xref: /openbsd/gnu/usr.bin/perl/pod/perldebguts.pod (revision 274d7c50)
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> 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<$DB::sub> as a reference
103by which to call the called subroutine, which it will normally want to do.
104
105X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub>
106is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling
107back to C<&DB::sub>(I<args>).
108
109=item *
110
111When execution of the program uses C<goto> to enter a non-XS subroutine
112and the 0x80 bit is set in C<$^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.
120
121=back
122
123Note that if C<&DB::sub> needs external data for it to work, no
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.
129
130=head2 Writing Your Own Debugger
131
132=head3 Environment Variables
133
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:
137
138  sub DB::DB {}
139
140It can easily be defined like this:
141
142  $ PERL5DB="sub DB::DB {}" perl -d your-script
143
144Another brief debugger, slightly more useful, can be created
145with only the line:
146
147  sub DB::DB {print ++$i; scalar <STDIN>}
148
149This debugger prints a number which increments for each statement
150encountered and waits for you to hit a newline before continuing
151to the next statement.
152
153The following debugger is actually useful:
154
155  {
156    package DB;
157    sub DB  {}
158    sub sub {print ++$i, " $sub\n"; &$sub}
159  }
160
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.
164
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.)
169
170After the rc file is read, the debugger reads the PERLDB_OPTS
171environment variable and uses it to set debugger options. The
172contents of this variable are treated as if they were the argument
173of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
174
175=head3 Debugger Internal Variables
176
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 *
183
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
194C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
195contains breakpoints and actions keyed by line number in
196the currently-selected file, either explicitly chosen with the
197debugger's C<f> command, or implicitly by flow of execution.
198
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
206=head3 Debugger Customization Functions
207
208Some functions are provided to simplify customization.
209
210=over 4
211
212=item *
213
214See L<perldebug/"Configurable Options"> for a description of options parsed by
215C<DB::parse_options(string)>.
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
223name, or info about C<eval>), C<args> (C<undef> or a reference to
224an array), C<file>, and C<line>.
225
226=item *
227
228C<DB::print_trace(FH, skip[, count[, short]])> prints
229formatted info about caller frames.  The last two functions may be
230convenient as arguments to C<< < >>, C<< << >> commands.
231
232=back
233
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
249 Enter h or 'h h' for help.
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
262with this one, once the C<o>ption C<frame=2> has been set:
263
264   DB<4> o f=2
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.
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
281is not a complete listing, but only excerpts.
282
283=over 4
284
285=item 1
286
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
300
301=item 2
302
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
322
323=item 3
324
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
340
341=item 4
342
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
364
365=item 5
366
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
385
386=item 6
387
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 $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
402  scalar context return from Exporter::import: ''
403
404=back
405
406In all cases shown above, the line indentation shows the call tree.
407If bit 2 of C<frame> is set, a line is printed on exit from a
408subroutine as well.  If bit 4 is set, the arguments are printed
409along with the caller info.  If bit 8 is set, the arguments are
410printed even if they are tied or references.  If bit 16 is set, the
411return value is printed, too.
412
413When a package is compiled, a line like this
414
415    Package lib/Carp.pm.
416
417is printed with proper indentation.
418
419=head1 Debugging Regular Expressions
420
421There are two ways to enable debugging output for regular expressions.
422
423If your perl is compiled with C<-DDEBUGGING>, you may use the
424B<-Dr> flag on the command line.
425
426Otherwise, one can C<use re 'debug'>, which has effects at
427compile time and run time.  Since Perl 5.9.5, this pragma is lexically
428scoped.
429
430=head2 Compile-time Output
431
432The debugging output at compile time looks like this:
433
434  Compiling REx '[bc]d(ef*g)+h[ij]k$'
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)
455  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
456        stclass 'ANYOF[bc]' minlen 7
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.
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
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.
469
470The
471
472  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
473        stclass 'ANYOF[bc]' minlen 7
474
475line (split into two lines above) contains optimizer
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
482C<first> I<id>) with a character class, and no string
483shorter than 7 characters can possibly match.
484
485The fields of interest which may appear in this line are
486
487=over 4
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
513Means that the optimizer information is all that the regular
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
536If the pattern may match only at a handful of places, with C<TYPE>
537being C<SBOL>, C<MBOL>, or C<GPOS>.  See the table below.
538
539=back
540
541If a substring is known to match at end-of-line only, it may be
542followed by C<$>, as in C<floating 'k'$>.
543
544The optimizer-specific information is used to avoid entering (a slow) regex
545engine on strings that will not definitely match.  If the C<isall> flag
546is set, a call to the regex engine may be avoided even when the optimizer
547found an appropriate place for the match.
548
549Above the optimizer section is the list of I<nodes> of the compiled
550form of the regex.  Each line has format
551
552C<   >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
553
554=head2 Types of Nodes
555
556Here are the current possible types, with short descriptions:
557
558=for comment
559This table is generated by regen/regcomp.pl.  Any changes made here
560will be lost.
561
562=for regcomp.pl begin
563
564 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
565
566 # Exit points
567
568 END              no         End of program.
569 SUCCEED          no         Return from a subroutine, basically.
570
571 # Line Start Anchors:
572 SBOL             no         Match "" at beginning of line: /^/, /\A/
573 MBOL             no         Same, assuming multiline: /^/m
574
575 # Line End Anchors:
576 SEOL             no         Match "" at end of line: /$/
577 MEOL             no         Same, assuming multiline: /$/m
578 EOS              no         Match "" at end of string: /\z/
579
580 # Match Start Anchors:
581 GPOS             no         Matches where last m//g left off.
582
583 # Word Boundary Opcodes:
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]
600
601 # [Special] alternatives:
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
608 ANYOFM           byte 1     Like ANYOF, but matches an invariant byte
609                             as determined by the mask and arg
610
611 # POSIX Character Classes:
612 POSIXD           none       Some [[:class:]] under /d; the FLAGS field
613                             gives which one
614 POSIXL           none       Some [[:class:]] under /l; the FLAGS field
615                             gives which one
616 POSIXU           none       Some [[:class:]] under /u; the FLAGS field
617                             gives which one
618 POSIXA           none       Some [[:class:]] under /a; the FLAGS field
619                             gives which one
620 NPOSIXD          none       complement of POSIXD, [[:^class:]]
621 NPOSIXL          none       complement of POSIXL, [[:^class:]]
622 NPOSIXU          none       complement of POSIXU, [[:^class:]]
623 NPOSIXA          none       complement of POSIXA, [[:^class:]]
624
625 ASCII            none       [[:ascii:]]
626 NASCII           none       [[:^ascii:]]
627
628 CLUMP            no         Match any extended grapheme cluster
629                             sequence
630
631 # Alternation
632
633 # BRANCH        The set of branches constituting a single choice are
634 #               hooked together with their "next" pointers, since
635 #               precedence prevents anything being concatenated to
636 #               any individual branch.  The "next" pointer of the last
637 #               BRANCH in a choice points to the thing following the
638 #               whole choice.  This is also where the final "next"
639 #               pointer of each individual branch points; each branch
640 #               starts with the operand node of a BRANCH node.
641 #
642 BRANCH           node       Match this alternative, or the next...
643
644 # Literals
645
646 EXACT            str        Match this string (preceded by length).
647 EXACTL           str        Like EXACT, but /l is in effect (used so
648                             locale-related warnings can be checked
649                             for).
650 EXACTF           str        Match this non-UTF-8 string (not guaranteed
651                             to be folded) using /id rules (w/len).
652 EXACTFL          str        Match this string (not guaranteed to be
653                             folded) using /il rules (w/len).
654 EXACTFU          str        Match this string (folded iff in UTF-8,
655                             length in folding doesn't change if not in
656                             UTF-8) using /iu rules (w/len).
657 EXACTFAA         str        Match this string (not guaranteed to be
658                             folded) using /iaa rules (w/len).
659
660 EXACTFU_SS       str        Match this string (folded iff in UTF-8,
661                             length in folding may change even if not in
662                             UTF-8) using /iu rules (w/len).
663 EXACTFLU8        str        Rare circumstances: like EXACTFU, but is
664                             under /l, UTF-8, folded, and everything in
665                             it is above 255.
666 EXACTFAA_NO_TRIE str        Match this string (which is not trie-able;
667                             not guaranteed to be folded) using /iaa
668                             rules (w/len).
669
670 # Do nothing types
671
672 NOTHING          no         Match empty string.
673 # A variant of above which delimits a group, thus stops optimizations
674 TAIL             no         Match empty string. Can jump here from
675                             outside.
676
677 # Loops
678
679 # STAR,PLUS    '?', and complex '*' and '+', are implemented as
680 #               circular BRANCH structures.  Simple cases
681 #               (one character per match) are implemented with STAR
682 #               and PLUS for speed and to minimize recursive plunges.
683 #
684 STAR             node       Match this (simple) thing 0 or more times.
685 PLUS             node       Match this (simple) thing 1 or more times.
686
687 CURLY            sv 2       Match this simple thing {n,m} times.
688 CURLYN           no 2       Capture next-after-this simple thing
689 CURLYM           no 2       Capture this medium-complex thing {n,m}
690                             times.
691 CURLYX           sv 2       Match this complex thing {n,m} times.
692
693 # This terminator creates a loop structure for CURLYX
694 WHILEM           no         Do curly processing and see if rest
695                             matches.
696
697 # Buffer related
698
699 # OPEN,CLOSE,GROUPP     ...are numbered at compile time.
700 OPEN             num 1      Mark this point in input as start of #n.
701 CLOSE            num 1      Close corresponding OPEN of #n.
702 SROPEN           none       Same as OPEN, but for script run
703 SRCLOSE          none       Close preceding SROPEN
704
705 REF              num 1      Match some already matched string
706 REFF             num 1      Match already matched string, folded using
707                             native charset rules for non-utf8
708 REFFL            num 1      Match already matched string, folded in
709                             loc.
710 REFFU            num 1      Match already matched string, folded using
711                             unicode rules for non-utf8
712 REFFA            num 1      Match already matched string, folded using
713                             unicode rules for non-utf8, no mixing
714                             ASCII, non-ASCII
715
716 # Named references.  Code in regcomp.c assumes that these all are after
717 # the numbered references
718 NREF             no-sv 1    Match some already matched string
719 NREFF            no-sv 1    Match already matched string, folded using
720                             native charset rules for non-utf8
721 NREFFL           no-sv 1    Match already matched string, folded in
722                             loc.
723 NREFFU           num 1      Match already matched string, folded using
724                             unicode rules for non-utf8
725 NREFFA           num 1      Match already matched string, folded using
726                             unicode rules for non-utf8, no mixing
727                             ASCII, non-ASCII
728
729 # Support for long RE
730 LONGJMP          off 1 1    Jump far away.
731 BRANCHJ          off 1 1    BRANCH with long offset.
732
733 # Special Case Regops
734 IFMATCH          off 1 2    Succeeds if the following matches.
735 UNLESSM          off 1 2    Fails if the following matches.
736 SUSPEND          off 1 1    "Independent" sub-RE.
737 IFTHEN           off 1 1    Switch, should be preceded by switcher.
738 GROUPP           num 1      Whether the group matched.
739
740 # The heavy worker
741
742 EVAL             evl/flags  Execute some Perl code.
743                  2L
744
745 # Modifiers
746
747 MINMOD           no         Next operator is not greedy.
748 LOGICAL          no         Next opcode should set the flag only.
749
750 # This is not used yet
751 RENUM            off 1 1    Group with independently numbered parens.
752
753 # Trie Related
754
755 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
756 # have inline charclass data (ascii only), the 'C' store it in the
757 # structure.
758
759 TRIE             trie 1     Match many EXACT(F[ALU]?)? at once.
760                             flags==type
761 TRIEC            trie       Same as TRIE, but with embedded charclass
762                  charclass  data
763
764 AHOCORASICK      trie 1     Aho Corasick stclass. flags==type
765 AHOCORASICKC     trie       Same as AHOCORASICK, but with embedded
766                  charclass  charclass data
767
768 # Regex Subroutines
769 GOSUB            num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
770
771 # Special conditionals
772 NGROUPP          no-sv 1    Whether the group matched.
773 INSUBP           num 1      Whether we are in a specific recurse.
774 DEFINEP          none 1     Never execute directly.
775
776 # Backtracking Verbs
777 ENDLIKE          none       Used only for the type field of verbs
778 OPFAIL           no-sv 1    Same as (?!), but with verb arg
779 ACCEPT           no-sv/num  Accepts the current matched string, with
780                  2L         verbar
781
782 # Verbs With Arguments
783 VERB             no-sv 1    Used only for the type field of verbs
784 PRUNE            no-sv 1    Pattern fails at this startpoint if no-
785                             backtracking through this
786 MARKPOINT        no-sv 1    Push the current location for rollback by
787                             cut.
788 SKIP             no-sv 1    On failure skip forward (to the mark)
789                             before retrying
790 COMMIT           no-sv 1    Pattern fails outright if backtracking
791                             through this
792 CUTGROUP         no-sv 1    On failure go to the next alternation in
793                             the group
794
795 # Control what to keep in $&.
796 KEEPS            no         $& begins here.
797
798 # New charclass like patterns
799 LNBREAK          none       generic newline pattern
800
801 # SPECIAL  REGOPS
802
803 # This is not really a node, but an optimized away piece of a "long"
804 # node.  To simplify debugging output, we mark it as if it were a node
805 OPTIMIZED        off        Placeholder for dump.
806
807 # Special opcode with the property that no opcode in a compiled program
808 # will ever be of this type. Thus it can be used as a flag value that
809 # no other opcode has been seen. END is used similarly, in that an END
810 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
811 # mean "not seen anything to optimize yet".
812 PSEUDO           off        Pseudo opcode for internal use.
813
814=for regcomp.pl end
815
816=for unprinted-credits
817Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
818
819Following the optimizer information is a dump of the offset/length
820table, here split across several lines:
821
822  Offsets: [45]
823  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
824  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
825  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
826  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
827
828The first line here indicates that the offset/length table contains 45
829entries.  Each entry is a pair of integers, denoted by C<offset[length]>.
830Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
831entry #12 is C<5[1]>.  C<1[4]> indicates that the node labeled C<1:>
832(the C<1: ANYOF[bc]>) begins at character position 1 in the
833pre-compiled form of the regex, and has a length of 4 characters.
834C<5[1]> in position 12
835indicates that the node labeled C<12:>
836(the C<< 12: EXACT <d> >>) begins at character position 5 in the
837pre-compiled form of the regex, and has a length of 1 character.
838C<12[1]> in position 14
839indicates that the node labeled C<14:>
840(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
841pre-compiled form of the regex, and has a length of 1 character---that
842is, it corresponds to the C<+> symbol in the precompiled regex.
843
844C<0[0]> items indicate that there is no corresponding node.
845
846=head2 Run-time Output
847
848First of all, when doing a match, one may get no run-time output even
849if debugging is enabled.  This means that the regex engine was never
850entered and that all of the job was therefore done by the optimizer.
851
852If the regex engine was entered, the output may look like this:
853
854  Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
855    Setting an EVAL scope, savestack=3
856     2 <ab> <cdefg__gh_>    |  1: ANYOF
857     3 <abc> <defg__gh_>    | 11: EXACT <d>
858     4 <abcd> <efg__gh_>    | 13: CURLYX {1,32767}
859     4 <abcd> <efg__gh_>    | 26:   WHILEM
860				0 out of 1..32767  cc=effff31c
861     4 <abcd> <efg__gh_>    | 15:     OPEN1
862     4 <abcd> <efg__gh_>    | 17:     EXACT <e>
863     5 <abcde> <fg__gh_>    | 19:     STAR
864			     EXACT <f> can match 1 times out of 32767...
865    Setting an EVAL scope, savestack=3
866     6 <bcdef> <g__gh__>    | 22:       EXACT <g>
867     7 <bcdefg> <__gh__>    | 24:       CLOSE1
868     7 <bcdefg> <__gh__>    | 26:       WHILEM
869				    1 out of 1..32767  cc=effff31c
870    Setting an EVAL scope, savestack=12
871     7 <bcdefg> <__gh__>    | 15:         OPEN1
872     7 <bcdefg> <__gh__>    | 17:         EXACT <e>
873       restoring \1 to 4(4)..7
874				    failed, try continuation...
875     7 <bcdefg> <__gh__>    | 27:         NOTHING
876     7 <bcdefg> <__gh__>    | 28:         EXACT <h>
877				    failed...
878				failed...
879
880The most significant information in the output is about the particular I<node>
881of the compiled regex that is currently being tested against the target string.
882The format of these lines is
883
884C<    >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>>   |I<ID>:  I<TYPE>
885
886The I<TYPE> info is indented with respect to the backtracking level.
887Other incidental information appears interspersed within.
888
889=head1 Debugging Perl Memory Usage
890
891Perl is a profligate wastrel when it comes to memory use.  There
892is a saying that to estimate memory usage of Perl, assume a reasonable
893algorithm for memory allocation, multiply that estimate by 10, and
894while you still may miss the mark, at least you won't be quite so
895astonished.  This is not absolutely true, but may provide a good
896grasp of what happens.
897
898Assume that an integer cannot take less than 20 bytes of memory, a
899float cannot take less than 24 bytes, a string cannot take less
900than 32 bytes (all these examples assume 32-bit architectures, the
901result are quite a bit worse on 64-bit architectures).  If a variable
902is accessed in two of three different ways (which require an integer,
903a float, or a string), the memory footprint may increase yet another
90420 bytes.  A sloppy malloc(3) implementation can inflate these
905numbers dramatically.
906
907On the opposite end of the scale, a declaration like
908
909  sub foo;
910
911may take up to 500 bytes of memory, depending on which release of Perl
912you're running.
913
914Anecdotal estimates of source-to-compiled code bloat suggest an
915eightfold increase.  This means that the compiled form of reasonable
916(normally commented, properly indented etc.) code will take
917about eight times more space in memory than the code took
918on disk.
919
920The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
921(it was available only if Perl was built with C<-DDEBUGGING>).
922The switch was used to track Perl's memory allocations and possible
923memory leaks.  These days the use of malloc debugging tools like
924F<Purify> or F<valgrind> is suggested instead.  See also
925L<perlhacktips/PERL_MEM_LOG>.
926
927One way to find out how much memory is being used by Perl data
928structures is to install the Devel::Size module from CPAN: it gives
929you the minimum number of bytes required to store a particular data
930structure.  Please be mindful of the difference between the size()
931and total_size().
932
933If Perl has been compiled using Perl's malloc you can analyze Perl
934memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
935
936=head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
937
938If your perl is using Perl's malloc() and was compiled with the
939necessary switches (this is the default), then it will print memory
940usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
941> 1 >>, and before termination of the program when C<<
942$ENV{PERL_DEBUG_MSTATS} >= 1 >>.  The report format is similar to
943the following example:
944
945 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
946 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
947    14216 free:   130   117    28     7     9   0   2     2   1 0 0
948		437    61    36     0     5
949    60924 used:   125   137   161    55     7   8   6    16   2 0 1
950		 74   109   304    84    20
951 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
952 Memory allocation statistics after execution:   (buckets 4(4)..8188(8192)
953    30888 free:   245    78    85    13     6   2   1     3   2 0 1
954		315   162    39    42    11
955   175816 used:   265   176  1112   111    26  22  11    27   2 1 1
956		196   178  1066   798    39
957 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
958
959It is possible to ask for such a statistic at arbitrary points in
960your execution using the mstat() function out of the standard
961Devel::Peek module.
962
963Here is some explanation of that format:
964
965=over 4
966
967=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
968
969Perl's malloc() uses bucketed allocations.  Every request is rounded
970up to the closest bucket size available, and a bucket is taken from
971the pool of buckets of that size.
972
973The line above describes the limits of buckets currently in use.
974Each bucket has two sizes: memory footprint and the maximal size
975of user data that can fit into this bucket.  Suppose in the above
976example that the smallest bucket were size 4.  The biggest bucket
977would have usable size 8188, and the memory footprint would be 8192.
978
979In a Perl built for debugging, some buckets may have negative usable
980size.  This means that these buckets cannot (and will not) be used.
981For larger buckets, the memory footprint may be one page greater
982than a power of 2.  If so, the corresponding power of two is
983printed in the C<APPROX> field above.
984
985=item Free/Used
986
987The 1 or 2 rows of numbers following that correspond to the number
988of buckets of each size between C<SMALLEST> and C<GREATEST>.  In
989the first row, the sizes (memory footprints) of buckets are powers
990of two--or possibly one page greater.  In the second row, if present,
991the memory footprints of the buckets are between the memory footprints
992of two buckets "above".
993
994For example, suppose under the previous example, the memory footprints
995were
996
997   free:    8     16    32    64    128  256 512 1024 2048 4096 8192
998	   4     12    24    48    80
999
1000With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
1001a 4-byte overhead, and thus an 8192-long bucket may take up to
10028188-byte allocations.
1003
1004=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
1005
1006The first two fields give the total amount of memory perl sbrk(2)ed
1007(ess-broken? :-) and number of sbrk(2)s used.  The third number is
1008what perl thinks about continuity of returned chunks.  So long as
1009this number is positive, malloc() will assume that it is probable
1010that sbrk(2) will provide continuous memory.
1011
1012Memory allocated by external libraries is not counted.
1013
1014=item C<pad: 0>
1015
1016The amount of sbrk(2)ed memory needed to keep buckets aligned.
1017
1018=item C<heads: 2192>
1019
1020Although memory overhead of bigger buckets is kept inside the bucket, for
1021smaller buckets, it is kept in separate areas.  This field gives the
1022total size of these areas.
1023
1024=item C<chain: 0>
1025
1026malloc() may want to subdivide a bigger bucket into smaller buckets.
1027If only a part of the deceased bucket is left unsubdivided, the rest
1028is kept as an element of a linked list.  This field gives the total
1029size of these chunks.
1030
1031=item C<tail: 6144>
1032
1033To minimize the number of sbrk(2)s, malloc() asks for more memory.  This
1034field gives the size of the yet unused part, which is sbrk(2)ed, but
1035never touched.
1036
1037=back
1038
1039=head1 SEE ALSO
1040
1041L<perldebug>,
1042L<perlguts>,
1043L<perlrun>
1044L<re>,
1045and
1046L<Devel::DProf>.
1047