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> set to identify
98 the called subroutine. (This doesn't happen if the calling subroutine
99 was compiled in the C<DB> package.) C<$DB::sub> normally holds the name
100 of the called subroutine, if it has a name by which it can be looked up.
101 Failing that, C<$DB::sub> will hold a reference to the called subroutine.
102 Either way, the C<&DB::sub> subroutine can use C<$DB::sub> as a reference
103 by which to call the called subroutine, which it will normally want to do.
105 X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub>
106 is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling
107 back to C<&DB::sub>(I<args>).
111 When execution of the program uses C<goto> to enter a non-XS subroutine
112 and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto> is made, with
113 C<$DB::sub> set to identify the subroutine being entered. The call to
114 C<&DB::goto> does not replace the C<goto>; the requested subroutine will
115 still be entered once C<&DB::goto> has returned. C<$DB::sub> normally
116 holds the name of the subroutine being entered, if it has one. Failing
117 that, C<$DB::sub> will hold a reference to the subroutine being entered.
118 Unlike when C<&DB::sub> is called, it is not guaranteed that C<$DB::sub>
119 can be used as a reference to operate on the subroutine being entered.
123 Note that if C<&DB::sub> needs external data for it to work, no
124 subroutine call is possible without it. As an example, the standard
125 debugger's C<&DB::sub> depends on the C<$DB::deep> variable
126 (it defines how many levels of recursion deep into the debugger you can go
127 before a mandatory break). If C<$DB::deep> is not defined, subroutine
128 calls are not possible, even though C<&DB::sub> exists.
130 =head2 Writing Your Own Debugger
132 =head3 Environment Variables
134 The C<PERL5DB> environment variable can be used to define a debugger.
135 For example, the minimal "working" debugger (it actually doesn't do anything)
136 consists of one line:
140 It can easily be defined like this:
142 $ PERL5DB="sub DB::DB {}" perl -d your-script
144 Another brief debugger, slightly more useful, can be created
147 sub DB::DB {print ++$i; scalar <STDIN>}
149 This debugger prints a number which increments for each statement
150 encountered and waits for you to hit a newline before continuing
151 to the next statement.
153 The following debugger is actually useful:
158 sub sub {print ++$i, " $sub\n"; &$sub}
161 It prints the sequence number of each subroutine call and the name of the
162 called subroutine. Note that C<&DB::sub> is being compiled into the
163 package C<DB> through the use of the C<package> directive.
165 When it starts, the debugger reads your rc file (F<./.perldb> or
166 F<~/.perldb> under Unix), which can set important options.
167 (A subroutine (C<&afterinit>) can be defined here as well; it is executed
168 after the debugger completes its own initialization.)
170 After the rc file is read, the debugger reads the PERLDB_OPTS
171 environment variable and uses it to set debugger options. The
172 contents of this variable are treated as if they were the argument
173 of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
175 =head3 Debugger Internal Variables
177 In addition to the file and subroutine-related variables mentioned above,
178 the debugger also maintains various magical internal variables.
184 C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
185 holds the lines of the currently-selected file (compiled by Perl), either
186 explicitly chosen with the debugger's C<f> command, or implicitly by flow
189 Values in this array are magical in numeric context: they compare
190 equal to zero only if the line is not breakable.
194 C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
195 contains breakpoints and actions keyed by line number in
196 the currently-selected file, either explicitly chosen with the
197 debugger's C<f> command, or implicitly by flow of execution.
199 As previously noted, individual entries (as opposed to the whole hash)
200 are settable. Perl only cares about Boolean true here, although
201 the values used by F<perl5db.pl> have the form
202 C<"$break_condition\0$action">.
206 =head3 Debugger Customization Functions
208 Some functions are provided to simplify customization.
214 See L<perldebug/"Configurable Options"> for a description of options parsed by
215 C<DB::parse_options(string)>.
219 C<DB::dump_trace(skip[,count])> skips the specified number of frames
220 and returns a list containing information about the calling frames (all
221 of them, if C<count> is missing). Each entry is reference to a hash
222 with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
223 name, or info about C<eval>), C<args> (C<undef> or a reference to
224 an array), C<file>, and C<line>.
228 C<DB::print_trace(FH, skip[, count[, short]])> prints
229 formatted info about caller frames. The last two functions may be
230 convenient as arguments to C<< < >>, C<< << >> commands.
234 Note that any variables and functions that are not documented in
235 this manpages (or in L<perldebug>) are considered for internal
236 use only, and as such are subject to change without notice.
238 =head1 Frame Listing Output Examples
240 The C<frame> option can be used to control the output of frame
241 information. For example, contrast this expression trace:
244 Stack dump during die enabled outside of evals.
246 Loading DB routines from perl5db.pl patch level 0.94
247 Emacs support available.
249 Enter h or 'h h' for help.
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):
262 with this one, once the C<o>ption C<frame=2> has been set:
266 DB<5> t print foo() * bar()
276 By way of demonstration, we present below a laborious listing
277 resulting from setting your C<PERLDB_OPTS> environment variable to
278 the value C<f=n N>, and running I<perl -d -V> from the command line.
279 Examples using various values of C<n> are shown to give you a feel
280 for the difference between settings. Long though it may be, this
281 is not a complete listing, but only excerpts.
288 entering Config::BEGIN
289 Package lib/Exporter.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
304 entering Config::BEGIN
305 Package lib/Exporter.pm.
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
316 entering Config::myconfig
317 entering Config::FETCH
319 entering Config::FETCH
321 entering Config::FETCH
325 in $=main::BEGIN() from /dev/null:0
326 in $=Config::BEGIN() from lib/Config.pm:2
327 Package lib/Exporter.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
343 in $=main::BEGIN() from /dev/null:0
344 in $=Config::BEGIN() from lib/Config.pm:2
345 Package lib/Exporter.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
367 in $=main::BEGIN() from /dev/null:0
368 in $=Config::BEGIN() from lib/Config.pm:2
369 Package lib/Exporter.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
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 $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
402 scalar context return from Exporter::import: ''
406 In all cases shown above, the line indentation shows the call tree.
407 If bit 2 of C<frame> is set, a line is printed on exit from a
408 subroutine as well. If bit 4 is set, the arguments are printed
409 along with the caller info. If bit 8 is set, the arguments are
410 printed even if they are tied or references. If bit 16 is set, the
411 return value is printed, too.
413 When a package is compiled, a line like this
417 is printed with proper indentation.
419 =head1 Debugging Regular Expressions
421 There are two ways to enable debugging output for regular expressions.
423 If your perl is compiled with C<-DDEBUGGING>, you may use the
424 B<-Dr> flag on the command line, and C<-Drv> for more verbose
427 Otherwise, one can C<use re 'debug'>, which has effects at both
428 compile time and run time. Since Perl 5.9.5, this pragma is lexically
431 =head2 Compile-time Output
433 The debugging output at compile time looks like this:
435 Compiling REx '[bc]d(ef*g)+h[ij]k$'
436 size 45 Got 364 bytes for offset annotations.
442 14: CURLYX[0] {1,32767}(28)
456 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
457 stclass 'ANYOF[bc]' minlen 7
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.
465 The first line shows the pre-compiled form of the regex. The second
466 shows the size of the compiled form (in arbitrary units, usually
467 4-byte words) and the total number of bytes allocated for the
468 offset/length table, usually 4+C<size>*8. The next line shows the
469 label I<id> of the first node that does a match.
473 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
474 stclass 'ANYOF[bc]' minlen 7
476 line (split into two lines above) contains optimizer
477 information. In the example shown, the optimizer found that the match
478 should contain a substring C<de> at offset 1, plus substring C<gh>
479 at some offset between 3 and infinity. Moreover, when checking for
480 these substrings (to abandon impossible matches quickly), Perl will check
481 for the substring C<gh> before checking for the substring C<de>. The
482 optimizer may also use the knowledge that the match starts (at the
483 C<first> I<id>) with a character class, and no string
484 shorter than 7 characters can possibly match.
486 The fields of interest which may appear in this line are
490 =item C<anchored> I<STRING> C<at> I<POS>
492 =item C<floating> I<STRING> C<at> I<POS1..POS2>
496 =item C<matching floating/anchored>
498 Which substring to check first.
502 The minimal length of the match.
504 =item C<stclass> I<TYPE>
506 Type of first matching node.
510 Don't scan for the found substrings.
514 Means that the optimizer information is all that the regular
515 expression contains, and thus one does not need to enter the regex engine at
520 Set if the pattern contains C<\G>.
524 Set if the pattern starts with a repeated char (as in C<x+y>).
528 Set if the pattern starts with C<.*>.
532 Set if the pattern contain eval-groups, such as C<(?{ code })> and
535 =item C<anchored(TYPE)>
537 If the pattern may match only at a handful of places, with C<TYPE>
538 being C<SBOL>, C<MBOL>, or C<GPOS>. See the table below.
542 If a substring is known to match at end-of-line only, it may be
543 followed by C<$>, as in C<floating 'k'$>.
545 The optimizer-specific information is used to avoid entering (a slow) regex
546 engine on strings that will not definitely match. If the C<isall> flag
547 is set, a call to the regex engine may be avoided even when the optimizer
548 found an appropriate place for the match.
550 Above the optimizer section is the list of I<nodes> of the compiled
551 form of the regex. Each line has format
553 C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
555 =head2 Types of Nodes
557 Here are the current possible types, with short descriptions:
560 This table is generated by regen/regcomp.pl. Any changes made here
563 =for regcomp.pl begin
565 # TYPE arg-description [regnode-struct-suffix] [longjump-len] DESCRIPTION
569 END no End of program.
570 SUCCEED no Return from a subroutine, basically.
572 # Line Start Anchors:
573 SBOL no Match "" at beginning of line: /^/, /\A/
574 MBOL no Same, assuming multiline: /^/m
577 SEOL no Match "" at end of line: /$/
578 MEOL no Same, assuming multiline: /$/m
579 EOS no Match "" at end of string: /\z/
581 # Match Start Anchors:
582 GPOS no Matches where last m//g left off.
584 # Word Boundary Opcodes:
585 BOUND no Like BOUNDA for non-utf8, otherwise like
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
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
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
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
609 ANYOFL sv Like ANYOF, but /l is in effect
611 ANYOFPOSIXL sv Like ANYOFL, but matches [[:posix:]]
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
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
624 ANYOFHs sv 1 Like ANYOFHb, but has a string field that
625 gives the leading matchable UTF-8 bytes;
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
632 ANYOFRb packed 1 Like ANYOFR, but all matches share the same
633 UTF-8 start byte, given in the flags field
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
639 # POSIX Character Classes:
640 POSIXD none Some [[:class:]] under /d; the FLAGS field
642 POSIXL none Some [[:class:]] under /l; the FLAGS field
644 POSIXU none Some [[:class:]] under /u; the FLAGS field
646 POSIXA none Some [[:class:]] under /a; the FLAGS field
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:]]
653 CLUMP no Match any extended grapheme cluster
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.
667 BRANCH node Match this alternative, or the next...
671 EXACT str Match this string (flags field is the
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;
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
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;
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)
692 EXACTFUP str Like EXACT, but match using /iu rules;
693 (string not UTF-8, folded except MICRO,
694 SHARP S: hence Problematic)
696 EXACTFLU8 str Like EXACTFU, but use /il, UTF-8, (string
697 is folded, and everything in it is above
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)
703 EXACT_REQ8 str Like EXACT, but only UTF-8 encoded targets
705 LEXACT_REQ8 len:str 1 Like LEXACT, but only UTF-8 encoded targets
707 EXACTFU_REQ8 str Like EXACTFU, but only UTF-8 encoded
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)
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
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.
728 STAR node Match this (simple) thing 0 or more times.
729 PLUS node Match this (simple) thing 1 or more times.
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}
735 CURLYX sv 2 Match this complex thing {n,m} times.
737 # This terminator creates a loop structure for CURLYX
738 WHILEM no Do curly processing and see if rest
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
749 REF num 1 Match some already matched string
750 REFF num 1 Match already matched string, using /di
752 REFFL num 1 Match already matched string, using /li
754 REFFU num 1 Match already matched string, usng /ui.
755 REFFA num 1 Match already matched string, using /aai
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
763 REFFLN no-sv 1 Match already matched string, using /li
765 REFFUN num 1 Match already matched string, using /ui
767 REFFAN num 1 Match already matched string, using /aai
770 # Support for long RE
771 LONGJMP off 1 1 Jump far away.
772 BRANCHJ off 1 1 BRANCH with long offset.
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
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
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.
789 EVAL evl/flags Execute some Perl code.
794 MINMOD no Next operator is not greedy.
795 LOGICAL no Next opcode should set the flag only.
797 # This is not used yet
798 RENUM off 1 1 Group with independently numbered parens.
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
806 TRIE trie 1 Match many EXACT(F[ALU]?)? at once.
808 TRIEC trie Same as TRIE, but with embedded charclass
811 AHOCORASICK trie 1 Aho Corasick stclass. flags==type
812 AHOCORASICKC trie Same as AHOCORASICK, but with embedded
813 charclass charclass data
816 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
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.
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
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
835 SKIP no-sv 1 On failure skip forward (to the mark)
837 COMMIT no-sv 1 Pattern fails outright if backtracking
839 CUTGROUP no-sv 1 On failure go to the next alternation in
842 # Control what to keep in $&.
843 KEEPS no $& begins here.
845 # New charclass like patterns
846 LNBREAK none generic newline pattern
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.
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.
863 =for unprinted-credits
864 Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
866 Following the optimizer information is a dump of the offset/length
867 table, here split across several lines:
870 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
871 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
872 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
873 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
875 The first line here indicates that the offset/length table contains 45
876 entries. Each entry is a pair of integers, denoted by C<offset[length]>.
877 Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
878 entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:>
879 (the C<1: ANYOF[bc]>) begins at character position 1 in the
880 pre-compiled form of the regex, and has a length of 4 characters.
881 C<5[1]> in position 12
882 indicates that the node labeled C<12:>
883 (the C<< 12: EXACT <d> >>) begins at character position 5 in the
884 pre-compiled form of the regex, and has a length of 1 character.
885 C<12[1]> in position 14
886 indicates that the node labeled C<14:>
887 (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
888 pre-compiled form of the regex, and has a length of 1 character---that
889 is, it corresponds to the C<+> symbol in the precompiled regex.
891 C<0[0]> items indicate that there is no corresponding node.
893 =head2 Run-time Output
895 First of all, when doing a match, one may get no run-time output even
896 if debugging is enabled. This means that the regex engine was never
897 entered and that all of the job was therefore done by the optimizer.
899 If the regex engine was entered, the output may look like this:
901 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
902 Setting an EVAL scope, savestack=3
903 2 <ab> <cdefg__gh_> | 1: ANYOF
904 3 <abc> <defg__gh_> | 11: EXACT <d>
905 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
906 4 <abcd> <efg__gh_> | 26: WHILEM
907 0 out of 1..32767 cc=effff31c
908 4 <abcd> <efg__gh_> | 15: OPEN1
909 4 <abcd> <efg__gh_> | 17: EXACT <e>
910 5 <abcde> <fg__gh_> | 19: STAR
911 EXACT <f> can match 1 times out of 32767...
912 Setting an EVAL scope, savestack=3
913 6 <bcdef> <g__gh__> | 22: EXACT <g>
914 7 <bcdefg> <__gh__> | 24: CLOSE1
915 7 <bcdefg> <__gh__> | 26: WHILEM
916 1 out of 1..32767 cc=effff31c
917 Setting an EVAL scope, savestack=12
918 7 <bcdefg> <__gh__> | 15: OPEN1
919 7 <bcdefg> <__gh__> | 17: EXACT <e>
920 restoring \1 to 4(4)..7
921 failed, try continuation...
922 7 <bcdefg> <__gh__> | 27: NOTHING
923 7 <bcdefg> <__gh__> | 28: EXACT <h>
927 The most significant information in the output is about the particular I<node>
928 of the compiled regex that is currently being tested against the target string.
929 The format of these lines is
931 C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE>
933 The I<TYPE> info is indented with respect to the backtracking level.
934 Other incidental information appears interspersed within.
936 =head1 Debugging Perl Memory Usage
938 Perl is a profligate wastrel when it comes to memory use. There
939 is a saying that to estimate memory usage of Perl, assume a reasonable
940 algorithm for memory allocation, multiply that estimate by 10, and
941 while you still may miss the mark, at least you won't be quite so
942 astonished. This is not absolutely true, but may provide a good
943 grasp of what happens.
945 Assume that an integer cannot take less than 20 bytes of memory, a
946 float cannot take less than 24 bytes, a string cannot take less
947 than 32 bytes (all these examples assume 32-bit architectures, the
948 result are quite a bit worse on 64-bit architectures). If a variable
949 is accessed in two of three different ways (which require an integer,
950 a float, or a string), the memory footprint may increase yet another
951 20 bytes. A sloppy malloc(3) implementation can inflate these
952 numbers dramatically.
954 On the opposite end of the scale, a declaration like
958 may take up to 500 bytes of memory, depending on which release of Perl
961 Anecdotal estimates of source-to-compiled code bloat suggest an
962 eightfold increase. This means that the compiled form of reasonable
963 (normally commented, properly indented etc.) code will take
964 about eight times more space in memory than the code took
967 The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
968 (it was available only if Perl was built with C<-DDEBUGGING>).
969 The switch was used to track Perl's memory allocations and possible
970 memory leaks. These days the use of malloc debugging tools like
971 F<Purify> or F<valgrind> is suggested instead. See also
972 L<perlhacktips/PERL_MEM_LOG>.
974 One way to find out how much memory is being used by Perl data
975 structures is to install the Devel::Size module from CPAN: it gives
976 you the minimum number of bytes required to store a particular data
977 structure. Please be mindful of the difference between the size()
980 If Perl has been compiled using Perl's malloc you can analyze Perl
981 memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
983 =head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
985 If your perl is using Perl's malloc() and was compiled with the
986 necessary switches (this is the default), then it will print memory
987 usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
988 > 1 >>, and before termination of the program when C<<
989 $ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to
990 the following example:
992 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
993 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
994 14216 free: 130 117 28 7 9 0 2 2 1 0 0
996 60924 used: 125 137 161 55 7 8 6 16 2 0 1
998 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
999 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
1000 30888 free: 245 78 85 13 6 2 1 3 2 0 1
1002 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
1004 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
1006 It is possible to ask for such a statistic at arbitrary points in
1007 your execution using the mstat() function out of the standard
1010 Here is some explanation of that format:
1014 =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
1016 Perl's malloc() uses bucketed allocations. Every request is rounded
1017 up to the closest bucket size available, and a bucket is taken from
1018 the pool of buckets of that size.
1020 The line above describes the limits of buckets currently in use.
1021 Each bucket has two sizes: memory footprint and the maximal size
1022 of user data that can fit into this bucket. Suppose in the above
1023 example that the smallest bucket were size 4. The biggest bucket
1024 would have usable size 8188, and the memory footprint would be 8192.
1026 In a Perl built for debugging, some buckets may have negative usable
1027 size. This means that these buckets cannot (and will not) be used.
1028 For larger buckets, the memory footprint may be one page greater
1029 than a power of 2. If so, the corresponding power of two is
1030 printed in the C<APPROX> field above.
1034 The 1 or 2 rows of numbers following that correspond to the number
1035 of buckets of each size between C<SMALLEST> and C<GREATEST>. In
1036 the first row, the sizes (memory footprints) of buckets are powers
1037 of two--or possibly one page greater. In the second row, if present,
1038 the memory footprints of the buckets are between the memory footprints
1039 of two buckets "above".
1041 For example, suppose under the previous example, the memory footprints
1044 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
1047 With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
1048 a 4-byte overhead, and thus an 8192-long bucket may take up to
1049 8188-byte allocations.
1051 =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
1053 The first two fields give the total amount of memory perl sbrk(2)ed
1054 (ess-broken? :-) and number of sbrk(2)s used. The third number is
1055 what perl thinks about continuity of returned chunks. So long as
1056 this number is positive, malloc() will assume that it is probable
1057 that sbrk(2) will provide continuous memory.
1059 Memory allocated by external libraries is not counted.
1063 The amount of sbrk(2)ed memory needed to keep buckets aligned.
1065 =item C<heads: 2192>
1067 Although memory overhead of bigger buckets is kept inside the bucket, for
1068 smaller buckets, it is kept in separate areas. This field gives the
1069 total size of these areas.
1073 malloc() may want to subdivide a bigger bucket into smaller buckets.
1074 If only a part of the deceased bucket is left unsubdivided, the rest
1075 is kept as an element of a linked list. This field gives the total
1076 size of these chunks.
1080 To minimize the number of sbrk(2)s, malloc() asks for more memory. This
1081 field gives the size of the yet unused part, which is sbrk(2)ed, but