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 $filename for C<eval>ed strings looks like C<(eval 34)>.
42 Values in this array are magical in numeric context: they compare
43 equal to zero only if the line is not breakable.
47 Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed
48 by line number. Individual entries (as opposed to the whole hash)
49 are settable. Perl only cares about Boolean true here, although
50 the values used by F<perl5db.pl> have the form
51 C<"$break_condition\0$action">.
53 The same holds for evaluated strings that contain subroutines, or
54 which are currently being executed. The $filename for C<eval>ed strings
55 looks like C<(eval 34)>.
59 Each scalar C<${"_<$filename"}> contains C<"_<$filename">. This is
60 also the case for evaluated strings that contain subroutines, or
61 which are currently being executed. The $filename for C<eval>ed
62 strings looks like C<(eval 34)>.
66 After each C<require>d file is compiled, but before it is executed,
67 C<DB::postponed(*{"_<$filename"})> is called if the subroutine
68 C<DB::postponed> exists. Here, the $filename is the expanded name of
69 the C<require>d file, as found in the values of %INC.
73 After each subroutine C<subname> is compiled, the existence of
74 C<$DB::postponed{subname}> is checked. If this key exists,
75 C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine
80 A hash C<%DB::sub> is maintained, whose keys are subroutine names
81 and whose values have the form C<filename:startline-endline>.
82 C<filename> has the form C<(eval 34)> for subroutines defined inside
87 When the execution of your program reaches a point that can hold a
88 breakpoint, the C<DB::DB()> subroutine is called if any of the variables
89 C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true. These variables
90 are not C<local>izable. This feature is disabled when executing
91 inside C<DB::DB()>, including functions called from it
92 unless C<< $^D & (1<<30) >> is true.
96 When execution of the program reaches a subroutine call, a call to
97 C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> holding the
98 name of the called subroutine. (This doesn't happen if the subroutine
99 was compiled in the C<DB> package.)
101 X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub>
102 is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling
103 back to C<&DB::sub>(I<args>).
107 When execution of the program uses C<goto> to enter a non-XS
108 subroutine and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto>
109 is made, with C<$DB::sub> holding the name of the subroutine being
114 Note that if C<&DB::sub> needs external data for it to work, no
115 subroutine call is possible without it. As an example, the standard
116 debugger's C<&DB::sub> depends on the C<$DB::deep> variable
117 (it defines how many levels of recursion deep into the debugger you can go
118 before a mandatory break). If C<$DB::deep> is not defined, subroutine
119 calls are not possible, even though C<&DB::sub> exists.
121 =head2 Writing Your Own Debugger
123 =head3 Environment Variables
125 The C<PERL5DB> environment variable can be used to define a debugger.
126 For example, the minimal "working" debugger (it actually doesn't do anything)
127 consists of one line:
131 It can easily be defined like this:
133 $ PERL5DB="sub DB::DB {}" perl -d your-script
135 Another brief debugger, slightly more useful, can be created
138 sub DB::DB {print ++$i; scalar <STDIN>}
140 This debugger prints a number which increments for each statement
141 encountered and waits for you to hit a newline before continuing
142 to the next statement.
144 The following debugger is actually useful:
149 sub sub {print ++$i, " $sub\n"; &$sub}
152 It prints the sequence number of each subroutine call and the name of the
153 called subroutine. Note that C<&DB::sub> is being compiled into the
154 package C<DB> through the use of the C<package> directive.
156 When it starts, the debugger reads your rc file (F<./.perldb> or
157 F<~/.perldb> under Unix), which can set important options.
158 (A subroutine (C<&afterinit>) can be defined here as well; it is executed
159 after the debugger completes its own initialization.)
161 After the rc file is read, the debugger reads the PERLDB_OPTS
162 environment variable and uses it to set debugger options. The
163 contents of this variable are treated as if they were the argument
164 of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
166 =head3 Debugger Internal Variables
168 In addition to the file and subroutine-related variables mentioned above,
169 the debugger also maintains various magical internal variables.
175 C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
176 holds the lines of the currently-selected file (compiled by Perl), either
177 explicitly chosen with the debugger's C<f> command, or implicitly by flow
180 Values in this array are magical in numeric context: they compare
181 equal to zero only if the line is not breakable.
185 C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
186 contains breakpoints and actions keyed by line number in
187 the currently-selected file, either explicitly chosen with the
188 debugger's C<f> command, or implicitly by flow of execution.
190 As previously noted, individual entries (as opposed to the whole hash)
191 are settable. Perl only cares about Boolean true here, although
192 the values used by F<perl5db.pl> have the form
193 C<"$break_condition\0$action">.
197 =head3 Debugger Customization Functions
199 Some functions are provided to simplify customization.
205 See L<perldebug/"Configurable Options"> for a description of options parsed by
206 C<DB::parse_options(string)>.
210 C<DB::dump_trace(skip[,count])> skips the specified number of frames
211 and returns a list containing information about the calling frames (all
212 of them, if C<count> is missing). Each entry is reference to a hash
213 with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
214 name, or info about C<eval>), C<args> (C<undef> or a reference to
215 an array), C<file>, and C<line>.
219 C<DB::print_trace(FH, skip[, count[, short]])> prints
220 formatted info about caller frames. The last two functions may be
221 convenient as arguments to C<< < >>, C<< << >> commands.
225 Note that any variables and functions that are not documented in
226 this manpages (or in L<perldebug>) are considered for internal
227 use only, and as such are subject to change without notice.
229 =head1 Frame Listing Output Examples
231 The C<frame> option can be used to control the output of frame
232 information. For example, contrast this expression trace:
235 Stack dump during die enabled outside of evals.
237 Loading DB routines from perl5db.pl patch level 0.94
238 Emacs support available.
240 Enter h or 'h h' for help.
247 DB<3> t print foo() * bar()
248 main::((eval 172):3): print foo() + bar();
249 main::foo((eval 168):2):
250 main::bar((eval 170):2):
253 with this one, once the C<o>ption C<frame=2> has been set:
257 DB<5> t print foo() * bar()
267 By way of demonstration, we present below a laborious listing
268 resulting from setting your C<PERLDB_OPTS> environment variable to
269 the value C<f=n N>, and running I<perl -d -V> from the command line.
270 Examples using various values of C<n> are shown to give you a feel
271 for the difference between settings. Long though it may be, this
272 is not a complete listing, but only excerpts.
279 entering Config::BEGIN
280 Package lib/Exporter.pm.
282 Package lib/Config.pm.
283 entering Config::TIEHASH
284 entering Exporter::import
285 entering Exporter::export
286 entering Config::myconfig
287 entering Config::FETCH
288 entering Config::FETCH
289 entering Config::FETCH
290 entering Config::FETCH
295 entering Config::BEGIN
296 Package lib/Exporter.pm.
299 Package lib/Config.pm.
300 entering Config::TIEHASH
301 exited Config::TIEHASH
302 entering Exporter::import
303 entering Exporter::export
304 exited Exporter::export
305 exited Exporter::import
307 entering Config::myconfig
308 entering Config::FETCH
310 entering Config::FETCH
312 entering Config::FETCH
316 in $=main::BEGIN() from /dev/null:0
317 in $=Config::BEGIN() from lib/Config.pm:2
318 Package lib/Exporter.pm.
320 Package lib/Config.pm.
321 in $=Config::TIEHASH('Config') from lib/Config.pm:644
322 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
323 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
324 in @=Config::myconfig() from /dev/null:0
325 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
326 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
327 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
328 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
329 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
330 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
334 in $=main::BEGIN() from /dev/null:0
335 in $=Config::BEGIN() from lib/Config.pm:2
336 Package lib/Exporter.pm.
338 out $=Config::BEGIN() from lib/Config.pm:0
339 Package lib/Config.pm.
340 in $=Config::TIEHASH('Config') from lib/Config.pm:644
341 out $=Config::TIEHASH('Config') from lib/Config.pm:644
342 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
343 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
344 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
345 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
346 out $=main::BEGIN() from /dev/null:0
347 in @=Config::myconfig() from /dev/null:0
348 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
349 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
350 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
351 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
352 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
353 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
354 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
358 in $=main::BEGIN() from /dev/null:0
359 in $=Config::BEGIN() from lib/Config.pm:2
360 Package lib/Exporter.pm.
362 out $=Config::BEGIN() from lib/Config.pm:0
363 Package lib/Config.pm.
364 in $=Config::TIEHASH('Config') from lib/Config.pm:644
365 out $=Config::TIEHASH('Config') from lib/Config.pm:644
366 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
367 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
368 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
369 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
370 out $=main::BEGIN() from /dev/null:0
371 in @=Config::myconfig() from /dev/null:0
372 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
373 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
374 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
375 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
379 in $=CODE(0x15eca4)() from /dev/null:0
380 in $=CODE(0x182528)() from lib/Config.pm:2
381 Package lib/Exporter.pm.
382 out $=CODE(0x182528)() from lib/Config.pm:0
383 scalar context return from CODE(0x182528): undef
384 Package lib/Config.pm.
385 in $=Config::TIEHASH('Config') from lib/Config.pm:628
386 out $=Config::TIEHASH('Config') from lib/Config.pm:628
387 scalar context return from Config::TIEHASH: empty hash
388 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
389 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
390 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
391 scalar context return from Exporter::export: ''
392 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
393 scalar context return from Exporter::import: ''
397 In all cases shown above, the line indentation shows the call tree.
398 If bit 2 of C<frame> is set, a line is printed on exit from a
399 subroutine as well. If bit 4 is set, the arguments are printed
400 along with the caller info. If bit 8 is set, the arguments are
401 printed even if they are tied or references. If bit 16 is set, the
402 return value is printed, too.
404 When a package is compiled, a line like this
408 is printed with proper indentation.
410 =head1 Debugging Regular Expressions
412 There are two ways to enable debugging output for regular expressions.
414 If your perl is compiled with C<-DDEBUGGING>, you may use the
415 B<-Dr> flag on the command line.
417 Otherwise, one can C<use re 'debug'>, which has effects at
418 compile time and run time. Since Perl 5.9.5, this pragma is lexically
421 =head2 Compile-time Output
423 The debugging output at compile time looks like this:
425 Compiling REx '[bc]d(ef*g)+h[ij]k$'
426 size 45 Got 364 bytes for offset annotations.
432 14: CURLYX[0] {1,32767}(28)
446 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
447 stclass 'ANYOF[bc]' minlen 7
449 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
450 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
451 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
452 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
453 Omitting $` $& $' support.
455 The first line shows the pre-compiled form of the regex. The second
456 shows the size of the compiled form (in arbitrary units, usually
457 4-byte words) and the total number of bytes allocated for the
458 offset/length table, usually 4+C<size>*8. The next line shows the
459 label I<id> of the first node that does a match.
463 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
464 stclass 'ANYOF[bc]' minlen 7
466 line (split into two lines above) contains optimizer
467 information. In the example shown, the optimizer found that the match
468 should contain a substring C<de> at offset 1, plus substring C<gh>
469 at some offset between 3 and infinity. Moreover, when checking for
470 these substrings (to abandon impossible matches quickly), Perl will check
471 for the substring C<gh> before checking for the substring C<de>. The
472 optimizer may also use the knowledge that the match starts (at the
473 C<first> I<id>) with a character class, and no string
474 shorter than 7 characters can possibly match.
476 The fields of interest which may appear in this line are
480 =item C<anchored> I<STRING> C<at> I<POS>
482 =item C<floating> I<STRING> C<at> I<POS1..POS2>
486 =item C<matching floating/anchored>
488 Which substring to check first.
492 The minimal length of the match.
494 =item C<stclass> I<TYPE>
496 Type of first matching node.
500 Don't scan for the found substrings.
504 Means that the optimizer information is all that the regular
505 expression contains, and thus one does not need to enter the regex engine at
510 Set if the pattern contains C<\G>.
514 Set if the pattern starts with a repeated char (as in C<x+y>).
518 Set if the pattern starts with C<.*>.
522 Set if the pattern contain eval-groups, such as C<(?{ code })> and
525 =item C<anchored(TYPE)>
527 If the pattern may match only at a handful of places, with C<TYPE>
528 being C<SBOL>, C<MBOL>, or C<GPOS>. See the table below.
532 If a substring is known to match at end-of-line only, it may be
533 followed by C<$>, as in C<floating 'k'$>.
535 The optimizer-specific information is used to avoid entering (a slow) regex
536 engine on strings that will not definitely match. If the C<isall> flag
537 is set, a call to the regex engine may be avoided even when the optimizer
538 found an appropriate place for the match.
540 Above the optimizer section is the list of I<nodes> of the compiled
541 form of the regex. Each line has format
543 C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
545 =head2 Types of Nodes
547 Here are the current possible types, with short descriptions:
550 This table is generated by regen/regcomp.pl. Any changes made here
553 =for regcomp.pl begin
555 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
559 END no End of program.
560 SUCCEED no Return from a subroutine, basically.
562 # Line Start Anchors:
563 SBOL no Match "" at beginning of line: /^/, /\A/
564 MBOL no Same, assuming multiline: /^/m
567 SEOL no Match "" at end of line: /$/
568 MEOL no Same, assuming multiline: /$/m
569 EOS no Match "" at end of string: /\z/
571 # Match Start Anchors:
572 GPOS no Matches where last m//g left off.
574 # Word Boundary Opcodes:
575 BOUND no Like BOUNDA for non-utf8, otherwise match ""
576 between any Unicode \w\W or \W\w
577 BOUNDL no Like BOUND/BOUNDU, but \w and \W are defined
579 BOUNDU no Match "" at any boundary of a given type
581 BOUNDA no Match "" at any boundary between \w\W or
582 \W\w, where \w is [_a-zA-Z0-9]
583 NBOUND no Like NBOUNDA for non-utf8, otherwise match
584 "" between any Unicode \w\w or \W\W
585 NBOUNDL no Like NBOUND/NBOUNDU, but \w and \W are
586 defined by current locale
587 NBOUNDU no Match "" at any non-boundary of a given type
588 using using Unicode rules
589 NBOUNDA no Match "" betweeen any \w\w or \W\W, where \w
592 # [Special] alternatives:
593 REG_ANY no Match any one character (except newline).
594 SANY no Match any one character.
595 ANYOF sv 1 Match character in (or not in) this class,
596 single char match only
597 ANYOFD sv 1 Like ANYOF, but /d is in effect
598 ANYOFL sv 1 Like ANYOF, but /l is in effect
600 # POSIX Character Classes:
601 POSIXD none Some [[:class:]] under /d; the FLAGS field
603 POSIXL none Some [[:class:]] under /l; the FLAGS field
605 POSIXU none Some [[:class:]] under /u; the FLAGS field
607 POSIXA none Some [[:class:]] under /a; the FLAGS field
609 NPOSIXD none complement of POSIXD, [[:^class:]]
610 NPOSIXL none complement of POSIXL, [[:^class:]]
611 NPOSIXU none complement of POSIXU, [[:^class:]]
612 NPOSIXA none complement of POSIXA, [[:^class:]]
614 CLUMP no Match any extended grapheme cluster sequence
618 # BRANCH The set of branches constituting a single choice are
619 # hooked together with their "next" pointers, since
620 # precedence prevents anything being concatenated to
621 # any individual branch. The "next" pointer of the last
622 # BRANCH in a choice points to the thing following the
623 # whole choice. This is also where the final "next"
624 # pointer of each individual branch points; each branch
625 # starts with the operand node of a BRANCH node.
627 BRANCH node Match this alternative, or the next...
631 EXACT str Match this string (preceded by length).
632 EXACTL str Like EXACT, but /l is in effect (used so
633 locale-related warnings can be checked for).
634 EXACTF str Match this non-UTF-8 string (not guaranteed
635 to be folded) using /id rules (w/len).
636 EXACTFL str Match this string (not guaranteed to be
637 folded) using /il rules (w/len).
638 EXACTFU str Match this string (folded iff in UTF-8,
639 length in folding doesn't change if not in
640 UTF-8) using /iu rules (w/len).
641 EXACTFA str Match this string (not guaranteed to be
642 folded) using /iaa rules (w/len).
644 EXACTFU_SS str Match this string (folded iff in UTF-8,
645 length in folding may change even if not in
646 UTF-8) using /iu rules (w/len).
647 EXACTFLU8 str Rare cirucmstances: like EXACTFU, but is
648 under /l, UTF-8, folded, and everything in
650 EXACTFA_NO_TRIE str Match this string (which is not trie-able;
651 not guaranteed to be folded) using /iaa
656 NOTHING no Match empty string.
657 # A variant of above which delimits a group, thus stops optimizations
658 TAIL no Match empty string. Can jump here from
663 # STAR,PLUS '?', and complex '*' and '+', are implemented as
664 # circular BRANCH structures. Simple cases
665 # (one character per match) are implemented with STAR
666 # and PLUS for speed and to minimize recursive plunges.
668 STAR node Match this (simple) thing 0 or more times.
669 PLUS node Match this (simple) thing 1 or more times.
671 CURLY sv 2 Match this simple thing {n,m} times.
672 CURLYN no 2 Capture next-after-this simple thing
673 CURLYM no 2 Capture this medium-complex thing {n,m}
675 CURLYX sv 2 Match this complex thing {n,m} times.
677 # This terminator creates a loop structure for CURLYX
678 WHILEM no Do curly processing and see if rest matches.
682 # OPEN,CLOSE,GROUPP ...are numbered at compile time.
683 OPEN num 1 Mark this point in input as start of #n.
684 CLOSE num 1 Analogous to OPEN.
686 REF num 1 Match some already matched string
687 REFF num 1 Match already matched string, folded using
688 native charset rules for non-utf8
689 REFFL num 1 Match already matched string, folded in loc.
690 REFFU num 1 Match already matched string, folded using
691 unicode rules for non-utf8
692 REFFA num 1 Match already matched string, folded using
693 unicode rules for non-utf8, no mixing ASCII,
696 # Named references. Code in regcomp.c assumes that these all are after
697 # the numbered references
698 NREF no-sv 1 Match some already matched string
699 NREFF no-sv 1 Match already matched string, folded using
700 native charset rules for non-utf8
701 NREFFL no-sv 1 Match already matched string, folded in loc.
702 NREFFU num 1 Match already matched string, folded using
703 unicode rules for non-utf8
704 NREFFA num 1 Match already matched string, folded using
705 unicode rules for non-utf8, no mixing ASCII,
708 # Support for long RE
709 LONGJMP off 1 1 Jump far away.
710 BRANCHJ off 1 1 BRANCH with long offset.
712 # Special Case Regops
713 IFMATCH off 1 2 Succeeds if the following matches.
714 UNLESSM off 1 2 Fails if the following matches.
715 SUSPEND off 1 1 "Independent" sub-RE.
716 IFTHEN off 1 1 Switch, should be preceded by switcher.
717 GROUPP num 1 Whether the group matched.
721 EVAL evl/flags Execute some Perl code.
726 MINMOD no Next operator is not greedy.
727 LOGICAL no Next opcode should set the flag only.
729 # This is not used yet
730 RENUM off 1 1 Group with independently numbered parens.
734 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
735 # have inline charclass data (ascii only), the 'C' store it in the
738 TRIE trie 1 Match many EXACT(F[ALU]?)? at once.
740 TRIEC trie Same as TRIE, but with embedded charclass
743 AHOCORASICK trie 1 Aho Corasick stclass. flags==type
744 AHOCORASICKC trie Same as AHOCORASICK, but with embedded
745 charclass charclass data
748 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
750 # Special conditionals
751 NGROUPP no-sv 1 Whether the group matched.
752 INSUBP num 1 Whether we are in a specific recurse.
753 DEFINEP none 1 Never execute directly.
756 ENDLIKE none Used only for the type field of verbs
757 OPFAIL no-sv 1 Same as (?!), but with verb arg
758 ACCEPT no-sv/num Accepts the current matched string, with
761 # Verbs With Arguments
762 VERB no-sv 1 Used only for the type field of verbs
763 PRUNE no-sv 1 Pattern fails at this startpoint if no-
764 backtracking through this
765 MARKPOINT no-sv 1 Push the current location for rollback by
767 SKIP no-sv 1 On failure skip forward (to the mark) before
769 COMMIT no-sv 1 Pattern fails outright if backtracking
771 CUTGROUP no-sv 1 On failure go to the next alternation in the
774 # Control what to keep in $&.
775 KEEPS no $& begins here.
777 # New charclass like patterns
778 LNBREAK none generic newline pattern
782 # This is not really a node, but an optimized away piece of a "long"
783 # node. To simplify debugging output, we mark it as if it were a node
784 OPTIMIZED off Placeholder for dump.
786 # Special opcode with the property that no opcode in a compiled program
787 # will ever be of this type. Thus it can be used as a flag value that
788 # no other opcode has been seen. END is used similarly, in that an END
789 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
790 # mean "not seen anything to optimize yet".
791 PSEUDO off Pseudo opcode for internal use.
795 =for unprinted-credits
796 Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
798 Following the optimizer information is a dump of the offset/length
799 table, here split across several lines:
802 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
803 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
804 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
805 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
807 The first line here indicates that the offset/length table contains 45
808 entries. Each entry is a pair of integers, denoted by C<offset[length]>.
809 Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
810 entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:>
811 (the C<1: ANYOF[bc]>) begins at character position 1 in the
812 pre-compiled form of the regex, and has a length of 4 characters.
813 C<5[1]> in position 12
814 indicates that the node labeled C<12:>
815 (the C<< 12: EXACT <d> >>) begins at character position 5 in the
816 pre-compiled form of the regex, and has a length of 1 character.
817 C<12[1]> in position 14
818 indicates that the node labeled C<14:>
819 (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
820 pre-compiled form of the regex, and has a length of 1 character---that
821 is, it corresponds to the C<+> symbol in the precompiled regex.
823 C<0[0]> items indicate that there is no corresponding node.
825 =head2 Run-time Output
827 First of all, when doing a match, one may get no run-time output even
828 if debugging is enabled. This means that the regex engine was never
829 entered and that all of the job was therefore done by the optimizer.
831 If the regex engine was entered, the output may look like this:
833 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
834 Setting an EVAL scope, savestack=3
835 2 <ab> <cdefg__gh_> | 1: ANYOF
836 3 <abc> <defg__gh_> | 11: EXACT <d>
837 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
838 4 <abcd> <efg__gh_> | 26: WHILEM
839 0 out of 1..32767 cc=effff31c
840 4 <abcd> <efg__gh_> | 15: OPEN1
841 4 <abcd> <efg__gh_> | 17: EXACT <e>
842 5 <abcde> <fg__gh_> | 19: STAR
843 EXACT <f> can match 1 times out of 32767...
844 Setting an EVAL scope, savestack=3
845 6 <bcdef> <g__gh__> | 22: EXACT <g>
846 7 <bcdefg> <__gh__> | 24: CLOSE1
847 7 <bcdefg> <__gh__> | 26: WHILEM
848 1 out of 1..32767 cc=effff31c
849 Setting an EVAL scope, savestack=12
850 7 <bcdefg> <__gh__> | 15: OPEN1
851 7 <bcdefg> <__gh__> | 17: EXACT <e>
852 restoring \1 to 4(4)..7
853 failed, try continuation...
854 7 <bcdefg> <__gh__> | 27: NOTHING
855 7 <bcdefg> <__gh__> | 28: EXACT <h>
859 The most significant information in the output is about the particular I<node>
860 of the compiled regex that is currently being tested against the target string.
861 The format of these lines is
863 C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE>
865 The I<TYPE> info is indented with respect to the backtracking level.
866 Other incidental information appears interspersed within.
868 =head1 Debugging Perl Memory Usage
870 Perl is a profligate wastrel when it comes to memory use. There
871 is a saying that to estimate memory usage of Perl, assume a reasonable
872 algorithm for memory allocation, multiply that estimate by 10, and
873 while you still may miss the mark, at least you won't be quite so
874 astonished. This is not absolutely true, but may provide a good
875 grasp of what happens.
877 Assume that an integer cannot take less than 20 bytes of memory, a
878 float cannot take less than 24 bytes, a string cannot take less
879 than 32 bytes (all these examples assume 32-bit architectures, the
880 result are quite a bit worse on 64-bit architectures). If a variable
881 is accessed in two of three different ways (which require an integer,
882 a float, or a string), the memory footprint may increase yet another
883 20 bytes. A sloppy malloc(3) implementation can inflate these
884 numbers dramatically.
886 On the opposite end of the scale, a declaration like
890 may take up to 500 bytes of memory, depending on which release of Perl
893 Anecdotal estimates of source-to-compiled code bloat suggest an
894 eightfold increase. This means that the compiled form of reasonable
895 (normally commented, properly indented etc.) code will take
896 about eight times more space in memory than the code took
899 The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
900 (it was available only if Perl was built with C<-DDEBUGGING>).
901 The switch was used to track Perl's memory allocations and possible
902 memory leaks. These days the use of malloc debugging tools like
903 F<Purify> or F<valgrind> is suggested instead. See also
904 L<perlhacktips/PERL_MEM_LOG>.
906 One way to find out how much memory is being used by Perl data
907 structures is to install the Devel::Size module from CPAN: it gives
908 you the minimum number of bytes required to store a particular data
909 structure. Please be mindful of the difference between the size()
912 If Perl has been compiled using Perl's malloc you can analyze Perl
913 memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
915 =head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
917 If your perl is using Perl's malloc() and was compiled with the
918 necessary switches (this is the default), then it will print memory
919 usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
920 > 1 >>, and before termination of the program when C<<
921 $ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to
922 the following example:
924 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
925 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
926 14216 free: 130 117 28 7 9 0 2 2 1 0 0
928 60924 used: 125 137 161 55 7 8 6 16 2 0 1
930 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
931 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
932 30888 free: 245 78 85 13 6 2 1 3 2 0 1
934 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
936 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
938 It is possible to ask for such a statistic at arbitrary points in
939 your execution using the mstat() function out of the standard
942 Here is some explanation of that format:
946 =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
948 Perl's malloc() uses bucketed allocations. Every request is rounded
949 up to the closest bucket size available, and a bucket is taken from
950 the pool of buckets of that size.
952 The line above describes the limits of buckets currently in use.
953 Each bucket has two sizes: memory footprint and the maximal size
954 of user data that can fit into this bucket. Suppose in the above
955 example that the smallest bucket were size 4. The biggest bucket
956 would have usable size 8188, and the memory footprint would be 8192.
958 In a Perl built for debugging, some buckets may have negative usable
959 size. This means that these buckets cannot (and will not) be used.
960 For larger buckets, the memory footprint may be one page greater
961 than a power of 2. If so, the corresponding power of two is
962 printed in the C<APPROX> field above.
966 The 1 or 2 rows of numbers following that correspond to the number
967 of buckets of each size between C<SMALLEST> and C<GREATEST>. In
968 the first row, the sizes (memory footprints) of buckets are powers
969 of two--or possibly one page greater. In the second row, if present,
970 the memory footprints of the buckets are between the memory footprints
971 of two buckets "above".
973 For example, suppose under the previous example, the memory footprints
976 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
979 With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
980 a 4-byte overhead, and thus an 8192-long bucket may take up to
981 8188-byte allocations.
983 =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
985 The first two fields give the total amount of memory perl sbrk(2)ed
986 (ess-broken? :-) and number of sbrk(2)s used. The third number is
987 what perl thinks about continuity of returned chunks. So long as
988 this number is positive, malloc() will assume that it is probable
989 that sbrk(2) will provide continuous memory.
991 Memory allocated by external libraries is not counted.
995 The amount of sbrk(2)ed memory needed to keep buckets aligned.
999 Although memory overhead of bigger buckets is kept inside the bucket, for
1000 smaller buckets, it is kept in separate areas. This field gives the
1001 total size of these areas.
1005 malloc() may want to subdivide a bigger bucket into smaller buckets.
1006 If only a part of the deceased bucket is left unsubdivided, the rest
1007 is kept as an element of a linked list. This field gives the total
1008 size of these chunks.
1012 To minimize the number of sbrk(2)s, malloc() asks for more memory. This
1013 field gives the size of the yet unused part, which is sbrk(2)ed, but