xref: /openbsd-src/gnu/usr.bin/perl/pod/perldebguts.pod (revision 50b7afb2c2c0993b0894d4e34bf857cb13ed9c80)
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
101=back
102
103Note that if C<&DB::sub> needs external data for it to work, no
104subroutine call is possible without it. As an example, the standard
105debugger's C<&DB::sub> depends on the C<$DB::deep> variable
106(it defines how many levels of recursion deep into the debugger you can go
107before a mandatory break).  If C<$DB::deep> is not defined, subroutine
108calls are not possible, even though C<&DB::sub> exists.
109
110=head2 Writing Your Own Debugger
111
112=head3 Environment Variables
113
114The C<PERL5DB> environment variable can be used to define a debugger.
115For example, the minimal "working" debugger (it actually doesn't do anything)
116consists of one line:
117
118  sub DB::DB {}
119
120It can easily be defined like this:
121
122  $ PERL5DB="sub DB::DB {}" perl -d your-script
123
124Another brief debugger, slightly more useful, can be created
125with only the line:
126
127  sub DB::DB {print ++$i; scalar <STDIN>}
128
129This debugger prints a number which increments for each statement
130encountered and waits for you to hit a newline before continuing
131to the next statement.
132
133The following debugger is actually useful:
134
135  {
136    package DB;
137    sub DB  {}
138    sub sub {print ++$i, " $sub\n"; &$sub}
139  }
140
141It prints the sequence number of each subroutine call and the name of the
142called subroutine.  Note that C<&DB::sub> is being compiled into the
143package C<DB> through the use of the C<package> directive.
144
145When it starts, the debugger reads your rc file (F<./.perldb> or
146F<~/.perldb> under Unix), which can set important options.
147(A subroutine (C<&afterinit>) can be defined here as well; it is executed
148after the debugger completes its own initialization.)
149
150After the rc file is read, the debugger reads the PERLDB_OPTS
151environment variable and uses it to set debugger options. The
152contents of this variable are treated as if they were the argument
153of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
154
155=head3 Debugger Internal Variables
156
157In addition to the file and subroutine-related variables mentioned above,
158the debugger also maintains various magical internal variables.
159
160=over 4
161
162=item *
163
164C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
165holds the lines of the currently-selected file (compiled by Perl), either
166explicitly chosen with the debugger's C<f> command, or implicitly by flow
167of execution.
168
169Values in this array are magical in numeric context: they compare
170equal to zero only if the line is not breakable.
171
172=item *
173
174C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
175contains breakpoints and actions keyed by line number in
176the currently-selected file, either explicitly chosen with the
177debugger's C<f> command, or implicitly by flow of execution.
178
179As previously noted, individual entries (as opposed to the whole hash)
180are settable.  Perl only cares about Boolean true here, although
181the values used by F<perl5db.pl> have the form
182C<"$break_condition\0$action">.
183
184=back
185
186=head3 Debugger Customization Functions
187
188Some functions are provided to simplify customization.
189
190=over 4
191
192=item *
193
194See L<perldebug/"Configurable Options"> for a description of options parsed by
195C<DB::parse_options(string)>.
196
197=item *
198
199C<DB::dump_trace(skip[,count])> skips the specified number of frames
200and returns a list containing information about the calling frames (all
201of them, if C<count> is missing).  Each entry is reference to a hash
202with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
203name, or info about C<eval>), C<args> (C<undef> or a reference to
204an array), C<file>, and C<line>.
205
206=item *
207
208C<DB::print_trace(FH, skip[, count[, short]])> prints
209formatted info about caller frames.  The last two functions may be
210convenient as arguments to C<< < >>, C<< << >> commands.
211
212=back
213
214Note that any variables and functions that are not documented in
215this manpages (or in L<perldebug>) are considered for internal
216use only, and as such are subject to change without notice.
217
218=head1 Frame Listing Output Examples
219
220The C<frame> option can be used to control the output of frame
221information.  For example, contrast this expression trace:
222
223 $ perl -de 42
224 Stack dump during die enabled outside of evals.
225
226 Loading DB routines from perl5db.pl patch level 0.94
227 Emacs support available.
228
229 Enter h or 'h h' for help.
230
231 main::(-e:1):   0
232   DB<1> sub foo { 14 }
233
234   DB<2> sub bar { 3 }
235
236   DB<3> t print foo() * bar()
237 main::((eval 172):3):   print foo() + bar();
238 main::foo((eval 168):2):
239 main::bar((eval 170):2):
240 42
241
242with this one, once the C<o>ption C<frame=2> has been set:
243
244   DB<4> o f=2
245                frame = '2'
246   DB<5> t print foo() * bar()
247 3:      foo() * bar()
248 entering main::foo
249  2:     sub foo { 14 };
250 exited main::foo
251 entering main::bar
252  2:     sub bar { 3 };
253 exited main::bar
254 42
255
256By way of demonstration, we present below a laborious listing
257resulting from setting your C<PERLDB_OPTS> environment variable to
258the value C<f=n N>, and running I<perl -d -V> from the command line.
259Examples using various values of C<n> are shown to give you a feel
260for the difference between settings.  Long though it may be, this
261is not a complete listing, but only excerpts.
262
263=over 4
264
265=item 1
266
267  entering main::BEGIN
268   entering Config::BEGIN
269    Package lib/Exporter.pm.
270    Package lib/Carp.pm.
271   Package lib/Config.pm.
272   entering Config::TIEHASH
273   entering Exporter::import
274    entering Exporter::export
275  entering Config::myconfig
276   entering Config::FETCH
277   entering Config::FETCH
278   entering Config::FETCH
279   entering Config::FETCH
280
281=item 2
282
283  entering main::BEGIN
284   entering Config::BEGIN
285    Package lib/Exporter.pm.
286    Package lib/Carp.pm.
287   exited Config::BEGIN
288   Package lib/Config.pm.
289   entering Config::TIEHASH
290   exited Config::TIEHASH
291   entering Exporter::import
292    entering Exporter::export
293    exited Exporter::export
294   exited Exporter::import
295  exited main::BEGIN
296  entering Config::myconfig
297   entering Config::FETCH
298   exited Config::FETCH
299   entering Config::FETCH
300   exited Config::FETCH
301   entering Config::FETCH
302
303=item 3
304
305  in  $=main::BEGIN() from /dev/null:0
306   in  $=Config::BEGIN() from lib/Config.pm:2
307    Package lib/Exporter.pm.
308    Package lib/Carp.pm.
309   Package lib/Config.pm.
310   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
311   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
312    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
313  in  @=Config::myconfig() from /dev/null:0
314   in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
315   in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
316   in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
317   in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
318   in  $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
319   in  $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
320
321=item 4
322
323  in  $=main::BEGIN() from /dev/null:0
324   in  $=Config::BEGIN() from lib/Config.pm:2
325    Package lib/Exporter.pm.
326    Package lib/Carp.pm.
327   out $=Config::BEGIN() from lib/Config.pm:0
328   Package lib/Config.pm.
329   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
330   out $=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 lib/
333    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
334   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
335  out $=main::BEGIN() from /dev/null:0
336  in  @=Config::myconfig() from /dev/null:0
337   in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
338   out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
339   in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
340   out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
341   in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
342   out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
343   in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
344
345=item 5
346
347  in  $=main::BEGIN() from /dev/null:0
348   in  $=Config::BEGIN() from lib/Config.pm:2
349    Package lib/Exporter.pm.
350    Package lib/Carp.pm.
351   out $=Config::BEGIN() from lib/Config.pm:0
352   Package lib/Config.pm.
353   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
354   out $=Config::TIEHASH('Config') from lib/Config.pm:644
355   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
356    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
357    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
358   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
359  out $=main::BEGIN() from /dev/null:0
360  in  @=Config::myconfig() from /dev/null:0
361   in  $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
362   out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
363   in  $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
364   out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
365
366=item 6
367
368  in  $=CODE(0x15eca4)() from /dev/null:0
369   in  $=CODE(0x182528)() from lib/Config.pm:2
370    Package lib/Exporter.pm.
371   out $=CODE(0x182528)() from lib/Config.pm:0
372   scalar context return from CODE(0x182528): undef
373   Package lib/Config.pm.
374   in  $=Config::TIEHASH('Config') from lib/Config.pm:628
375   out $=Config::TIEHASH('Config') from lib/Config.pm:628
376   scalar context return from Config::TIEHASH:   empty hash
377   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
378    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
379    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
380    scalar context return from Exporter::export: ''
381   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
382   scalar context return from Exporter::import: ''
383
384=back
385
386In all cases shown above, the line indentation shows the call tree.
387If bit 2 of C<frame> is set, a line is printed on exit from a
388subroutine as well.  If bit 4 is set, the arguments are printed
389along with the caller info.  If bit 8 is set, the arguments are
390printed even if they are tied or references.  If bit 16 is set, the
391return value is printed, too.
392
393When a package is compiled, a line like this
394
395    Package lib/Carp.pm.
396
397is printed with proper indentation.
398
399=head1 Debugging Regular Expressions
400
401There are two ways to enable debugging output for regular expressions.
402
403If your perl is compiled with C<-DDEBUGGING>, you may use the
404B<-Dr> flag on the command line.
405
406Otherwise, one can C<use re 'debug'>, which has effects at
407compile time and run time.  Since Perl 5.9.5, this pragma is lexically
408scoped.
409
410=head2 Compile-time Output
411
412The debugging output at compile time looks like this:
413
414  Compiling REx '[bc]d(ef*g)+h[ij]k$'
415  size 45 Got 364 bytes for offset annotations.
416  first at 1
417  rarest char g at 0
418  rarest char d at 0
419     1: ANYOF[bc](12)
420    12: EXACT <d>(14)
421    14: CURLYX[0] {1,32767}(28)
422    16:   OPEN1(18)
423    18:     EXACT <e>(20)
424    20:     STAR(23)
425    21:       EXACT <f>(0)
426    23:     EXACT <g>(25)
427    25:   CLOSE1(27)
428    27:   WHILEM[1/1](0)
429    28: NOTHING(29)
430    29: EXACT <h>(31)
431    31: ANYOF[ij](42)
432    42: EXACT <k>(44)
433    44: EOL(45)
434    45: END(0)
435  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
436        stclass 'ANYOF[bc]' minlen 7
437  Offsets: [45]
438  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
439  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
440  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
441  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
442  Omitting $` $& $' support.
443
444The first line shows the pre-compiled form of the regex.  The second
445shows the size of the compiled form (in arbitrary units, usually
4464-byte words) and the total number of bytes allocated for the
447offset/length table, usually 4+C<size>*8.  The next line shows the
448label I<id> of the first node that does a match.
449
450The
451
452  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
453        stclass 'ANYOF[bc]' minlen 7
454
455line (split into two lines above) contains optimizer
456information.  In the example shown, the optimizer found that the match
457should contain a substring C<de> at offset 1, plus substring C<gh>
458at some offset between 3 and infinity.  Moreover, when checking for
459these substrings (to abandon impossible matches quickly), Perl will check
460for the substring C<gh> before checking for the substring C<de>.  The
461optimizer may also use the knowledge that the match starts (at the
462C<first> I<id>) with a character class, and no string
463shorter than 7 characters can possibly match.
464
465The fields of interest which may appear in this line are
466
467=over 4
468
469=item C<anchored> I<STRING> C<at> I<POS>
470
471=item C<floating> I<STRING> C<at> I<POS1..POS2>
472
473See above.
474
475=item C<matching floating/anchored>
476
477Which substring to check first.
478
479=item C<minlen>
480
481The minimal length of the match.
482
483=item C<stclass> I<TYPE>
484
485Type of first matching node.
486
487=item C<noscan>
488
489Don't scan for the found substrings.
490
491=item C<isall>
492
493Means that the optimizer information is all that the regular
494expression contains, and thus one does not need to enter the regex engine at
495all.
496
497=item C<GPOS>
498
499Set if the pattern contains C<\G>.
500
501=item C<plus>
502
503Set if the pattern starts with a repeated char (as in C<x+y>).
504
505=item C<implicit>
506
507Set if the pattern starts with C<.*>.
508
509=item C<with eval>
510
511Set if the pattern contain eval-groups, such as C<(?{ code })> and
512C<(??{ code })>.
513
514=item C<anchored(TYPE)>
515
516If the pattern may match only at a handful of places, with C<TYPE>
517being C<BOL>, C<MBOL>, or C<GPOS>.  See the table below.
518
519=back
520
521If a substring is known to match at end-of-line only, it may be
522followed by C<$>, as in C<floating 'k'$>.
523
524The optimizer-specific information is used to avoid entering (a slow) regex
525engine on strings that will not definitely match.  If the C<isall> flag
526is set, a call to the regex engine may be avoided even when the optimizer
527found an appropriate place for the match.
528
529Above the optimizer section is the list of I<nodes> of the compiled
530form of the regex.  Each line has format
531
532C<   >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
533
534=head2 Types of Nodes
535
536Here are the possible types, with short descriptions:
537
538=for comment
539This table is generated by regen/regcomp.pl.  Any changes made here
540will be lost.
541
542=for regcomp.pl begin
543
544 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
545
546 # Exit points
547
548 END                no         End of program.
549 SUCCEED            no         Return from a subroutine, basically.
550
551 # Anchors:
552
553 BOL                no         Match "" at beginning of line.
554 MBOL               no         Same, assuming multiline.
555 SBOL               no         Same, assuming singleline.
556 EOS                no         Match "" at end of string.
557 EOL                no         Match "" at end of line.
558 MEOL               no         Same, assuming multiline.
559 SEOL               no         Same, assuming singleline.
560 BOUND              no         Match "" at any word boundary using
561                               native charset semantics for non-utf8
562 BOUNDL             no         Match "" at any locale word boundary
563 BOUNDU             no         Match "" at any word boundary using
564                               Unicode semantics
565 BOUNDA             no         Match "" at any word boundary using ASCII
566                               semantics
567 NBOUND             no         Match "" at any word non-boundary using
568                               native charset semantics for non-utf8
569 NBOUNDL            no         Match "" at any locale word non-boundary
570 NBOUNDU            no         Match "" at any word non-boundary using
571                               Unicode semantics
572 NBOUNDA            no         Match "" at any word non-boundary using
573                               ASCII semantics
574 GPOS               no         Matches where last m//g left off.
575
576 # [Special] alternatives:
577
578 REG_ANY            no         Match any one character (except newline).
579 SANY               no         Match any one character.
580 CANY               no         Match any one byte.
581 ANYOF              sv         Match character in (or not in) this
582                               class, single char match only
583 ANYOF_WARN_SUPER   sv         Match character in (or not in) this
584                               class, warn (if enabled) upon matching a
585                               char above Unicode max;
586 ANYOF_SYNTHETIC    sv         Synthetic start class
587
588 POSIXD             none       Some [[:class:]] under /d; the FLAGS
589                               field gives which one
590 POSIXL             none       Some [[:class:]] under /l; the FLAGS
591                               field gives which one
592 POSIXU             none       Some [[:class:]] under /u; the FLAGS
593                               field gives which one
594 POSIXA             none       Some [[:class:]] under /a; the FLAGS
595                               field gives which one
596 NPOSIXD            none       complement of POSIXD, [[:^class:]]
597 NPOSIXL            none       complement of POSIXL, [[:^class:]]
598 NPOSIXU            none       complement of POSIXU, [[:^class:]]
599 NPOSIXA            none       complement of POSIXA, [[:^class:]]
600
601 CLUMP              no         Match any extended grapheme cluster
602                               sequence
603
604 # Alternation
605
606 # BRANCH        The set of branches constituting a single choice are
607 #               hooked together with their "next" pointers, since
608 #               precedence prevents anything being concatenated to
609 #               any individual branch.  The "next" pointer of the last
610 #               BRANCH in a choice points to the thing following the
611 #               whole choice.  This is also where the final "next"
612 #               pointer of each individual branch points; each branch
613 #               starts with the operand node of a BRANCH node.
614 #
615 BRANCH             node       Match this alternative, or the next...
616
617 # Back pointer
618
619 # BACK          Normal "next" pointers all implicitly point forward;
620 #               BACK exists to make loop structures possible.
621 # not used
622 BACK               no         Match "", "next" ptr points backward.
623
624 # Literals
625
626 EXACT              str        Match this string (preceded by length).
627 EXACTF             str        Match this non-UTF-8 string (not
628                               guaranteed to be folded) using /id rules
629                               (w/len).
630 EXACTFL            str        Match this string (not guaranteed to be
631                               folded) using /il rules (w/len).
632 EXACTFU            str        Match this string (folded iff in UTF-8,
633                               length in folding doesn't change if not
634                               in UTF-8) using /iu rules (w/len).
635 EXACTFA            str        Match this string (not guaranteed to be
636                               folded) using /iaa rules (w/len).
637 EXACTFU_SS         str        Match this string (folded iff in UTF-8,
638                               length in folding may change even if not
639                               in UTF-8) using /iu rules (w/len).
640 EXACTFU_TRICKYFOLD str        Match this folded UTF-8 string using /iu
641                               rules
642
643 # Do nothing types
644
645 NOTHING            no         Match empty string.
646 # A variant of above which delimits a group, thus stops optimizations
647 TAIL               no         Match empty string. Can jump here from
648                               outside.
649
650 # Loops
651
652 # STAR,PLUS    '?', and complex '*' and '+', are implemented as
653 #               circular BRANCH structures using BACK.  Simple cases
654 #               (one character per match) are implemented with STAR
655 #               and PLUS for speed and to minimize recursive plunges.
656 #
657 STAR               node       Match this (simple) thing 0 or more
658                               times.
659 PLUS               node       Match this (simple) thing 1 or more
660                               times.
661
662 CURLY              sv 2       Match this simple thing {n,m} times.
663 CURLYN             no 2       Capture next-after-this simple thing
664 CURLYM             no 2       Capture this medium-complex thing {n,m}
665                               times.
666 CURLYX             sv 2       Match this complex thing {n,m} times.
667
668 # This terminator creates a loop structure for CURLYX
669 WHILEM             no         Do curly processing and see if rest
670                               matches.
671
672 # Buffer related
673
674 # OPEN,CLOSE,GROUPP     ...are numbered at compile time.
675 OPEN               num 1      Mark this point in input as start of #n.
676 CLOSE              num 1      Analogous to OPEN.
677
678 REF                num 1      Match some already matched string
679 REFF               num 1      Match already matched string, folded
680                               using native charset semantics for non-
681                               utf8
682 REFFL              num 1      Match already matched string, folded in
683                               loc.
684 REFFU              num 1      Match already matched string, folded
685                               using unicode semantics for non-utf8
686 REFFA              num 1      Match already matched string, folded
687                               using unicode semantics for non-utf8, no
688                               mixing ASCII, non-ASCII
689
690 # Named references.  Code in regcomp.c assumes that these all are after
691 # the numbered references
692 NREF               no-sv 1    Match some already matched string
693 NREFF              no-sv 1    Match already matched string, folded
694                               using native charset semantics for non-
695                               utf8
696 NREFFL             no-sv 1    Match already matched string, folded in
697                               loc.
698 NREFFU             num 1      Match already matched string, folded
699                               using unicode semantics for non-utf8
700 NREFFA             num 1      Match already matched string, folded
701                               using unicode semantics for non-utf8, no
702                               mixing ASCII, non-ASCII
703
704 IFMATCH            off 1 2    Succeeds if the following matches.
705 UNLESSM            off 1 2    Fails if the following matches.
706 SUSPEND            off 1 1    "Independent" sub-RE.
707 IFTHEN             off 1 1    Switch, should be preceded by switcher.
708 GROUPP             num 1      Whether the group matched.
709
710 # Support for long RE
711
712 LONGJMP            off 1 1    Jump far away.
713 BRANCHJ            off 1 1    BRANCH with long offset.
714
715 # The heavy worker
716
717 EVAL               evl 1      Execute some Perl code.
718
719 # Modifiers
720
721 MINMOD             no         Next operator is not greedy.
722 LOGICAL            no         Next opcode should set the flag only.
723
724 # This is not used yet
725 RENUM              off 1 1    Group with independently numbered parens.
726
727 # Trie Related
728
729 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
730 # have inline charclass data (ascii only), the 'C' store it in the
731 # structure.
732
733 TRIE               trie 1     Match many EXACT(F[ALU]?)? at once.
734                               flags==type
735 TRIEC              trie       Same as TRIE, but with embedded charclass
736                    charclass  data
737
738 AHOCORASICK        trie 1     Aho Corasick stclass. flags==type
739 AHOCORASICKC       trie       Same as AHOCORASICK, but with embedded
740                    charclass  charclass data
741
742 # Regex Subroutines
743 GOSUB              num/ofs 2L recurse to paren arg1 at (signed) ofs
744                               arg2
745 GOSTART            no         recurse to start of pattern
746
747 # Special conditionals
748 NGROUPP            no-sv 1    Whether the group matched.
749 INSUBP             num 1      Whether we are in a specific recurse.
750 DEFINEP            none 1     Never execute directly.
751
752 # Backtracking Verbs
753 ENDLIKE            none       Used only for the type field of verbs
754 OPFAIL             none       Same as (?!)
755 ACCEPT             parno 1    Accepts the current matched string.
756
757 # Verbs With Arguments
758 VERB               no-sv 1    Used only for the type field of verbs
759 PRUNE              no-sv 1    Pattern fails at this startpoint if no-
760                               backtracking through this
761 MARKPOINT          no-sv 1    Push the current location for rollback by
762                               cut.
763 SKIP               no-sv 1    On failure skip forward (to the mark)
764                               before retrying
765 COMMIT             no-sv 1    Pattern fails outright if backtracking
766                               through this
767 CUTGROUP           no-sv 1    On failure go to the next alternation in
768                               the group
769
770 # Control what to keep in $&.
771 KEEPS              no         $& begins here.
772
773 # New charclass like patterns
774 LNBREAK            none       generic newline pattern
775
776 # SPECIAL  REGOPS
777
778 # This is not really a node, but an optimized away piece of a "long"
779 # node.  To simplify debugging output, we mark it as if it were a node
780 OPTIMIZED          off        Placeholder for dump.
781
782 # Special opcode with the property that no opcode in a compiled program
783 # will ever be of this type. Thus it can be used as a flag value that
784 # no other opcode has been seen. END is used similarly, in that an END
785 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
786 # mean "not seen anything to optimize yet".
787 PSEUDO             off        Pseudo opcode for internal use.
788
789=for regcomp.pl end
790
791=for unprinted-credits
792Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
793
794Following the optimizer information is a dump of the offset/length
795table, here split across several lines:
796
797  Offsets: [45]
798  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
799  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
800  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
801  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
802
803The first line here indicates that the offset/length table contains 45
804entries.  Each entry is a pair of integers, denoted by C<offset[length]>.
805Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
806entry #12 is C<5[1]>.  C<1[4]> indicates that the node labeled C<1:>
807(the C<1: ANYOF[bc]>) begins at character position 1 in the
808pre-compiled form of the regex, and has a length of 4 characters.
809C<5[1]> in position 12
810indicates that the node labeled C<12:>
811(the C<< 12: EXACT <d> >>) begins at character position 5 in the
812pre-compiled form of the regex, and has a length of 1 character.
813C<12[1]> in position 14
814indicates that the node labeled C<14:>
815(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
816pre-compiled form of the regex, and has a length of 1 character---that
817is, it corresponds to the C<+> symbol in the precompiled regex.
818
819C<0[0]> items indicate that there is no corresponding node.
820
821=head2 Run-time Output
822
823First of all, when doing a match, one may get no run-time output even
824if debugging is enabled.  This means that the regex engine was never
825entered and that all of the job was therefore done by the optimizer.
826
827If the regex engine was entered, the output may look like this:
828
829  Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
830    Setting an EVAL scope, savestack=3
831     2 <ab> <cdefg__gh_>    |  1: ANYOF
832     3 <abc> <defg__gh_>    | 11: EXACT <d>
833     4 <abcd> <efg__gh_>    | 13: CURLYX {1,32767}
834     4 <abcd> <efg__gh_>    | 26:   WHILEM
835				0 out of 1..32767  cc=effff31c
836     4 <abcd> <efg__gh_>    | 15:     OPEN1
837     4 <abcd> <efg__gh_>    | 17:     EXACT <e>
838     5 <abcde> <fg__gh_>    | 19:     STAR
839			     EXACT <f> can match 1 times out of 32767...
840    Setting an EVAL scope, savestack=3
841     6 <bcdef> <g__gh__>    | 22:       EXACT <g>
842     7 <bcdefg> <__gh__>    | 24:       CLOSE1
843     7 <bcdefg> <__gh__>    | 26:       WHILEM
844				    1 out of 1..32767  cc=effff31c
845    Setting an EVAL scope, savestack=12
846     7 <bcdefg> <__gh__>    | 15:         OPEN1
847     7 <bcdefg> <__gh__>    | 17:         EXACT <e>
848       restoring \1 to 4(4)..7
849				    failed, try continuation...
850     7 <bcdefg> <__gh__>    | 27:         NOTHING
851     7 <bcdefg> <__gh__>    | 28:         EXACT <h>
852				    failed...
853				failed...
854
855The most significant information in the output is about the particular I<node>
856of the compiled regex that is currently being tested against the target string.
857The format of these lines is
858
859C<    >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>>   |I<ID>:  I<TYPE>
860
861The I<TYPE> info is indented with respect to the backtracking level.
862Other incidental information appears interspersed within.
863
864=head1 Debugging Perl Memory Usage
865
866Perl is a profligate wastrel when it comes to memory use.  There
867is a saying that to estimate memory usage of Perl, assume a reasonable
868algorithm for memory allocation, multiply that estimate by 10, and
869while you still may miss the mark, at least you won't be quite so
870astonished.  This is not absolutely true, but may provide a good
871grasp of what happens.
872
873Assume that an integer cannot take less than 20 bytes of memory, a
874float cannot take less than 24 bytes, a string cannot take less
875than 32 bytes (all these examples assume 32-bit architectures, the
876result are quite a bit worse on 64-bit architectures).  If a variable
877is accessed in two of three different ways (which require an integer,
878a float, or a string), the memory footprint may increase yet another
87920 bytes.  A sloppy malloc(3) implementation can inflate these
880numbers dramatically.
881
882On the opposite end of the scale, a declaration like
883
884  sub foo;
885
886may take up to 500 bytes of memory, depending on which release of Perl
887you're running.
888
889Anecdotal estimates of source-to-compiled code bloat suggest an
890eightfold increase.  This means that the compiled form of reasonable
891(normally commented, properly indented etc.) code will take
892about eight times more space in memory than the code took
893on disk.
894
895The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
896(it was available only if Perl was built with C<-DDEBUGGING>).
897The switch was used to track Perl's memory allocations and possible
898memory leaks.  These days the use of malloc debugging tools like
899F<Purify> or F<valgrind> is suggested instead.  See also
900L<perlhacktips/PERL_MEM_LOG>.
901
902One way to find out how much memory is being used by Perl data
903structures is to install the Devel::Size module from CPAN: it gives
904you the minimum number of bytes required to store a particular data
905structure.  Please be mindful of the difference between the size()
906and total_size().
907
908If Perl has been compiled using Perl's malloc you can analyze Perl
909memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
910
911=head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
912
913If your perl is using Perl's malloc() and was compiled with the
914necessary switches (this is the default), then it will print memory
915usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
916> 1 >>, and before termination of the program when C<<
917$ENV{PERL_DEBUG_MSTATS} >= 1 >>.  The report format is similar to
918the following example:
919
920  $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
921  Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
922     14216 free:   130   117    28     7     9   0   2     2   1 0 0
923		437    61    36     0     5
924     60924 used:   125   137   161    55     7   8   6    16   2 0 1
925		 74   109   304    84    20
926  Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
927  Memory allocation statistics after execution:   (buckets 4(4)..8188(8192)
928     30888 free:   245    78    85    13     6   2   1     3   2 0 1
929		315   162    39    42    11
930    175816 used:   265   176  1112   111    26  22  11    27   2 1 1
931		196   178  1066   798    39
932  Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
933
934It is possible to ask for such a statistic at arbitrary points in
935your execution using the mstat() function out of the standard
936Devel::Peek module.
937
938Here is some explanation of that format:
939
940=over 4
941
942=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
943
944Perl's malloc() uses bucketed allocations.  Every request is rounded
945up to the closest bucket size available, and a bucket is taken from
946the pool of buckets of that size.
947
948The line above describes the limits of buckets currently in use.
949Each bucket has two sizes: memory footprint and the maximal size
950of user data that can fit into this bucket.  Suppose in the above
951example that the smallest bucket were size 4.  The biggest bucket
952would have usable size 8188, and the memory footprint would be 8192.
953
954In a Perl built for debugging, some buckets may have negative usable
955size.  This means that these buckets cannot (and will not) be used.
956For larger buckets, the memory footprint may be one page greater
957than a power of 2.  If so, the corresponding power of two is
958printed in the C<APPROX> field above.
959
960=item Free/Used
961
962The 1 or 2 rows of numbers following that correspond to the number
963of buckets of each size between C<SMALLEST> and C<GREATEST>.  In
964the first row, the sizes (memory footprints) of buckets are powers
965of two--or possibly one page greater.  In the second row, if present,
966the memory footprints of the buckets are between the memory footprints
967of two buckets "above".
968
969For example, suppose under the previous example, the memory footprints
970were
971
972     free:    8     16    32    64    128  256 512 1024 2048 4096 8192
973	   4     12    24    48    80
974
975With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
976a 4-byte overhead, and thus an 8192-long bucket may take up to
9778188-byte allocations.
978
979=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
980
981The first two fields give the total amount of memory perl sbrk(2)ed
982(ess-broken? :-) and number of sbrk(2)s used.  The third number is
983what perl thinks about continuity of returned chunks.  So long as
984this number is positive, malloc() will assume that it is probable
985that sbrk(2) will provide continuous memory.
986
987Memory allocated by external libraries is not counted.
988
989=item C<pad: 0>
990
991The amount of sbrk(2)ed memory needed to keep buckets aligned.
992
993=item C<heads: 2192>
994
995Although memory overhead of bigger buckets is kept inside the bucket, for
996smaller buckets, it is kept in separate areas.  This field gives the
997total size of these areas.
998
999=item C<chain: 0>
1000
1001malloc() may want to subdivide a bigger bucket into smaller buckets.
1002If only a part of the deceased bucket is left unsubdivided, the rest
1003is kept as an element of a linked list.  This field gives the total
1004size of these chunks.
1005
1006=item C<tail: 6144>
1007
1008To minimize the number of sbrk(2)s, malloc() asks for more memory.  This
1009field gives the size of the yet unused part, which is sbrk(2)ed, but
1010never touched.
1011
1012=back
1013
1014=head1 SEE ALSO
1015
1016L<perldebug>,
1017L<perlguts>,
1018L<perlrun>
1019L<re>,
1020and
1021L<Devel::DProf>.
1022