xref: /openbsd-src/gnu/usr.bin/perl/pod/perldebguts.pod (revision 4e1ee0786f11cc571bd0be17d38e46f635c719fc)
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|perlrun/-Dletters>,
18which is usable only if a special Perl is built per the instructions in
19the F<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 $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
402  scalar context return from Exporter::import: ''
403
404=back
405
406In all cases shown above, the line indentation shows the call tree.
407If bit 2 of C<frame> is set, a line is printed on exit from a
408subroutine as well.  If bit 4 is set, the arguments are printed
409along with the caller info.  If bit 8 is set, the arguments are
410printed even if they are tied or references.  If bit 16 is set, the
411return value is printed, too.
412
413When a package is compiled, a line like this
414
415    Package lib/Carp.pm.
416
417is printed with proper indentation.
418
419=head1 Debugging Regular Expressions
420
421There are two ways to enable debugging output for regular expressions.
422
423If your perl is compiled with C<-DDEBUGGING>, you may use the
424B<-Dr> flag on the command line, and C<-Drv> for more verbose
425information.
426
427Otherwise, one can C<use re 'debug'>, which has effects at both
428compile time and run time.  Since Perl 5.9.5, this pragma is lexically
429scoped.
430
431=head2 Compile-time Output
432
433The debugging output at compile time looks like this:
434
435  Compiling REx '[bc]d(ef*g)+h[ij]k$'
436  size 45 Got 364 bytes for offset annotations.
437  first at 1
438  rarest char g at 0
439  rarest char d at 0
440     1: ANYOF[bc](12)
441    12: EXACT <d>(14)
442    14: CURLYX[0] {1,32767}(28)
443    16:   OPEN1(18)
444    18:     EXACT <e>(20)
445    20:     STAR(23)
446    21:       EXACT <f>(0)
447    23:     EXACT <g>(25)
448    25:   CLOSE1(27)
449    27:   WHILEM[1/1](0)
450    28: NOTHING(29)
451    29: EXACT <h>(31)
452    31: ANYOF[ij](42)
453    42: EXACT <k>(44)
454    44: EOL(45)
455    45: END(0)
456  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
457        stclass 'ANYOF[bc]' minlen 7
458  Offsets: [45]
459  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
460  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
461  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
462  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
463  Omitting $` $& $' support.
464
465The first line shows the pre-compiled form of the regex.  The second
466shows the size of the compiled form (in arbitrary units, usually
4674-byte words) and the total number of bytes allocated for the
468offset/length table, usually 4+C<size>*8.  The next line shows the
469label I<id> of the first node that does a match.
470
471The
472
473  anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
474        stclass 'ANYOF[bc]' minlen 7
475
476line (split into two lines above) contains optimizer
477information.  In the example shown, the optimizer found that the match
478should contain a substring C<de> at offset 1, plus substring C<gh>
479at some offset between 3 and infinity.  Moreover, when checking for
480these substrings (to abandon impossible matches quickly), Perl will check
481for the substring C<gh> before checking for the substring C<de>.  The
482optimizer may also use the knowledge that the match starts (at the
483C<first> I<id>) with a character class, and no string
484shorter than 7 characters can possibly match.
485
486The fields of interest which may appear in this line are
487
488=over 4
489
490=item C<anchored> I<STRING> C<at> I<POS>
491
492=item C<floating> I<STRING> C<at> I<POS1..POS2>
493
494See above.
495
496=item C<matching floating/anchored>
497
498Which substring to check first.
499
500=item C<minlen>
501
502The minimal length of the match.
503
504=item C<stclass> I<TYPE>
505
506Type of first matching node.
507
508=item C<noscan>
509
510Don't scan for the found substrings.
511
512=item C<isall>
513
514Means that the optimizer information is all that the regular
515expression contains, and thus one does not need to enter the regex engine at
516all.
517
518=item C<GPOS>
519
520Set if the pattern contains C<\G>.
521
522=item C<plus>
523
524Set if the pattern starts with a repeated char (as in C<x+y>).
525
526=item C<implicit>
527
528Set if the pattern starts with C<.*>.
529
530=item C<with eval>
531
532Set if the pattern contain eval-groups, such as C<(?{ code })> and
533C<(??{ code })>.
534
535=item C<anchored(TYPE)>
536
537If the pattern may match only at a handful of places, with C<TYPE>
538being C<SBOL>, C<MBOL>, or C<GPOS>.  See the table below.
539
540=back
541
542If a substring is known to match at end-of-line only, it may be
543followed by C<$>, as in C<floating 'k'$>.
544
545The optimizer-specific information is used to avoid entering (a slow) regex
546engine on strings that will not definitely match.  If the C<isall> flag
547is set, a call to the regex engine may be avoided even when the optimizer
548found an appropriate place for the match.
549
550Above the optimizer section is the list of I<nodes> of the compiled
551form of the regex.  Each line has format
552
553C<   >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
554
555=head2 Types of Nodes
556
557Here are the current possible types, with short descriptions:
558
559=for comment
560This table is generated by regen/regcomp.pl.  Any changes made here
561will be lost.
562
563=for regcomp.pl begin
564
565 # TYPE arg-description [regnode-struct-suffix] [longjump-len] DESCRIPTION
566
567 # Exit points
568
569 END              no         End of program.
570 SUCCEED          no         Return from a subroutine, basically.
571
572 # Line Start Anchors:
573 SBOL             no         Match "" at beginning of line: /^/, /\A/
574 MBOL             no         Same, assuming multiline: /^/m
575
576 # Line End Anchors:
577 SEOL             no         Match "" at end of line: /$/
578 MEOL             no         Same, assuming multiline: /$/m
579 EOS              no         Match "" at end of string: /\z/
580
581 # Match Start Anchors:
582 GPOS             no         Matches where last m//g left off.
583
584 # Word Boundary Opcodes:
585 BOUND            no         Like BOUNDA for non-utf8, otherwise like
586                             BOUNDU
587 BOUNDL           no         Like BOUND/BOUNDU, but \w and \W are
588                             defined by current locale
589 BOUNDU           no         Match "" at any boundary of a given type
590                             using /u rules.
591 BOUNDA           no         Match "" at any boundary between \w\W or
592                             \W\w, where \w is [_a-zA-Z0-9]
593 NBOUND           no         Like NBOUNDA for non-utf8, otherwise like
594                             BOUNDU
595 NBOUNDL          no         Like NBOUND/NBOUNDU, but \w and \W are
596                             defined by current locale
597 NBOUNDU          no         Match "" at any non-boundary of a given
598                             type using using /u rules.
599 NBOUNDA          no         Match "" betweeen any \w\w or \W\W, where
600                             \w is [_a-zA-Z0-9]
601
602 # [Special] alternatives:
603 REG_ANY          no         Match any one character (except newline).
604 SANY             no         Match any one character.
605 ANYOF            sv         Match character in (or not in) this class,
606                  charclass  single char match only
607 ANYOFD           sv         Like ANYOF, but /d is in effect
608                  charclass
609 ANYOFL           sv         Like ANYOF, but /l is in effect
610                  charclass
611 ANYOFPOSIXL      sv         Like ANYOFL, but matches [[:posix:]]
612                  charclass_ classes
613                  posixl
614
615 ANYOFH           sv 1       Like ANYOF, but only has "High" matches,
616                             none in the bitmap; the flags field
617                             contains the lowest matchable UTF-8 start
618                             byte
619 ANYOFHb          sv 1       Like ANYOFH, but all matches share the same
620                             UTF-8 start byte, given in the flags field
621 ANYOFHr          sv 1       Like ANYOFH, but the flags field contains
622                             packed bounds for all matchable UTF-8 start
623                             bytes.
624 ANYOFHs          sv 1       Like ANYOFHb, but has a string field that
625                             gives the leading matchable UTF-8 bytes;
626                             flags field is len
627 ANYOFR           packed 1   Matches any character in the range given by
628                             its packed args: upper 12 bits is the max
629                             delta from the base lower 20; the flags
630                             field contains the lowest matchable UTF-8
631                             start byte
632 ANYOFRb          packed 1   Like ANYOFR, but all matches share the same
633                             UTF-8 start byte, given in the flags field
634
635 ANYOFM           byte 1     Like ANYOF, but matches an invariant byte
636                             as determined by the mask and arg
637 NANYOFM          byte 1     complement of ANYOFM
638
639 # POSIX Character Classes:
640 POSIXD           none       Some [[:class:]] under /d; the FLAGS field
641                             gives which one
642 POSIXL           none       Some [[:class:]] under /l; the FLAGS field
643                             gives which one
644 POSIXU           none       Some [[:class:]] under /u; the FLAGS field
645                             gives which one
646 POSIXA           none       Some [[:class:]] under /a; the FLAGS field
647                             gives which one
648 NPOSIXD          none       complement of POSIXD, [[:^class:]]
649 NPOSIXL          none       complement of POSIXL, [[:^class:]]
650 NPOSIXU          none       complement of POSIXU, [[:^class:]]
651 NPOSIXA          none       complement of POSIXA, [[:^class:]]
652
653 CLUMP            no         Match any extended grapheme cluster
654                             sequence
655
656 # Alternation
657
658 # BRANCH        The set of branches constituting a single choice are
659 #               hooked together with their "next" pointers, since
660 #               precedence prevents anything being concatenated to
661 #               any individual branch.  The "next" pointer of the last
662 #               BRANCH in a choice points to the thing following the
663 #               whole choice.  This is also where the final "next"
664 #               pointer of each individual branch points; each branch
665 #               starts with the operand node of a BRANCH node.
666 #
667 BRANCH           node       Match this alternative, or the next...
668
669 # Literals
670
671 EXACT            str        Match this string (flags field is the
672                             length).
673
674 # In a long string node, the U32 argument is the length, and is
675 # immediately followed by the string.
676 LEXACT           len:str 1  Match this long string (preceded by length;
677                             flags unused).
678 EXACTL           str        Like EXACT, but /l is in effect (used so
679                             locale-related warnings can be checked for)
680 EXACTF           str        Like EXACT, but match using /id rules;
681                             (string not UTF-8, ASCII folded; non-ASCII
682                             not)
683 EXACTFL          str        Like EXACT, but match using /il rules;
684                             (string not likely to be folded)
685 EXACTFU          str        Like EXACT, but match using /iu rules;
686                             (string folded)
687
688 EXACTFAA         str        Like EXACT, but match using /iaa rules;
689                             (string folded except in non-UTF8 patterns:
690                             MICRO, SHARP S; folded length <= unfolded)
691
692 EXACTFUP         str        Like EXACT, but match using /iu rules;
693                             (string not UTF-8, folded except MICRO,
694                             SHARP S: hence Problematic)
695
696 EXACTFLU8        str        Like EXACTFU, but use /il, UTF-8, (string
697                             is folded, and everything in it is above
698                             255
699 EXACTFAA_NO_TRIE str        Like EXACT, but match using /iaa rules
700                             (string not UTF-8, not guaranteed to be
701                             folded, not currently trie-able)
702
703 EXACT_REQ8       str        Like EXACT, but only UTF-8 encoded targets
704                             can match
705 LEXACT_REQ8      len:str 1  Like LEXACT, but only UTF-8 encoded targets
706                             can match
707 EXACTFU_REQ8     str        Like EXACTFU, but only UTF-8 encoded
708                             targets can match
709
710 EXACTFU_S_EDGE   str        /di rules, but nothing in it precludes /ui,
711                             except begins and/or ends with [Ss];
712                             (string not UTF-8; compile-time only)
713
714 # Do nothing types
715
716 NOTHING          no         Match empty string.
717 # A variant of above which delimits a group, thus stops optimizations
718 TAIL             no         Match empty string. Can jump here from
719                             outside.
720
721 # Loops
722
723 # STAR,PLUS    '?', and complex '*' and '+', are implemented as
724 #               circular BRANCH structures.  Simple cases
725 #               (one character per match) are implemented with STAR
726 #               and PLUS for speed and to minimize recursive plunges.
727 #
728 STAR             node       Match this (simple) thing 0 or more times.
729 PLUS             node       Match this (simple) thing 1 or more times.
730
731 CURLY            sv 2       Match this simple thing {n,m} times.
732 CURLYN           no 2       Capture next-after-this simple thing
733 CURLYM           no 2       Capture this medium-complex thing {n,m}
734                             times.
735 CURLYX           sv 2       Match this complex thing {n,m} times.
736
737 # This terminator creates a loop structure for CURLYX
738 WHILEM           no         Do curly processing and see if rest
739                             matches.
740
741 # Buffer related
742
743 # OPEN,CLOSE,GROUPP     ...are numbered at compile time.
744 OPEN             num 1      Mark this point in input as start of #n.
745 CLOSE            num 1      Close corresponding OPEN of #n.
746 SROPEN           none       Same as OPEN, but for script run
747 SRCLOSE          none       Close preceding SROPEN
748
749 REF              num 1      Match some already matched string
750 REFF             num 1      Match already matched string, using /di
751                             rules.
752 REFFL            num 1      Match already matched string, using /li
753                             rules.
754 REFFU            num 1      Match already matched string, usng /ui.
755 REFFA            num 1      Match already matched string, using /aai
756                             rules.
757
758 # Named references.  Code in regcomp.c assumes that these all are after
759 # the numbered references
760 REFN             no-sv 1    Match some already matched string
761 REFFN            no-sv 1    Match already matched string, using /di
762                             rules.
763 REFFLN           no-sv 1    Match already matched string, using /li
764                             rules.
765 REFFUN           num 1      Match already matched string, using /ui
766                             rules.
767 REFFAN           num 1      Match already matched string, using /aai
768                             rules.
769
770 # Support for long RE
771 LONGJMP          off 1 1    Jump far away.
772 BRANCHJ          off 1 1    BRANCH with long offset.
773
774 # Special Case Regops
775 IFMATCH          off 1 1    Succeeds if the following matches; non-zero
776                             flags "f", next_off "o" means lookbehind
777                             assertion starting "f..(f-o)" characters
778                             before current
779 UNLESSM          off 1 1    Fails if the following matches; non-zero
780                             flags "f", next_off "o" means lookbehind
781                             assertion starting "f..(f-o)" characters
782                             before current
783 SUSPEND          off 1 1    "Independent" sub-RE.
784 IFTHEN           off 1 1    Switch, should be preceded by switcher.
785 GROUPP           num 1      Whether the group matched.
786
787 # The heavy worker
788
789 EVAL             evl/flags  Execute some Perl code.
790                  2L
791
792 # Modifiers
793
794 MINMOD           no         Next operator is not greedy.
795 LOGICAL          no         Next opcode should set the flag only.
796
797 # This is not used yet
798 RENUM            off 1 1    Group with independently numbered parens.
799
800 # Trie Related
801
802 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
803 # have inline charclass data (ascii only), the 'C' store it in the
804 # structure.
805
806 TRIE             trie 1     Match many EXACT(F[ALU]?)? at once.
807                             flags==type
808 TRIEC            trie       Same as TRIE, but with embedded charclass
809                  charclass  data
810
811 AHOCORASICK      trie 1     Aho Corasick stclass. flags==type
812 AHOCORASICKC     trie       Same as AHOCORASICK, but with embedded
813                  charclass  charclass data
814
815 # Regex Subroutines
816 GOSUB            num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
817
818 # Special conditionals
819 GROUPPN          no-sv 1    Whether the group matched.
820 INSUBP           num 1      Whether we are in a specific recurse.
821 DEFINEP          none 1     Never execute directly.
822
823 # Backtracking Verbs
824 ENDLIKE          none       Used only for the type field of verbs
825 OPFAIL           no-sv 1    Same as (?!), but with verb arg
826 ACCEPT           no-sv/num  Accepts the current matched string, with
827                  2L         verbar
828
829 # Verbs With Arguments
830 VERB             no-sv 1    Used only for the type field of verbs
831 PRUNE            no-sv 1    Pattern fails at this startpoint if no-
832                             backtracking through this
833 MARKPOINT        no-sv 1    Push the current location for rollback by
834                             cut.
835 SKIP             no-sv 1    On failure skip forward (to the mark)
836                             before retrying
837 COMMIT           no-sv 1    Pattern fails outright if backtracking
838                             through this
839 CUTGROUP         no-sv 1    On failure go to the next alternation in
840                             the group
841
842 # Control what to keep in $&.
843 KEEPS            no         $& begins here.
844
845 # New charclass like patterns
846 LNBREAK          none       generic newline pattern
847
848 # SPECIAL  REGOPS
849
850 # This is not really a node, but an optimized away piece of a "long"
851 # node.  To simplify debugging output, we mark it as if it were a node
852 OPTIMIZED        off        Placeholder for dump.
853
854 # Special opcode with the property that no opcode in a compiled program
855 # will ever be of this type. Thus it can be used as a flag value that
856 # no other opcode has been seen. END is used similarly, in that an END
857 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
858 # mean "not seen anything to optimize yet".
859 PSEUDO           off        Pseudo opcode for internal use.
860
861 REGEX_SET        depth p    Regex set, temporary node used in pre-
862                             optimization compilation
863
864=for regcomp.pl end
865
866=for unprinted-credits
867Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
868
869Following the optimizer information is a dump of the offset/length
870table, here split across several lines:
871
872  Offsets: [45]
873  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
874  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
875  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
876  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
877
878The first line here indicates that the offset/length table contains 45
879entries.  Each entry is a pair of integers, denoted by C<offset[length]>.
880Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
881entry #12 is C<5[1]>.  C<1[4]> indicates that the node labeled C<1:>
882(the C<1: ANYOF[bc]>) begins at character position 1 in the
883pre-compiled form of the regex, and has a length of 4 characters.
884C<5[1]> in position 12
885indicates that the node labeled C<12:>
886(the C<< 12: EXACT <d> >>) begins at character position 5 in the
887pre-compiled form of the regex, and has a length of 1 character.
888C<12[1]> in position 14
889indicates that the node labeled C<14:>
890(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
891pre-compiled form of the regex, and has a length of 1 character---that
892is, it corresponds to the C<+> symbol in the precompiled regex.
893
894C<0[0]> items indicate that there is no corresponding node.
895
896=head2 Run-time Output
897
898First of all, when doing a match, one may get no run-time output even
899if debugging is enabled.  This means that the regex engine was never
900entered and that all of the job was therefore done by the optimizer.
901
902If the regex engine was entered, the output may look like this:
903
904  Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
905    Setting an EVAL scope, savestack=3
906     2 <ab> <cdefg__gh_>    |  1: ANYOF
907     3 <abc> <defg__gh_>    | 11: EXACT <d>
908     4 <abcd> <efg__gh_>    | 13: CURLYX {1,32767}
909     4 <abcd> <efg__gh_>    | 26:   WHILEM
910				0 out of 1..32767  cc=effff31c
911     4 <abcd> <efg__gh_>    | 15:     OPEN1
912     4 <abcd> <efg__gh_>    | 17:     EXACT <e>
913     5 <abcde> <fg__gh_>    | 19:     STAR
914			     EXACT <f> can match 1 times out of 32767...
915    Setting an EVAL scope, savestack=3
916     6 <bcdef> <g__gh__>    | 22:       EXACT <g>
917     7 <bcdefg> <__gh__>    | 24:       CLOSE1
918     7 <bcdefg> <__gh__>    | 26:       WHILEM
919				    1 out of 1..32767  cc=effff31c
920    Setting an EVAL scope, savestack=12
921     7 <bcdefg> <__gh__>    | 15:         OPEN1
922     7 <bcdefg> <__gh__>    | 17:         EXACT <e>
923       restoring \1 to 4(4)..7
924				    failed, try continuation...
925     7 <bcdefg> <__gh__>    | 27:         NOTHING
926     7 <bcdefg> <__gh__>    | 28:         EXACT <h>
927				    failed...
928				failed...
929
930The most significant information in the output is about the particular I<node>
931of the compiled regex that is currently being tested against the target string.
932The format of these lines is
933
934C<    >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>>   |I<ID>:  I<TYPE>
935
936The I<TYPE> info is indented with respect to the backtracking level.
937Other incidental information appears interspersed within.
938
939=head1 Debugging Perl Memory Usage
940
941Perl is a profligate wastrel when it comes to memory use.  There
942is a saying that to estimate memory usage of Perl, assume a reasonable
943algorithm for memory allocation, multiply that estimate by 10, and
944while you still may miss the mark, at least you won't be quite so
945astonished.  This is not absolutely true, but may provide a good
946grasp of what happens.
947
948Assume that an integer cannot take less than 20 bytes of memory, a
949float cannot take less than 24 bytes, a string cannot take less
950than 32 bytes (all these examples assume 32-bit architectures, the
951result are quite a bit worse on 64-bit architectures).  If a variable
952is accessed in two of three different ways (which require an integer,
953a float, or a string), the memory footprint may increase yet another
95420 bytes.  A sloppy malloc(3) implementation can inflate these
955numbers dramatically.
956
957On the opposite end of the scale, a declaration like
958
959  sub foo;
960
961may take up to 500 bytes of memory, depending on which release of Perl
962you're running.
963
964Anecdotal estimates of source-to-compiled code bloat suggest an
965eightfold increase.  This means that the compiled form of reasonable
966(normally commented, properly indented etc.) code will take
967about eight times more space in memory than the code took
968on disk.
969
970The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
971(it was available only if Perl was built with C<-DDEBUGGING>).
972The switch was used to track Perl's memory allocations and possible
973memory leaks.  These days the use of malloc debugging tools like
974F<Purify> or F<valgrind> is suggested instead.  See also
975L<perlhacktips/PERL_MEM_LOG>.
976
977One way to find out how much memory is being used by Perl data
978structures is to install the Devel::Size module from CPAN: it gives
979you the minimum number of bytes required to store a particular data
980structure.  Please be mindful of the difference between the size()
981and total_size().
982
983If Perl has been compiled using Perl's malloc you can analyze Perl
984memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
985
986=head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
987
988If your perl is using Perl's malloc() and was compiled with the
989necessary switches (this is the default), then it will print memory
990usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
991> 1 >>, and before termination of the program when C<<
992$ENV{PERL_DEBUG_MSTATS} >= 1 >>.  The report format is similar to
993the following example:
994
995 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
996 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
997    14216 free:   130   117    28     7     9   0   2     2   1 0 0
998		437    61    36     0     5
999    60924 used:   125   137   161    55     7   8   6    16   2 0 1
1000		 74   109   304    84    20
1001 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
1002 Memory allocation statistics after execution:   (buckets 4(4)..8188(8192)
1003    30888 free:   245    78    85    13     6   2   1     3   2 0 1
1004		315   162    39    42    11
1005   175816 used:   265   176  1112   111    26  22  11    27   2 1 1
1006		196   178  1066   798    39
1007 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
1008
1009It is possible to ask for such a statistic at arbitrary points in
1010your execution using the mstat() function out of the standard
1011Devel::Peek module.
1012
1013Here is some explanation of that format:
1014
1015=over 4
1016
1017=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
1018
1019Perl's malloc() uses bucketed allocations.  Every request is rounded
1020up to the closest bucket size available, and a bucket is taken from
1021the pool of buckets of that size.
1022
1023The line above describes the limits of buckets currently in use.
1024Each bucket has two sizes: memory footprint and the maximal size
1025of user data that can fit into this bucket.  Suppose in the above
1026example that the smallest bucket were size 4.  The biggest bucket
1027would have usable size 8188, and the memory footprint would be 8192.
1028
1029In a Perl built for debugging, some buckets may have negative usable
1030size.  This means that these buckets cannot (and will not) be used.
1031For larger buckets, the memory footprint may be one page greater
1032than a power of 2.  If so, the corresponding power of two is
1033printed in the C<APPROX> field above.
1034
1035=item Free/Used
1036
1037The 1 or 2 rows of numbers following that correspond to the number
1038of buckets of each size between C<SMALLEST> and C<GREATEST>.  In
1039the first row, the sizes (memory footprints) of buckets are powers
1040of two--or possibly one page greater.  In the second row, if present,
1041the memory footprints of the buckets are between the memory footprints
1042of two buckets "above".
1043
1044For example, suppose under the previous example, the memory footprints
1045were
1046
1047   free:    8     16    32    64    128  256 512 1024 2048 4096 8192
1048	   4     12    24    48    80
1049
1050With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
1051a 4-byte overhead, and thus an 8192-long bucket may take up to
10528188-byte allocations.
1053
1054=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
1055
1056The first two fields give the total amount of memory perl sbrk(2)ed
1057(ess-broken? :-) and number of sbrk(2)s used.  The third number is
1058what perl thinks about continuity of returned chunks.  So long as
1059this number is positive, malloc() will assume that it is probable
1060that sbrk(2) will provide continuous memory.
1061
1062Memory allocated by external libraries is not counted.
1063
1064=item C<pad: 0>
1065
1066The amount of sbrk(2)ed memory needed to keep buckets aligned.
1067
1068=item C<heads: 2192>
1069
1070Although memory overhead of bigger buckets is kept inside the bucket, for
1071smaller buckets, it is kept in separate areas.  This field gives the
1072total size of these areas.
1073
1074=item C<chain: 0>
1075
1076malloc() may want to subdivide a bigger bucket into smaller buckets.
1077If only a part of the deceased bucket is left unsubdivided, the rest
1078is kept as an element of a linked list.  This field gives the total
1079size of these chunks.
1080
1081=item C<tail: 6144>
1082
1083To minimize the number of sbrk(2)s, malloc() asks for more memory.  This
1084field gives the size of the yet unused part, which is sbrk(2)ed, but
1085never touched.
1086
1087=back
1088
1089=head1 SEE ALSO
1090
1091L<perldebug>,
1092L<perlguts>,
1093L<perlrun>
1094L<re>,
1095and
1096L<Devel::DProf>.
1097