options parsed by
206C<DB::parse_options(string)>.
207
208=item *
209
210C<DB::dump_trace(skip[,count])> skips the specified number of frames
211and returns a list containing information about the calling frames (all
212of them, if C<count> is missing).  Each entry is reference to a hash
213with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
214name, or info about C<eval>), C<args> (C<undef> or a reference to
215an array), C<file>, and C<line>.
216
217=item *
218
219C<DB::print_trace(FH, skip[, count[, short]])> prints
220formatted info about caller frames.  The last two functions may be
221convenient as arguments to C<< < >>, C<< << >> commands.
222
223=back
224
225Note that any variables and functions that are not documented in
226this manpages (or in L<perldebug>) are considered for internal
227use only, and as such are subject to change without notice.
228
229=head1 Frame Listing Output Examples
230
231The C<frame> option can be used to control the output of frame
232information.  For example, contrast this expression trace:
233
234 $ perl -de 42
235 Stack dump during die enabled outside of evals.
236
237 Loading DB routines from perl5db.pl patch level 0.94
238 Emacs support available.
239
240 Enter h or 'h h' for help.
241
242 main::(-e:1):   0
243   DB<1> sub foo { 14 }
244
245   DB<2> sub bar { 3 }
246
247   DB<3> t print foo() * bar()
248 main::((eval 172):3):   print foo() + bar();
249 main::foo((eval 168):2):
250 main::bar((eval 170):2):
251 42
252
253with this one, once the C<o>ption C<frame=2> has been set:
254
255   DB<4> o f=2
256                frame = '2'
257   DB<5> t print foo() * bar()
258 3:      foo() * bar()
259 entering main::foo
260  2:     sub foo { 14 };
261 exited main::foo
262 entering main::bar
263  2:     sub bar { 3 };
264 exited main::bar
265 42
266
267By way of demonstration, we present below a laborious listing
268resulting from setting your C<PERLDB_OPTS> environment variable to
269the value C<f=n N>, and running I<perl -d -V> from the command line.
270Examples using various values of C<n> are shown to give you a feel
271for the difference between settings.  Long though it may be, this
272is not a complete listing, but only excerpts.
273
274=over 4
275
276=item 1
277
278  entering main::BEGIN
279   entering Config::BEGIN
280    Package lib/Exporter.pm.
281    Package lib/Carp.pm.
282   Package lib/Config.pm.
283   entering Config::TIEHASH
284   entering Exporter::import
285    entering Exporter::export
286  entering Config::myconfig
287   entering Config::FETCH
288   entering Config::FETCH
289   entering Config::FETCH
290   entering Config::FETCH
291
292=item 2
293
294  entering main::BEGIN
295   entering Config::BEGIN
296    Package lib/Exporter.pm.
297    Package lib/Carp.pm.
298   exited Config::BEGIN
299   Package lib/Config.pm.
300   entering Config::TIEHASH
301   exited Config::TIEHASH
302   entering Exporter::import
303    entering Exporter::export
304    exited Exporter::export
305   exited Exporter::import
306  exited main::BEGIN
307  entering Config::myconfig
308   entering Config::FETCH
309   exited Config::FETCH
310   entering Config::FETCH
311   exited Config::FETCH
312   entering Config::FETCH
313
314=item 3
315
316  in  $=main::BEGIN() from /dev/null:0
317   in  $=Config::BEGIN() from lib/Config.pm:2
318    Package lib/Exporter.pm.
319    Package lib/Carp.pm.
320   Package lib/Config.pm.
321   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
322   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
323    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
324  in  @=Config::myconfig() from /dev/null:0
325   in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
326   in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
327   in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
328   in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
329   in  $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
330   in  $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
331
332=item 4
333
334  in  $=main::BEGIN() from /dev/null:0
335   in  $=Config::BEGIN() from lib/Config.pm:2
336    Package lib/Exporter.pm.
337    Package lib/Carp.pm.
338   out $=Config::BEGIN() from lib/Config.pm:0
339   Package lib/Config.pm.
340   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
341   out $=Config::TIEHASH('Config') from lib/Config.pm:644
342   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
343    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
344    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
345   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
346  out $=main::BEGIN() from /dev/null:0
347  in  @=Config::myconfig() from /dev/null:0
348   in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
349   out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
350   in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
351   out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
352   in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
353   out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
354   in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
355
356=item 5
357
358  in  $=main::BEGIN() from /dev/null:0
359   in  $=Config::BEGIN() from lib/Config.pm:2
360    Package lib/Exporter.pm.
361    Package lib/Carp.pm.
362   out $=Config::BEGIN() from lib/Config.pm:0
363   Package lib/Config.pm.
364   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
365   out $=Config::TIEHASH('Config') from lib/Config.pm:644
366   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
367    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
368    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
369   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
370  out $=main::BEGIN() from /dev/null:0
371  in  @=Config::myconfig() from /dev/null:0
372   in  $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
373   out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
374   in  $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
375   out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
376
377=item 6
378
379  in  $=CODE(0x15eca4)() from /dev/null:0
380   in  $=CODE(0x182528)() from lib/Config.pm:2
381    Package lib/Exporter.pm.
382   out $=CODE(0x182528)() from lib/Config.pm:0
383   scalar context return from CODE(0x182528): undef
384   Package lib/Config.pm.
385   in  $=Config::TIEHASH('Config') from lib/Config.pm:628
386   out $=Config::TIEHASH('Config') from lib/Config.pm:628
387   scalar context return from Config::TIEHASH:   empty hash
388   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
389    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
390    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
391    scalar context return from Exporter::export: ''
392   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
393   scalar context return from Exporter::import: ''
394
395=back
396
397In all cases shown above, the line indentation shows the call tree.
398If bit 2 of C<frame> is set, a line is printed on exit from a
399subroutine as well.  If bit 4 is set, the arguments are printed
400along with the caller info.  If bit 8 is set, the arguments are
401printed even if they are tied or references.  If bit 16 is set, the
402return value is printed, too.
403
404When a package is compiled, a line like this
405
406    Package lib/Carp.pm.
407
408is printed with proper indentation.
409
410=head1 Debugging Regular Expressions
411
412There are two ways to enable debugging output for regular expressions.
413
414If your perl is compiled with C<-DDEBUGGING>, you may use the
415B<-Dr> flag on the command line.
416
417Otherwise, one can C<use re 'debug'>, which has effects at
418compile time and run time.  Since Perl 5.9.5, this pragma is lexically
419scoped.
420
421=head2 Compile-time Output
422
423The debugging output at compile time looks like this:
424
425  Compiling REx '[bc]d(ef*g)+h[ij]k$'
426  size 45 Got 364 bytes for offset annotations.
427  first at 1
428  rarest char g at 0
429  rarest char d at 0
430     1: ANYOF[bc](12)
431    12: EXACT <d>(14)
432    14: CURLYX[0] {1,32767}(28)
433    16:   OPEN1(18)
434    18:     EXACT <e>(20)
435    20:     STAR(23)
436    21:       EXACT <f>(0)
437    23:     EXACT <g>(25)
438    25:   CLOSE1(27)
439    27:   WHILEM[1/1](0)
440    28: NOTHING(29)
441    29: EXACT <h>(31)
442    31: ANYOF[ij](42)
443    42: EXACT <k>(44)
444    44: EOL(45)
445    45: END(0)
446  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
447        stclass 'ANYOF[bc]' minlen 7
448  Offsets: [45]
449  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
450  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
451  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
452  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
453  Omitting $` $& $' support.
454
455The first line shows the pre-compiled form of the regex.  The second
456shows the size of the compiled form (in arbitrary units, usually
4574-byte words) and the total number of bytes allocated for the
458offset/length table, usually 4+C<size>*8.  The next line shows the
459label I<id> of the first node that does a match.
460
461The
462
463  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
464        stclass 'ANYOF[bc]' minlen 7
465
466line (split into two lines above) contains optimizer
467information.  In the example shown, the optimizer found that the match
468should contain a substring C<de> at offset 1, plus substring C<gh>
469at some offset between 3 and infinity.  Moreover, when checking for
470these substrings (to abandon impossible matches quickly), Perl will check
471for the substring C<gh> before checking for the substring C<de>.  The
472optimizer may also use the knowledge that the match starts (at the
473C<first> I<id>) with a character class, and no string
474shorter than 7 characters can possibly match.
475
476The fields of interest which may appear in this line are
477
478=over 4
479
480=item C<anchored> I<STRING> C<at> I<POS>
481
482=item C<floating> I<STRING> C<at> I<POS1..POS2>
483
484See above.
485
486=item C<matching floating/anchored>
487
488Which substring to check first.
489
490=item C<minlen>
491
492The minimal length of the match.
493
494=item C<stclass> I<TYPE>
495
496Type of first matching node.
497
498=item C<noscan>
499
500Don't scan for the found substrings.
501
502=item C<isall>
503
504Means that the optimizer information is all that the regular
505expression contains, and thus one does not need to enter the regex engine at
506all.
507
508=item C<GPOS>
509
510Set if the pattern contains C<\G>.
511
512=item C<plus>
513
514Set if the pattern starts with a repeated char (as in C<x+y>).
515
516=item C<implicit>
517
518Set if the pattern starts with C<.*>.
519
520=item C<with eval>
521
522Set if the pattern contain eval-groups, such as C<(?{ code })> and
523C<(??{ code })>.
524
525=item C<anchored(TYPE)>
526
527If the pattern may match only at a handful of places, with C<TYPE>
528being C<BOL>, C<MBOL>, or C<GPOS>.  See the table below.
529
530=back
531
532If a substring is known to match at end-of-line only, it may be
533followed by C<$>, as in C<floating 'k'$>.
534
535The optimizer-specific information is used to avoid entering (a slow) regex
536engine on strings that will not definitely match.  If the C<isall> flag
537is set, a call to the regex engine may be avoided even when the optimizer
538found an appropriate place for the match.
539
540Above the optimizer section is the list of I<nodes> of the compiled
541form of the regex.  Each line has format
542
543C<   >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
544
545=head2 Types of Nodes
546
547Here are the current possible types, with short descriptions:
548
549=for comment
550This table is generated by regen/regcomp.pl.  Any changes made here
551will be lost.
552
553=for regcomp.pl begin
554
555 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
556
557 # Exit points
558
559 END             no         End of program.
560 SUCCEED         no         Return from a subroutine, basically.
561
562 # Anchors:
563
564 BOL             no         Match "" at beginning of line.
565 MBOL            no         Same, assuming multiline.
566 SBOL            no         Same, assuming singleline.
567 EOS             no         Match "" at end of string.
568 EOL             no         Match "" at end of line.
569 MEOL            no         Same, assuming multiline.
570 SEOL            no         Same, assuming singleline.
571 BOUND           no         Match "" at any word boundary using native
572                            charset rules for non-utf8
573 BOUNDL          no         Match "" at any locale word boundary
574 BOUNDU          no         Match "" at any word boundary using Unicode
575                            rules
576 BOUNDA          no         Match "" at any word boundary using ASCII
577                            rules
578 NBOUND          no         Match "" at any word non-boundary using
579                            native charset rules for non-utf8
580 NBOUNDL         no         Match "" at any locale word non-boundary
581 NBOUNDU         no         Match "" at any word non-boundary using
582                            Unicode rules
583 NBOUNDA         no         Match "" at any word non-boundary using
584                            ASCII rules
585 GPOS            no         Matches where last m//g left off.
586
587 # [Special] alternatives:
588
589 REG_ANY         no         Match any one character (except newline).
590 SANY            no         Match any one character.
591 CANY            no         Match any one byte.
592 ANYOF           sv         Match character in (or not in) this class,
593                            single char match only
594
595 POSIXD          none       Some [[:class:]] under /d; the FLAGS field
596                            gives which one
597 POSIXL          none       Some [[:class:]] under /l; the FLAGS field
598                            gives which one
599 POSIXU          none       Some [[:class:]] under /u; the FLAGS field
600                            gives which one
601 POSIXA          none       Some [[:class:]] under /a; the FLAGS field
602                            gives which one
603 NPOSIXD         none       complement of POSIXD, [[:^class:]]
604 NPOSIXL         none       complement of POSIXL, [[:^class:]]
605 NPOSIXU         none       complement of POSIXU, [[:^class:]]
606 NPOSIXA         none       complement of POSIXA, [[:^class:]]
607
608 CLUMP           no         Match any extended grapheme cluster sequence
609
610 # Alternation
611
612 # BRANCH        The set of branches constituting a single choice are
613 #               hooked together with their "next" pointers, since
614 #               precedence prevents anything being concatenated to
615 #               any individual branch.  The "next" pointer of the last
616 #               BRANCH in a choice points to the thing following the
617 #               whole choice.  This is also where the final "next"
618 #               pointer of each individual branch points; each branch
619 #               starts with the operand node of a BRANCH node.
620 #
621 BRANCH          node       Match this alternative, or the next...
622
623 # Back pointer
624
625 # BACK          Normal "next" pointers all implicitly point forward;
626 #               BACK exists to make loop structures possible.
627 # not used
628 BACK            no         Match "", "next" ptr points backward.
629
630 # Literals
631
632 EXACT           str        Match this string (preceded by length).
633 EXACTF          str        Match this non-UTF-8 string (not guaranteed
634                            to be folded) using /id rules (w/len).
635 EXACTFL         str        Match this string (not guaranteed to be
636                            folded) using /il rules (w/len).
637 EXACTFU         str        Match this string (folded iff in UTF-8,
638                            length in folding doesn't change if not in
639                            UTF-8) using /iu rules (w/len).
640 EXACTFA         str        Match this string (not guaranteed to be
641                            folded) using /iaa rules (w/len).
642 EXACTFU_SS      str        Match this string (folded iff in UTF-8,
643                            length in folding may change even if not in
644                            UTF-8) using /iu rules (w/len).
645 EXACTFA_NO_TRIE str        Match this string (which is not trie-able;
646                            not guaranteed to be folded) using /iaa
647                            rules (w/len).
648
649 # Do nothing types
650
651 NOTHING         no         Match empty string.
652 # A variant of above which delimits a group, thus stops optimizations
653 TAIL            no         Match empty string. Can jump here from
654                            outside.
655
656 # Loops
657
658 # STAR,PLUS    '?', and complex '*' and '+', are implemented as
659 #               circular BRANCH structures using BACK.  Simple cases
660 #               (one character per match) are implemented with STAR
661 #               and PLUS for speed and to minimize recursive plunges.
662 #
663 STAR            node       Match this (simple) thing 0 or more times.
664 PLUS            node       Match this (simple) thing 1 or more times.
665
666 CURLY           sv 2       Match this simple thing {n,m} times.
667 CURLYN          no 2       Capture next-after-this simple thing
668 CURLYM          no 2       Capture this medium-complex thing {n,m}
669                            times.
670 CURLYX          sv 2       Match this complex thing {n,m} times.
671
672 # This terminator creates a loop structure for CURLYX
673 WHILEM          no         Do curly processing and see if rest matches.
674
675 # Buffer related
676
677 # OPEN,CLOSE,GROUPP     ...are numbered at compile time.
678 OPEN            num 1      Mark this point in input as start of #n.
679 CLOSE           num 1      Analogous to OPEN.
680
681 REF             num 1      Match some already matched string
682 REFF            num 1      Match already matched string, folded using
683                            native charset rules for non-utf8
684 REFFL           num 1      Match already matched string, folded in loc.
685 REFFU           num 1      Match already matched string, folded using
686                            unicode rules for non-utf8
687 REFFA           num 1      Match already matched string, folded using
688                            unicode rules for non-utf8, no mixing ASCII,
689                            non-ASCII
690
691 # Named references.  Code in regcomp.c assumes that these all are after
692 # the numbered references
693 NREF            no-sv 1    Match some already matched string
694 NREFF           no-sv 1    Match already matched string, folded using
695                            native charset rules for non-utf8
696 NREFFL          no-sv 1    Match already matched string, folded in loc.
697 NREFFU          num 1      Match already matched string, folded using
698                            unicode rules for non-utf8
699 NREFFA          num 1      Match already matched string, folded using
700                            unicode rules for non-utf8, no mixing ASCII,
701                            non-ASCII
702
703 IFMATCH         off 1 2    Succeeds if the following matches.
704 UNLESSM         off 1 2    Fails if the following matches.
705 SUSPEND         off 1 1    "Independent" sub-RE.
706 IFTHEN          off 1 1    Switch, should be preceded by switcher.
707 GROUPP          num 1      Whether the group matched.
708
709 # Support for long RE
710
711 LONGJMP         off 1 1    Jump far away.
712 BRANCHJ         off 1 1    BRANCH with long offset.
713
714 # The heavy worker
715
716 EVAL            evl 1      Execute some Perl code.
717
718 # Modifiers
719
720 MINMOD          no         Next operator is not greedy.
721 LOGICAL         no         Next opcode should set the flag only.
722
723 # This is not used yet
724 RENUM           off 1 1    Group with independently numbered parens.
725
726 # Trie Related
727
728 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
729 # have inline charclass data (ascii only), the 'C' store it in the
730 # structure.
731
732 TRIE            trie 1     Match many EXACT(F[ALU]?)? at once.
733                            flags==type
734 TRIEC           trie       Same as TRIE, but with embedded charclass
735                 charclass  data
736
737 AHOCORASICK     trie 1     Aho Corasick stclass. flags==type
738 AHOCORASICKC    trie       Same as AHOCORASICK, but with embedded
739                 charclass  charclass data
740
741 # Regex Subroutines
742 GOSUB           num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
743 GOSTART         no         recurse to start of pattern
744
745 # Special conditionals
746 NGROUPP         no-sv 1    Whether the group matched.
747 INSUBP          num 1      Whether we are in a specific recurse.
748 DEFINEP         none 1     Never execute directly.
749
750 # Backtracking Verbs
751 ENDLIKE         none       Used only for the type field of verbs
752 OPFAIL          none       Same as (?!)
753 ACCEPT          parno 1    Accepts the current matched string.
754
755 # Verbs With Arguments
756 VERB            no-sv 1    Used only for the type field of verbs
757 PRUNE           no-sv 1    Pattern fails at this startpoint if no-
758                            backtracking through this
759 MARKPOINT       no-sv 1    Push the current location for rollback by
760                            cut.
761 SKIP            no-sv 1    On failure skip forward (to the mark) before
762                            retrying
763 COMMIT          no-sv 1    Pattern fails outright if backtracking
764                            through this
765 CUTGROUP        no-sv 1    On failure go to the next alternation in the
766                            group
767
768 # Control what to keep in $&.
769 KEEPS           no         $& begins here.
770
771 # New charclass like patterns
772 LNBREAK         none       generic newline pattern
773
774 # SPECIAL  REGOPS
775
776 # This is not really a node, but an optimized away piece of a "long"
777 # node.  To simplify debugging output, we mark it as if it were a node
778 OPTIMIZED       off        Placeholder for dump.
779
780 # Special opcode with the property that no opcode in a compiled program
781 # will ever be of this type. Thus it can be used as a flag value that
782 # no other opcode has been seen. END is used similarly, in that an END
783 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
784 # mean "not seen anything to optimize yet".
785 PSEUDO          off        Pseudo opcode for internal use.
786
787=for regcomp.pl end
788
789=for unprinted-credits
790Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
791
792Following the optimizer information is a dump of the offset/length
793table, here split across several lines:
794
795  Offsets: [45]
796  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
797  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
798  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
799  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
800
801The first line here indicates that the offset/length table contains 45
802entries.  Each entry is a pair of integers, denoted by C<offset[length]>.
803Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
804entry #12 is C<5[1]>.  C<1[4]> indicates that the node labeled C<1:>
805(the C<1: ANYOF[bc]>) begins at character position 1 in the
806pre-compiled form of the regex, and has a length of 4 characters.
807C<5[1]> in position 12
808indicates that the node labeled C<12:>
809(the C<< 12: EXACT <d> >>) begins at character position 5 in the
810pre-compiled form of the regex, and has a length of 1 character.
811C<12[1]> in position 14
812indicates that the node labeled C<14:>
813(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
814pre-compiled form of the regex, and has a length of 1 character---that
815is, it corresponds to the C<+> symbol in the precompiled regex.
816
817C<0[0]> items indicate that there is no corresponding node.
818
819=head2 Run-time Output
820
821First of all, when doing a match, one may get no run-time output even
822if debugging is enabled.  This means that the regex engine was never
823entered and that all of the job was therefore done by the optimizer.
824
825If the regex engine was entered, the output may look like this:
826
827  Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
828    Setting an EVAL scope, savestack=3
829     2 <ab> <cdefg__gh_>    |  1: ANYOF
830     3 <abc> <defg__gh_>    | 11: EXACT <d>
831     4 <abcd> <efg__gh_>    | 13: CURLYX {1,32767}
832     4 <abcd> <efg__gh_>    | 26:   WHILEM
833				0 out of 1..32767  cc=effff31c
834     4 <abcd> <efg__gh_>    | 15:     OPEN1
835     4 <abcd> <efg__gh_>    | 17:     EXACT <e>
836     5 <abcde> <fg__gh_>    | 19:     STAR
837			     EXACT <f> can match 1 times out of 32767...
838    Setting an EVAL scope, savestack=3
839     6 <bcdef> <g__gh__>    | 22:       EXACT <g>
840     7 <bcdefg> <__gh__>    | 24:       CLOSE1
841     7 <bcdefg> <__gh__>    | 26:       WHILEM
842				    1 out of 1..32767  cc=effff31c
843    Setting an EVAL scope, savestack=12
844     7 <bcdefg> <__gh__>    | 15:         OPEN1
845     7 <bcdefg> <__gh__>    | 17:         EXACT <e>
846       restoring \1 to 4(4)..7
847				    failed, try continuation...
848     7 <bcdefg> <__gh__>    | 27:         NOTHING
849     7 <bcdefg> <__gh__>    | 28:         EXACT <h>
850				    failed...
851				failed...
852
853The most significant information in the output is about the particular I<node>
854of the compiled regex that is currently being tested against the target string.
855The format of these lines is
856
857C<    >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>>   |I<ID>:  I<TYPE>
858
859The I<TYPE> info is indented with respect to the backtracking level.
860Other incidental information appears interspersed within.
861
862=head1 Debugging Perl Memory Usage
863
864Perl is a profligate wastrel when it comes to memory use.  There
865is a saying that to estimate memory usage of Perl, assume a reasonable
866algorithm for memory allocation, multiply that estimate by 10, and
867while you still may miss the mark, at least you won't be quite so
868astonished.  This is not absolutely true, but may provide a good
869grasp of what happens.
870
871Assume that an integer cannot take less than 20 bytes of memory, a
872float cannot take less than 24 bytes, a string cannot take less
873than 32 bytes (all these examples assume 32-bit architectures, the
874result are quite a bit worse on 64-bit architectures).  If a variable
875is accessed in two of three different ways (which require an integer,
876a float, or a string), the memory footprint may increase yet another
87720 bytes.  A sloppy malloc(3) implementation can inflate these
878numbers dramatically.
879
880On the opposite end of the scale, a declaration like
881
882  sub foo;
883
884may take up to 500 bytes of memory, depending on which release of Perl
885you're running.
886
887Anecdotal estimates of source-to-compiled code bloat suggest an
888eightfold increase.  This means that the compiled form of reasonable
889(normally commented, properly indented etc.) code will take
890about eight times more space in memory than the code took
891on disk.
892
893The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
894(it was available only if Perl was built with C<-DDEBUGGING>).
895The switch was used to track Perl's memory allocations and possible
896memory leaks.  These days the use of malloc debugging tools like
897F<Purify> or F<valgrind> is suggested instead.  See also
898L<perlhacktips/PERL_MEM_LOG>.
899
900One way to find out how much memory is being used by Perl data
901structures is to install the Devel::Size module from CPAN: it gives
902you the minimum number of bytes required to store a particular data
903structure.  Please be mindful of the difference between the size()
904and total_size().
905
906If Perl has been compiled using Perl's malloc you can analyze Perl
907memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
908
909=head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
910
911If your perl is using Perl's malloc() and was compiled with the
912necessary switches (this is the default), then it will print memory
913usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
914> 1 >>, and before termination of the program when C<<
915$ENV{PERL_DEBUG_MSTATS} >= 1 >>.  The report format is similar to
916the following example:
917
918  $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
919  Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
920     14216 free:   130   117    28     7     9   0   2     2   1 0 0
921		437    61    36     0     5
922     60924 used:   125   137   161    55     7   8   6    16   2 0 1
923		 74   109   304    84    20
924  Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
925  Memory allocation statistics after execution:   (buckets 4(4)..8188(8192)
926     30888 free:   245    78    85    13     6   2   1     3   2 0 1
927		315   162    39    42    11
928    175816 used:   265   176  1112   111    26  22  11    27   2 1 1
929		196   178  1066   798    39
930  Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
931
932It is possible to ask for such a statistic at arbitrary points in
933your execution using the mstat() function out of the standard
934Devel::Peek module.
935
936Here is some explanation of that format:
937
938=over 4
939
940=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
941
942Perl's malloc() uses bucketed allocations.  Every request is rounded
943up to the closest bucket size available, and a bucket is taken from
944the pool of buckets of that size.
945
946The line above describes the limits of buckets currently in use.
947Each bucket has two sizes: memory footprint and the maximal size
948of user data that can fit into this bucket.  Suppose in the above
949example that the smallest bucket were size 4.  The biggest bucket
950would have usable size 8188, and the memory footprint would be 8192.
951
952In a Perl built for debugging, some buckets may have negative usable
953size.  This means that these buckets cannot (and will not) be used.
954For larger buckets, the memory footprint may be one page greater
955than a power of 2.  If so, the corresponding power of two is
956printed in the C<APPROX> field above.
957
958=item Free/Used
959
960The 1 or 2 rows of numbers following that correspond to the number
961of buckets of each size between C<SMALLEST> and C<GREATEST>.  In
962the first row, the sizes (memory footprints) of buckets are powers
963of two--or possibly one page greater.  In the second row, if present,
964the memory footprints of the buckets are between the memory footprints
965of two buckets "above".
966
967For example, suppose under the previous example, the memory footprints
968were
969
970     free:    8     16    32    64    128  256 512 1024 2048 4096 8192
971	   4     12    24    48    80
972
973With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
974a 4-byte overhead, and thus an 8192-long bucket may take up to
9758188-byte allocations.
976
977=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
978
979The first two fields give the total amount of memory perl sbrk(2)ed
980(ess-broken? :-) and number of sbrk(2)s used.  The third number is
981what perl thinks about continuity of returned chunks.  So long as
982this number is positive, malloc() will assume that it is probable
983that sbrk(2) will provide continuous memory.
984
985Memory allocated by external libraries is not counted.
986
987=item C<pad: 0>
988
989The amount of sbrk(2)ed memory needed to keep buckets aligned.
990
991=item C<heads: 2192>
992
993Although memory overhead of bigger buckets is kept inside the bucket, for
994smaller buckets, it is kept in separate areas.  This field gives the
995total size of these areas.
996
997=item C<chain: 0>
998
999malloc() may want to subdivide a bigger bucket into smaller buckets.
1000If only a part of the deceased bucket is left unsubdivided, the rest
1001is kept as an element of a linked list.  This field gives the total
1002size of these chunks.
1003
1004=item C<tail: 6144>
1005
1006To minimize the number of sbrk(2)s, malloc() asks for more memory.  This
1007field gives the size of the yet unused part, which is sbrk(2)ed, but
1008never touched.
1009
1010=back
1011
1012=head1 SEE ALSO
1013
1014L<perldebug>,
1015L<perlguts>,
1016L<perlrun>
1017L<re>,
1018and
1019L<Devel::DProf>.
1020