3 perldebguts - Guts of Perl debugging
7 This is not L<perldebug>, which tells you how to use
8 the debugger. This manpage describes low-level details concerning
9 the debugger's internals, which range from difficult to impossible
10 to understand for anyone who isn't incredibly intimate with Perl's guts.
13 =head1 Debugger Internals
15 Perl has special debugging hooks at compile-time and run-time used
16 to create debugging environments. These hooks are not to be confused
17 with the I<perl -Dxxx> command described in L<perlrun>, which is
18 usable only if a special Perl is built per the instructions in the
19 F<INSTALL> podpage in the Perl source tree.
21 For example, whenever you call Perl's built-in C<caller> function
22 from the package C<DB>, the arguments that the corresponding stack
23 frame was called with are copied to the C<@DB::args> array. These
24 mechanisms are enabled by calling Perl with the B<-d> switch.
25 Specifically, the following additional features are enabled
32 Perl inserts the contents of C<$ENV{PERL5DB}> (or C<BEGIN {require
33 'perl5db.pl'}> if not present) before the first line of your program.
37 Each array C<@{"_<$filename"}> holds the lines of $filename for a
38 file compiled by Perl. The same is also true for C<eval>ed strings
39 that contain subroutines, or which are currently being executed.
40 The $filename for C<eval>ed strings looks like C<(eval 34)>.
41 Code assertions in regexes look like C<(re_eval 19)>.
43 Values in this array are magical in numeric context: they compare
44 equal to zero only if the line is not breakable.
48 Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed
49 by line number. Individual entries (as opposed to the whole hash)
50 are settable. Perl only cares about Boolean true here, although
51 the values used by F<perl5db.pl> have the form
52 C<"$break_condition\0$action">.
54 The same holds for evaluated strings that contain subroutines, or
55 which are currently being executed. The $filename for C<eval>ed strings
56 looks like C<(eval 34)> or C<(re_eval 19)>.
60 Each scalar C<${"_<$filename"}> contains C<"_<$filename">. This is
61 also the case for evaluated strings that contain subroutines, or
62 which are currently being executed. The $filename for C<eval>ed
63 strings looks like C<(eval 34)> or C<(re_eval 19)>.
67 After each C<require>d file is compiled, but before it is executed,
68 C<DB::postponed(*{"_<$filename"})> is called if the subroutine
69 C<DB::postponed> exists. Here, the $filename is the expanded name of
70 the C<require>d file, as found in the values of %INC.
74 After each subroutine C<subname> is compiled, the existence of
75 C<$DB::postponed{subname}> is checked. If this key exists,
76 C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine
81 A hash C<%DB::sub> is maintained, whose keys are subroutine names
82 and whose values have the form C<filename:startline-endline>.
83 C<filename> has the form C<(eval 34)> for subroutines defined inside
84 C<eval>s, or C<(re_eval 19)> for those within regex code assertions.
88 When the execution of your program reaches a point that can hold a
89 breakpoint, the C<DB::DB()> subroutine is called if any of the variables
90 C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true. These variables
91 are not C<local>izable. This feature is disabled when executing
92 inside C<DB::DB()>, including functions called from it
93 unless C<< $^D & (1<<30) >> is true.
97 When execution of the program reaches a subroutine call, a call to
98 C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> holding the
99 name of the called subroutine. (This doesn't happen if the subroutine
100 was compiled in the C<DB> package.)
104 Note that if C<&DB::sub> needs external data for it to work, no
105 subroutine call is possible without it. As an example, the standard
106 debugger's C<&DB::sub> depends on the C<$DB::deep> variable
107 (it defines how many levels of recursion deep into the debugger you can go
108 before a mandatory break). If C<$DB::deep> is not defined, subroutine
109 calls are not possible, even though C<&DB::sub> exists.
111 =head2 Writing Your Own Debugger
113 =head3 Environment Variables
115 The C<PERL5DB> environment variable can be used to define a debugger.
116 For example, the minimal "working" debugger (it actually doesn't do anything)
117 consists of one line:
121 It can easily be defined like this:
123 $ PERL5DB="sub DB::DB {}" perl -d your-script
125 Another brief debugger, slightly more useful, can be created
128 sub DB::DB {print ++$i; scalar <STDIN>}
130 This debugger prints a number which increments for each statement
131 encountered and waits for you to hit a newline before continuing
132 to the next statement.
134 The following debugger is actually useful:
139 sub sub {print ++$i, " $sub\n"; &$sub}
142 It prints the sequence number of each subroutine call and the name of the
143 called subroutine. Note that C<&DB::sub> is being compiled into the
144 package C<DB> through the use of the C<package> directive.
146 When it starts, the debugger reads your rc file (F<./.perldb> or
147 F<~/.perldb> under Unix), which can set important options.
148 (A subroutine (C<&afterinit>) can be defined here as well; it is executed
149 after the debugger completes its own initialization.)
151 After the rc file is read, the debugger reads the PERLDB_OPTS
152 environment variable and uses it to set debugger options. The
153 contents of this variable are treated as if they were the argument
154 of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
156 =head3 Debugger Internal Variables
158 In addition to the file and subroutine-related variables mentioned above,
159 the debugger also maintains various magical internal variables.
165 C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
166 holds the lines of the currently-selected file (compiled by Perl), either
167 explicitly chosen with the debugger's C<f> command, or implicitly by flow
170 Values in this array are magical in numeric context: they compare
171 equal to zero only if the line is not breakable.
175 C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
176 contains breakpoints and actions keyed by line number in
177 the currently-selected file, either explicitly chosen with the
178 debugger's C<f> command, or implicitly by flow of execution.
180 As previously noted, individual entries (as opposed to the whole hash)
181 are settable. Perl only cares about Boolean true here, although
182 the values used by F<perl5db.pl> have the form
183 C<"$break_condition\0$action">.
187 =head3 Debugger Customization Functions
189 Some functions are provided to simplify customization.
195 See L<perldebug/"Configurable Options"> for a description of options parsed by
196 C<DB::parse_options(string)>.
200 C<DB::dump_trace(skip[,count])> skips the specified number of frames
201 and returns a list containing information about the calling frames (all
202 of them, if C<count> is missing). Each entry is reference to a hash
203 with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
204 name, or info about C<eval>), C<args> (C<undef> or a reference to
205 an array), C<file>, and C<line>.
209 C<DB::print_trace(FH, skip[, count[, short]])> prints
210 formatted info about caller frames. The last two functions may be
211 convenient as arguments to C<< < >>, C<< << >> commands.
215 Note that any variables and functions that are not documented in
216 this manpages (or in L<perldebug>) are considered for internal
217 use only, and as such are subject to change without notice.
219 =head1 Frame Listing Output Examples
221 The C<frame> option can be used to control the output of frame
222 information. For example, contrast this expression trace:
225 Stack dump during die enabled outside of evals.
227 Loading DB routines from perl5db.pl patch level 0.94
228 Emacs support available.
230 Enter h or `h h' for help.
237 DB<3> t print foo() * bar()
238 main::((eval 172):3): print foo() + bar();
239 main::foo((eval 168):2):
240 main::bar((eval 170):2):
243 with this one, once the C<o>ption C<frame=2> has been set:
247 DB<5> t print foo() * bar()
257 By way of demonstration, we present below a laborious listing
258 resulting from setting your C<PERLDB_OPTS> environment variable to
259 the value C<f=n N>, and running I<perl -d -V> from the command line.
260 Examples using various values of C<n> are shown to give you a feel
261 for the difference between settings. Long though it may be, this
262 is not a complete listing, but only excerpts.
269 entering Config::BEGIN
270 Package lib/Exporter.pm.
272 Package lib/Config.pm.
273 entering Config::TIEHASH
274 entering Exporter::import
275 entering Exporter::export
276 entering Config::myconfig
277 entering Config::FETCH
278 entering Config::FETCH
279 entering Config::FETCH
280 entering Config::FETCH
285 entering Config::BEGIN
286 Package lib/Exporter.pm.
289 Package lib/Config.pm.
290 entering Config::TIEHASH
291 exited Config::TIEHASH
292 entering Exporter::import
293 entering Exporter::export
294 exited Exporter::export
295 exited Exporter::import
297 entering Config::myconfig
298 entering Config::FETCH
300 entering Config::FETCH
302 entering Config::FETCH
306 in $=main::BEGIN() from /dev/null:0
307 in $=Config::BEGIN() from lib/Config.pm:2
308 Package lib/Exporter.pm.
310 Package lib/Config.pm.
311 in $=Config::TIEHASH('Config') from lib/Config.pm:644
312 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
313 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
314 in @=Config::myconfig() from /dev/null:0
315 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
316 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
317 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
318 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
319 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
320 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
324 in $=main::BEGIN() from /dev/null:0
325 in $=Config::BEGIN() from lib/Config.pm:2
326 Package lib/Exporter.pm.
328 out $=Config::BEGIN() from lib/Config.pm:0
329 Package lib/Config.pm.
330 in $=Config::TIEHASH('Config') from lib/Config.pm:644
331 out $=Config::TIEHASH('Config') from lib/Config.pm:644
332 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
333 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
334 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
335 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
336 out $=main::BEGIN() from /dev/null:0
337 in @=Config::myconfig() from /dev/null:0
338 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
339 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
340 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
341 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
342 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
343 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
344 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
348 in $=main::BEGIN() from /dev/null:0
349 in $=Config::BEGIN() from lib/Config.pm:2
350 Package lib/Exporter.pm.
352 out $=Config::BEGIN() from lib/Config.pm:0
353 Package lib/Config.pm.
354 in $=Config::TIEHASH('Config') from lib/Config.pm:644
355 out $=Config::TIEHASH('Config') from lib/Config.pm:644
356 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
357 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
358 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
359 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
360 out $=main::BEGIN() from /dev/null:0
361 in @=Config::myconfig() from /dev/null:0
362 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
363 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
364 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
365 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
369 in $=CODE(0x15eca4)() from /dev/null:0
370 in $=CODE(0x182528)() from lib/Config.pm:2
371 Package lib/Exporter.pm.
372 out $=CODE(0x182528)() from lib/Config.pm:0
373 scalar context return from CODE(0x182528): undef
374 Package lib/Config.pm.
375 in $=Config::TIEHASH('Config') from lib/Config.pm:628
376 out $=Config::TIEHASH('Config') from lib/Config.pm:628
377 scalar context return from Config::TIEHASH: empty hash
378 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
379 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
380 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
381 scalar context return from Exporter::export: ''
382 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
383 scalar context return from Exporter::import: ''
387 In all cases shown above, the line indentation shows the call tree.
388 If bit 2 of C<frame> is set, a line is printed on exit from a
389 subroutine as well. If bit 4 is set, the arguments are printed
390 along with the caller info. If bit 8 is set, the arguments are
391 printed even if they are tied or references. If bit 16 is set, the
392 return value is printed, too.
394 When a package is compiled, a line like this
398 is printed with proper indentation.
400 =head1 Debugging Regular Expressions
402 There are two ways to enable debugging output for regular expressions.
404 If your perl is compiled with C<-DDEBUGGING>, you may use the
405 B<-Dr> flag on the command line.
407 Otherwise, one can C<use re 'debug'>, which has effects at
408 compile time and run time. Since Perl 5.9.5, this pragma is lexically
411 =head2 Compile-time Output
413 The debugging output at compile time looks like this:
415 Compiling REx `[bc]d(ef*g)+h[ij]k$'
416 size 45 Got 364 bytes for offset annotations.
422 14: CURLYX[0] {1,32767}(28)
436 anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
437 stclass `ANYOF[bc]' minlen 7
439 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
440 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
441 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
442 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
443 Omitting $` $& $' support.
445 The first line shows the pre-compiled form of the regex. The second
446 shows the size of the compiled form (in arbitrary units, usually
447 4-byte words) and the total number of bytes allocated for the
448 offset/length table, usually 4+C<size>*8. The next line shows the
449 label I<id> of the first node that does a match.
453 anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
454 stclass `ANYOF[bc]' minlen 7
456 line (split into two lines above) contains optimizer
457 information. In the example shown, the optimizer found that the match
458 should contain a substring C<de> at offset 1, plus substring C<gh>
459 at some offset between 3 and infinity. Moreover, when checking for
460 these substrings (to abandon impossible matches quickly), Perl will check
461 for the substring C<gh> before checking for the substring C<de>. The
462 optimizer may also use the knowledge that the match starts (at the
463 C<first> I<id>) with a character class, and no string
464 shorter than 7 characters can possibly match.
466 The fields of interest which may appear in this line are
470 =item C<anchored> I<STRING> C<at> I<POS>
472 =item C<floating> I<STRING> C<at> I<POS1..POS2>
476 =item C<matching floating/anchored>
478 Which substring to check first.
482 The minimal length of the match.
484 =item C<stclass> I<TYPE>
486 Type of first matching node.
490 Don't scan for the found substrings.
494 Means that the optimizer information is all that the regular
495 expression contains, and thus one does not need to enter the regex engine at
500 Set if the pattern contains C<\G>.
504 Set if the pattern starts with a repeated char (as in C<x+y>).
508 Set if the pattern starts with C<.*>.
512 Set if the pattern contain eval-groups, such as C<(?{ code })> and
515 =item C<anchored(TYPE)>
517 If the pattern may match only at a handful of places, with C<TYPE>
518 being C<BOL>, C<MBOL>, or C<GPOS>. See the table below.
522 If a substring is known to match at end-of-line only, it may be
523 followed by C<$>, as in C<floating `k'$>.
525 The optimizer-specific information is used to avoid entering (a slow) regex
526 engine on strings that will not definitely match. If the C<isall> flag
527 is set, a call to the regex engine may be avoided even when the optimizer
528 found an appropriate place for the match.
530 Above the optimizer section is the list of I<nodes> of the compiled
531 form of the regex. Each line has format
533 C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
535 =head2 Types of Nodes
537 Here are the possible types, with short descriptions:
539 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
542 END no End of program.
543 SUCCEED no Return from a subroutine, basically.
547 BOL no Match "" at beginning of line.
548 MBOL no Same, assuming multiline.
549 SBOL no Same, assuming singleline.
550 EOS no Match "" at end of string.
551 EOL no Match "" at end of line.
552 MEOL no Same, assuming multiline.
553 SEOL no Same, assuming singleline.
554 BOUND no Match "" at any word boundary using native charset
555 semantics for non-utf8
556 BOUNDL no Match "" at any locale word boundary
557 BOUNDU no Match "" at any word boundary using Unicode semantics
558 BOUNDA no Match "" at any word boundary using ASCII semantics
559 NBOUND no Match "" at any word non-boundary using native charset
560 semantics for non-utf8
561 NBOUNDL no Match "" at any locale word non-boundary
562 NBOUNDU no Match "" at any word non-boundary using Unicode semantics
563 NBOUNDA no Match "" at any word non-boundary using ASCII semantics
564 GPOS no Matches where last m//g left off.
566 # [Special] alternatives:
568 REG_ANY no Match any one character (except newline).
569 SANY no Match any one character.
570 CANY no Match any one byte.
571 ANYOF sv Match character in (or not in) this class, single char
573 ANYOFV sv Match character in (or not in) this class, can
575 ALNUM no Match any alphanumeric character using native charset
576 semantics for non-utf8
577 ALNUML no Match any alphanumeric char in locale
578 ALNUMU no Match any alphanumeric char using Unicode semantics
579 ALNUMA no Match [A-Za-z_0-9]
580 NALNUM no Match any non-alphanumeric character using native charset
581 semantics for non-utf8
582 NALNUML no Match any non-alphanumeric char in locale
583 NALNUMU no Match any non-alphanumeric char using Unicode semantics
584 NALNUMA no Match [^A-Za-z_0-9]
585 SPACE no Match any whitespace character using native charset
586 semantics for non-utf8
587 SPACEL no Match any whitespace char in locale
588 SPACEU no Match any whitespace char using Unicode semantics
589 SPACEA no Match [ \t\n\f\r]
590 NSPACE no Match any non-whitespace character using native charset
591 semantics for non-utf8
592 NSPACEL no Match any non-whitespace char in locale
593 NSPACEU no Match any non-whitespace char using Unicode semantics
594 NSPACEA no Match [^ \t\n\f\r]
595 DIGIT no Match any numeric character using native charset semantics
597 DIGITL no Match any numeric character in locale
598 DIGITA no Match [0-9]
599 NDIGIT no Match any non-numeric character using native charset
600 i semantics for non-utf8
601 NDIGITL no Match any non-numeric character in locale
602 NDIGITA no Match [^0-9]
603 CLUMP no Match any extended grapheme cluster sequence
607 # BRANCH The set of branches constituting a single choice are hooked
608 # together with their "next" pointers, since precedence prevents
609 # anything being concatenated to any individual branch. The
610 # "next" pointer of the last BRANCH in a choice points to the
611 # thing following the whole choice. This is also where the
612 # final "next" pointer of each individual branch points; each
613 # branch starts with the operand node of a BRANCH node.
615 BRANCH node Match this alternative, or the next...
619 # BACK Normal "next" pointers all implicitly point forward; BACK
620 # exists to make loop structures possible.
622 BACK no Match "", "next" ptr points backward.
626 EXACT str Match this string (preceded by length).
627 EXACTF str Match this string, folded, native charset semantics for
628 non-utf8 (prec. by length).
629 EXACTFL str Match this string, folded in locale (w/len).
630 EXACTFU str Match this string, folded, Unicode semantics for non-utf8
632 EXACTFA str Match this string, folded, Unicode semantics for non-utf8,
633 but no ASCII-range character matches outside ASCII (prec.
638 NOTHING no Match empty string.
639 # A variant of above which delimits a group, thus stops optimizations
640 TAIL no Match empty string. Can jump here from outside.
644 # STAR,PLUS '?', and complex '*' and '+', are implemented as circular
645 # BRANCH structures using BACK. Simple cases (one character
646 # per match) are implemented with STAR and PLUS for speed
647 # and to minimize recursive plunges.
649 STAR node Match this (simple) thing 0 or more times.
650 PLUS node Match this (simple) thing 1 or more times.
652 CURLY sv 2 Match this simple thing {n,m} times.
653 CURLYN no 2 Capture next-after-this simple thing
654 CURLYM no 2 Capture this medium-complex thing {n,m} times.
655 CURLYX sv 2 Match this complex thing {n,m} times.
657 # This terminator creates a loop structure for CURLYX
658 WHILEM no Do curly processing and see if rest matches.
662 # OPEN,CLOSE,GROUPP ...are numbered at compile time.
663 OPEN num 1 Mark this point in input as start of #n.
664 CLOSE num 1 Analogous to OPEN.
666 REF num 1 Match some already matched string
667 REFF num 1 Match already matched string, folded using native charset
668 semantics for non-utf8
669 REFFL num 1 Match already matched string, folded in loc.
670 REFFU num 1 Match already matched string, folded using unicode
671 semantics for non-utf8
672 REFFA num 1 Match already matched string, folded using unicode
673 semantics for non-utf8, no mixing ASCII, non-ASCII
675 # Named references. Code in regcomp.c assumes that these all are after the
676 # numbered references
677 NREF no-sv 1 Match some already matched string
678 NREFF no-sv 1 Match already matched string, folded using native charset
679 semantics for non-utf8
680 NREFFL no-sv 1 Match already matched string, folded in loc.
681 NREFFU num 1 Match already matched string, folded using unicode
682 semantics for non-utf8
683 NREFFA num 1 Match already matched string, folded using unicode
684 semantics for non-utf8, no mixing ASCII, non-ASCII
686 IFMATCH off 1 2 Succeeds if the following matches.
687 UNLESSM off 1 2 Fails if the following matches.
688 SUSPEND off 1 1 "Independent" sub-RE.
689 IFTHEN off 1 1 Switch, should be preceded by switcher.
690 GROUPP num 1 Whether the group matched.
692 # Support for long RE
694 LONGJMP off 1 1 Jump far away.
695 BRANCHJ off 1 1 BRANCH with long offset.
699 EVAL evl 1 Execute some Perl code.
703 MINMOD no Next operator is not greedy.
704 LOGICAL no Next opcode should set the flag only.
706 # This is not used yet
707 RENUM off 1 1 Group with independently numbered parens.
711 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants have
712 # inline charclass data (ascii only), the 'C' store it in the structure.
713 # NOTE: the relative order of the TRIE-like regops is significant
715 TRIE trie 1 Match many EXACT(F[ALU]?)? at once. flags==type
716 TRIEC charclass Same as TRIE, but with embedded charclass data
718 # For start classes, contains an added fail table.
719 AHOCORASICK trie 1 Aho Corasick stclass. flags==type
720 AHOCORASICKC charclass Same as AHOCORASICK, but with embedded charclass data
723 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
724 GOSTART no recurse to start of pattern
726 # Special conditionals
727 NGROUPP no-sv 1 Whether the group matched.
728 INSUBP num 1 Whether we are in a specific recurse.
729 DEFINEP none 1 Never execute directly.
732 ENDLIKE none Used only for the type field of verbs
733 OPFAIL none Same as (?!)
734 ACCEPT parno 1 Accepts the current matched string.
737 # Verbs With Arguments
738 VERB no-sv 1 Used only for the type field of verbs
739 PRUNE no-sv 1 Pattern fails at this startpoint if no-backtracking through this
740 MARKPOINT no-sv 1 Push the current location for rollback by cut.
741 SKIP no-sv 1 On failure skip forward (to the mark) before retrying
742 COMMIT no-sv 1 Pattern fails outright if backtracking through this
743 CUTGROUP no-sv 1 On failure go to the next alternation in the group
745 # Control what to keep in $&.
746 KEEPS no $& begins here.
748 # New charclass like patterns
749 LNBREAK none generic newline pattern
750 VERTWS none vertical whitespace (Perl 6)
751 NVERTWS none not vertical whitespace (Perl 6)
752 HORIZWS none horizontal whitespace (Perl 6)
753 NHORIZWS none not horizontal whitespace (Perl 6)
755 FOLDCHAR codepoint 1 codepoint with tricky case folding properties.
759 # This is not really a node, but an optimized away piece of a "long" node.
760 # To simplify debugging output, we mark it as if it were a node
761 OPTIMIZED off Placeholder for dump.
763 # Special opcode with the property that no opcode in a compiled program
764 # will ever be of this type. Thus it can be used as a flag value that
765 # no other opcode has been seen. END is used similarly, in that an END
766 # node cant be optimized. So END implies "unoptimizable" and PSEUDO mean
767 # "not seen anything to optimize yet".
768 PSEUDO off Pseudo opcode for internal use.
770 =for unprinted-credits
771 Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
773 Following the optimizer information is a dump of the offset/length
774 table, here split across several lines:
777 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
778 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
779 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
780 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
782 The first line here indicates that the offset/length table contains 45
783 entries. Each entry is a pair of integers, denoted by C<offset[length]>.
784 Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
785 entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:>
786 (the C<1: ANYOF[bc]>) begins at character position 1 in the
787 pre-compiled form of the regex, and has a length of 4 characters.
788 C<5[1]> in position 12
789 indicates that the node labeled C<12:>
790 (the C<< 12: EXACT <d> >>) begins at character position 5 in the
791 pre-compiled form of the regex, and has a length of 1 character.
792 C<12[1]> in position 14
793 indicates that the node labeled C<14:>
794 (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
795 pre-compiled form of the regex, and has a length of 1 character---that
796 is, it corresponds to the C<+> symbol in the precompiled regex.
798 C<0[0]> items indicate that there is no corresponding node.
800 =head2 Run-time Output
802 First of all, when doing a match, one may get no run-time output even
803 if debugging is enabled. This means that the regex engine was never
804 entered and that all of the job was therefore done by the optimizer.
806 If the regex engine was entered, the output may look like this:
808 Matching `[bc]d(ef*g)+h[ij]k$' against `abcdefg__gh__'
809 Setting an EVAL scope, savestack=3
810 2 <ab> <cdefg__gh_> | 1: ANYOF
811 3 <abc> <defg__gh_> | 11: EXACT <d>
812 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
813 4 <abcd> <efg__gh_> | 26: WHILEM
814 0 out of 1..32767 cc=effff31c
815 4 <abcd> <efg__gh_> | 15: OPEN1
816 4 <abcd> <efg__gh_> | 17: EXACT <e>
817 5 <abcde> <fg__gh_> | 19: STAR
818 EXACT <f> can match 1 times out of 32767...
819 Setting an EVAL scope, savestack=3
820 6 <bcdef> <g__gh__> | 22: EXACT <g>
821 7 <bcdefg> <__gh__> | 24: CLOSE1
822 7 <bcdefg> <__gh__> | 26: WHILEM
823 1 out of 1..32767 cc=effff31c
824 Setting an EVAL scope, savestack=12
825 7 <bcdefg> <__gh__> | 15: OPEN1
826 7 <bcdefg> <__gh__> | 17: EXACT <e>
827 restoring \1 to 4(4)..7
828 failed, try continuation...
829 7 <bcdefg> <__gh__> | 27: NOTHING
830 7 <bcdefg> <__gh__> | 28: EXACT <h>
834 The most significant information in the output is about the particular I<node>
835 of the compiled regex that is currently being tested against the target string.
836 The format of these lines is
838 C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE>
840 The I<TYPE> info is indented with respect to the backtracking level.
841 Other incidental information appears interspersed within.
843 =head1 Debugging Perl Memory Usage
845 Perl is a profligate wastrel when it comes to memory use. There
846 is a saying that to estimate memory usage of Perl, assume a reasonable
847 algorithm for memory allocation, multiply that estimate by 10, and
848 while you still may miss the mark, at least you won't be quite so
849 astonished. This is not absolutely true, but may provide a good
850 grasp of what happens.
852 Assume that an integer cannot take less than 20 bytes of memory, a
853 float cannot take less than 24 bytes, a string cannot take less
854 than 32 bytes (all these examples assume 32-bit architectures, the
855 result are quite a bit worse on 64-bit architectures). If a variable
856 is accessed in two of three different ways (which require an integer,
857 a float, or a string), the memory footprint may increase yet another
858 20 bytes. A sloppy malloc(3) implementation can inflate these
859 numbers dramatically.
861 On the opposite end of the scale, a declaration like
865 may take up to 500 bytes of memory, depending on which release of Perl
868 Anecdotal estimates of source-to-compiled code bloat suggest an
869 eightfold increase. This means that the compiled form of reasonable
870 (normally commented, properly indented etc.) code will take
871 about eight times more space in memory than the code took
874 The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
875 (it was available only if Perl was built with C<-DDEBUGGING>).
876 The switch was used to track Perl's memory allocations and possible
877 memory leaks. These days the use of malloc debugging tools like
878 F<Purify> or F<valgrind> is suggested instead. See also
879 L<perlhacktips/PERL_MEM_LOG>.
881 One way to find out how much memory is being used by Perl data
882 structures is to install the Devel::Size module from CPAN: it gives
883 you the minimum number of bytes required to store a particular data
884 structure. Please be mindful of the difference between the size()
887 If Perl has been compiled using Perl's malloc you can analyze Perl
888 memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
890 =head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
892 If your perl is using Perl's malloc() and was compiled with the
893 necessary switches (this is the default), then it will print memory
894 usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
895 > 1 >>, and before termination of the program when C<<
896 $ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to
897 the following example:
899 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
900 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
901 14216 free: 130 117 28 7 9 0 2 2 1 0 0
903 60924 used: 125 137 161 55 7 8 6 16 2 0 1
905 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
906 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
907 30888 free: 245 78 85 13 6 2 1 3 2 0 1
909 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
911 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
913 It is possible to ask for such a statistic at arbitrary points in
914 your execution using the mstat() function out of the standard
917 Here is some explanation of that format:
921 =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
923 Perl's malloc() uses bucketed allocations. Every request is rounded
924 up to the closest bucket size available, and a bucket is taken from
925 the pool of buckets of that size.
927 The line above describes the limits of buckets currently in use.
928 Each bucket has two sizes: memory footprint and the maximal size
929 of user data that can fit into this bucket. Suppose in the above
930 example that the smallest bucket were size 4. The biggest bucket
931 would have usable size 8188, and the memory footprint would be 8192.
933 In a Perl built for debugging, some buckets may have negative usable
934 size. This means that these buckets cannot (and will not) be used.
935 For larger buckets, the memory footprint may be one page greater
936 than a power of 2. If so, the corresponding power of two is
937 printed in the C<APPROX> field above.
941 The 1 or 2 rows of numbers following that correspond to the number
942 of buckets of each size between C<SMALLEST> and C<GREATEST>. In
943 the first row, the sizes (memory footprints) of buckets are powers
944 of two--or possibly one page greater. In the second row, if present,
945 the memory footprints of the buckets are between the memory footprints
946 of two buckets "above".
948 For example, suppose under the previous example, the memory footprints
951 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
954 With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
955 a 4-byte overhead, and thus an 8192-long bucket may take up to
956 8188-byte allocations.
958 =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
960 The first two fields give the total amount of memory perl sbrk(2)ed
961 (ess-broken? :-) and number of sbrk(2)s used. The third number is
962 what perl thinks about continuity of returned chunks. So long as
963 this number is positive, malloc() will assume that it is probable
964 that sbrk(2) will provide continuous memory.
966 Memory allocated by external libraries is not counted.
970 The amount of sbrk(2)ed memory needed to keep buckets aligned.
974 Although memory overhead of bigger buckets is kept inside the bucket, for
975 smaller buckets, it is kept in separate areas. This field gives the
976 total size of these areas.
980 malloc() may want to subdivide a bigger bucket into smaller buckets.
981 If only a part of the deceased bucket is left unsubdivided, the rest
982 is kept as an element of a linked list. This field gives the total
983 size of these chunks.
987 To minimize the number of sbrk(2)s, malloc() asks for more memory. This
988 field gives the size of the yet unused part, which is sbrk(2)ed, but