1PERLDEBGUTS(1) Perl Programmers Reference Guide PERLDEBGUTS(1)
2
6 perldebguts - Guts of Perl debugging
7
9 This is not perldebug, which tells you how to use the debugger. This
10 manpage describes low-level details concerning the debugger's
11 internals, which range from difficult to impossible to understand for
12 anyone who isn't incredibly intimate with Perl's guts. Caveat lector.
13
15 Perl has special debugging hooks at compile-time and run-time used to
16 create debugging environments. These hooks are not to be confused with
17 the perl -Dxxx command described in perlrun, which is usable only if a
18 special Perl is built per the instructions in the INSTALL file in the
19 Perl source tree.
20 For example, whenever you call Perl's built-in "caller" function from
22 the package "DB", the arguments that the corresponding stack frame was
23 called with are copied to the @DB::args array. These mechanisms are
24 enabled by calling Perl with the -d switch. Specifically, the
25 following additional features are enabled (cf. "$^P" in perlvar):
26 • Perl inserts the contents of $ENV{PERL5DB} (or "BEGIN {require
28 'perl5db.pl'}" if not present) before the first line of your
29 program.
30 • Each array "@{"_<$filename"}" holds the lines of $filename for a
32 file compiled by Perl. The same is also true for "eval"ed strings
33 that contain subroutines, or which are currently being executed.
34 The $filename for "eval"ed strings looks like "(eval 34)".
35 Values in this array are magical in numeric context: they compare
37 equal to zero only if the line is not breakable.
38 • Each hash "%{"_<$filename"}" contains breakpoints and actions keyed
40 by line number. Individual entries (as opposed to the whole hash)
41 are settable. Perl only cares about Boolean true here, although
42 the values used by perl5db.pl have the form
43 "$break_condition\0$action".
44 The same holds for evaluated strings that contain subroutines, or
46 which are currently being executed. The $filename for "eval"ed
47 strings looks like "(eval 34)".
48 • Each scalar "${"_<$filename"}" contains $filename. This is also
50 the case for evaluated strings that contain subroutines, or which
51 are currently being executed. The $filename for "eval"ed strings
52 looks like "(eval 34)".
53 • After each "require"d file is compiled, but before it is executed,
55 DB::postponed(*{"_<$filename"}) is called if the subroutine
56 "DB::postponed" exists. Here, the $filename is the expanded name
57 of the "require"d file, as found in the values of %INC.
58 • After each subroutine "subname" is compiled, the existence of
60 $DB::postponed{subname} is checked. If this key exists,
61 DB::postponed(subname) is called if the "DB::postponed" subroutine
62 also exists.
63 • A hash %DB::sub is maintained, whose keys are subroutine names and
65 whose values have the form "filename:startline-endline".
66 "filename" has the form "(eval 34)" for subroutines defined inside
67 "eval"s.
68 • When the execution of your program reaches a point that can hold a
70 breakpoint, the DB::DB() subroutine is called if any of the
71 variables $DB::trace, $DB::single, or $DB::signal is true. These
72 variables are not "local"izable. This feature is disabled when
73 executing inside DB::DB(), including functions called from it
74 unless "$^D & (1<<30)" is true.
75 • When execution of the program reaches a subroutine call, a call to
77 &DB::sub(args) is made instead, with $DB::sub set to identify the
78 called subroutine. (This doesn't happen if the calling subroutine
79 was compiled in the "DB" package.) $DB::sub normally holds the
80 name of the called subroutine, if it has a name by which it can be
81 looked up. Failing that, $DB::sub will hold a reference to the
82 called subroutine. Either way, the &DB::sub subroutine can use
83 $DB::sub as a reference by which to call the called subroutine,
84 which it will normally want to do.
85 If the call is to an lvalue subroutine, and &DB::lsub is defined
87 &DB::lsub(args) is called instead, otherwise falling back to
88 &DB::sub(args).
89 • When execution of the program uses "goto" to enter a non-XS
91 subroutine and the 0x80 bit is set in $^P, a call to &DB::goto is
92 made, with $DB::sub set to identify the subroutine being entered.
93 The call to &DB::goto does not replace the "goto"; the requested
94 subroutine will still be entered once &DB::goto has returned.
95 $DB::sub normally holds the name of the subroutine being entered,
96 if it has one. Failing that, $DB::sub will hold a reference to the
97 subroutine being entered. Unlike when &DB::sub is called, it is
98 not guaranteed that $DB::sub can be used as a reference to operate
99 on the subroutine being entered.
100 Note that if &DB::sub needs external data for it to work, no subroutine
102 call is possible without it. As an example, the standard debugger's
103 &DB::sub depends on the $DB::deep variable (it defines how many levels
104 of recursion deep into the debugger you can go before a mandatory
105 break). If $DB::deep is not defined, subroutine calls are not
106 possible, even though &DB::sub exists.
107 Writing Your Own Debugger
109 Environment Variables
110 The "PERL5DB" environment variable can be used to define a debugger.
112 For example, the minimal "working" debugger (it actually doesn't do
113 anything) consists of one line:
114 sub DB::DB {}
116 It can easily be defined like this:
118 $ PERL5DB="sub DB::DB {}" perl -d your-script
120 Another brief debugger, slightly more useful, can be created with only
122 the line:
123 sub DB::DB {print ++$i; scalar <STDIN>}
125 This debugger prints a number which increments for each statement
127 encountered and waits for you to hit a newline before continuing to the
128 next statement.
129 The following debugger is actually useful:
131 {
133 package DB;
134 sub DB {}
135 sub sub {print ++$i, " $sub\n"; &$sub}
136 }
137 It prints the sequence number of each subroutine call and the name of
139 the called subroutine. Note that &DB::sub is being compiled into the
140 package "DB" through the use of the "package" directive.
141 When it starts, the debugger reads your rc file (./.perldb or ~/.perldb
143 under Unix), which can set important options. (A subroutine
144 (&afterinit) can be defined here as well; it is executed after the
145 debugger completes its own initialization.)
146 After the rc file is read, the debugger reads the PERLDB_OPTS
148 environment variable and uses it to set debugger options. The contents
149 of this variable are treated as if they were the argument of an "o ..."
150 debugger command (q.v. in "Configurable Options" in perldebug).
151 Debugger Internal Variables
153 In addition to the file and subroutine-related variables mentioned
155 above, the debugger also maintains various magical internal variables.
156 • @DB::dbline is an alias for "@{"::_<current_file"}", which holds
158 the lines of the currently-selected file (compiled by Perl), either
159 explicitly chosen with the debugger's "f" command, or implicitly by
160 flow of execution.
161 Values in this array are magical in numeric context: they compare
163 equal to zero only if the line is not breakable.
164 • %DB::dbline is an alias for "%{"::_<current_file"}", which contains
166 breakpoints and actions keyed by line number in the currently-
167 selected file, either explicitly chosen with the debugger's "f"
168 command, or implicitly by flow of execution.
169 As previously noted, individual entries (as opposed to the whole
171 hash) are settable. Perl only cares about Boolean true here,
172 although the values used by perl5db.pl have the form
173 "$break_condition\0$action".
174 Debugger Customization Functions
176 Some functions are provided to simplify customization.
178 • See "Configurable Options" in perldebug for a description of
180 options parsed by DB::parse_options(string).
181 • "DB::dump_trace(skip[,count])" skips the specified number of frames
183 and returns a list containing information about the calling frames
184 (all of them, if "count" is missing). Each entry is reference to a
185 hash with keys "context" (either ".", "$", or "@"), "sub"
186 (subroutine name, or info about "eval"), "args" ("undef" or a
187 reference to an array), "file", and "line".
188 • "DB::print_trace(FH, skip[, count[, short]])" prints formatted info
190 about caller frames. The last two functions may be convenient as
191 arguments to "<", "<<" commands.
192 Note that any variables and functions that are not documented in this
194 manpages (or in perldebug) are considered for internal use only, and as
195 such are subject to change without notice.
196
198 The "frame" option can be used to control the output of frame
199 information. For example, contrast this expression trace:
200 $ perl -de 42
202 Stack dump during die enabled outside of evals.
203 Loading DB routines from perl5db.pl patch level 0.94
205 Emacs support available.
206 Enter h or 'h h' for help.
208 main::(-e:1): 0
210 DB<1> sub foo { 14 }
211 DB<2> sub bar { 3 }
213 DB<3> t print foo() * bar()
215 main::((eval 172):3): print foo() + bar();
216 main::foo((eval 168):2):
217 main::bar((eval 170):2):
218 42
219 with this one, once the "o"ption "frame=2" has been set:
221 DB<4> o f=2
223 frame = '2'
224 DB<5> t print foo() * bar()
225 3: foo() * bar()
226 entering main::foo
227 2: sub foo { 14 };
228 exited main::foo
229 entering main::bar
230 2: sub bar { 3 };
231 exited main::bar
232 42
233 By way of demonstration, we present below a laborious listing resulting
235 from setting your "PERLDB_OPTS" environment variable to the value "f=n
236 N", and running perl -d -V from the command line. Examples using
237 various values of "n" are shown to give you a feel for the difference
238 between settings. Long though it may be, this is not a complete
239 listing, but only excerpts.
240 1.
242 entering main::BEGIN
243 entering Config::BEGIN
244 Package lib/Exporter.pm.
245 Package lib/Carp.pm.
246 Package lib/Config.pm.
247 entering Config::TIEHASH
248 entering Exporter::import
249 entering Exporter::export
250 entering Config::myconfig
251 entering Config::FETCH
252 entering Config::FETCH
253 entering Config::FETCH
254 entering Config::FETCH
255 2.
257 entering main::BEGIN
258 entering Config::BEGIN
259 Package lib/Exporter.pm.
260 Package lib/Carp.pm.
261 exited Config::BEGIN
262 Package lib/Config.pm.
263 entering Config::TIEHASH
264 exited Config::TIEHASH
265 entering Exporter::import
266 entering Exporter::export
267 exited Exporter::export
268 exited Exporter::import
269 exited main::BEGIN
270 entering Config::myconfig
271 entering Config::FETCH
272 exited Config::FETCH
273 entering Config::FETCH
274 exited Config::FETCH
275 entering Config::FETCH
276 3.
278 in $=main::BEGIN() from /dev/null:0
279 in $=Config::BEGIN() from lib/Config.pm:2
280 Package lib/Exporter.pm.
281 Package lib/Carp.pm.
282 Package lib/Config.pm.
283 in $=Config::TIEHASH('Config') from lib/Config.pm:644
284 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
285 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
286 in @=Config::myconfig() from /dev/null:0
287 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
288 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
289 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
290 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
291 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
292 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
293 4.
295 in $=main::BEGIN() from /dev/null:0
296 in $=Config::BEGIN() from lib/Config.pm:2
297 Package lib/Exporter.pm.
298 Package lib/Carp.pm.
299 out $=Config::BEGIN() from lib/Config.pm:0
300 Package lib/Config.pm.
301 in $=Config::TIEHASH('Config') from lib/Config.pm:644
302 out $=Config::TIEHASH('Config') from lib/Config.pm:644
303 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
304 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
305 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
306 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
307 out $=main::BEGIN() from /dev/null:0
308 in @=Config::myconfig() from /dev/null:0
309 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
310 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
311 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
312 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
313 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
314 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
315 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
316 5.
318 in $=main::BEGIN() from /dev/null:0
319 in $=Config::BEGIN() from lib/Config.pm:2
320 Package lib/Exporter.pm.
321 Package lib/Carp.pm.
322 out $=Config::BEGIN() from lib/Config.pm:0
323 Package lib/Config.pm.
324 in $=Config::TIEHASH('Config') from lib/Config.pm:644
325 out $=Config::TIEHASH('Config') from lib/Config.pm:644
326 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
327 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
328 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
329 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
330 out $=main::BEGIN() from /dev/null:0
331 in @=Config::myconfig() from /dev/null:0
332 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
333 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
334 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
335 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
336 6.
338 in $=CODE(0x15eca4)() from /dev/null:0
339 in $=CODE(0x182528)() from lib/Config.pm:2
340 Package lib/Exporter.pm.
341 out $=CODE(0x182528)() from lib/Config.pm:0
342 scalar context return from CODE(0x182528): undef
343 Package lib/Config.pm.
344 in $=Config::TIEHASH('Config') from lib/Config.pm:628
345 out $=Config::TIEHASH('Config') from lib/Config.pm:628
346 scalar context return from Config::TIEHASH: empty hash
347 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
348 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
349 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
350 scalar context return from Exporter::export: ''
351 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
352 scalar context return from Exporter::import: ''
353 In all cases shown above, the line indentation shows the call tree. If
355 bit 2 of "frame" is set, a line is printed on exit from a subroutine as
356 well. If bit 4 is set, the arguments are printed along with the caller
357 info. If bit 8 is set, the arguments are printed even if they are tied
358 or references. If bit 16 is set, the return value is printed, too.
359 When a package is compiled, a line like this
361 Package lib/Carp.pm.
363 is printed with proper indentation.
365
367 There are two ways to enable debugging output for regular expressions.
368 If your perl is compiled with "-DDEBUGGING", you may use the -Dr flag
370 on the command line, and "-Drv" for more verbose information.
371 Otherwise, one can "use re 'debug'", which has effects at both compile
373 time and run time. Since Perl 5.9.5, this pragma is lexically scoped.
374 Compile-time Output
376 The debugging output at compile time looks like this:
377 Compiling REx '[bc]d(ef*g)+h[ij]k$'
379 size 45 Got 364 bytes for offset annotations.
380 first at 1
381 rarest char g at 0
382 rarest char d at 0
383 1: ANYOF[bc](12)
384 12: EXACT <d>(14)
385 14: CURLYX[0] {1,32767}(28)
386 16: OPEN1(18)
387 18: EXACT <e>(20)
388 20: STAR(23)
389 21: EXACT <f>(0)
390 23: EXACT <g>(25)
391 25: CLOSE1(27)
392 27: WHILEM[1/1](0)
393 28: NOTHING(29)
394 29: EXACT <h>(31)
395 31: ANYOF[ij](42)
396 42: EXACT <k>(44)
397 44: EOL(45)
398 45: END(0)
399 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
400 stclass 'ANYOF[bc]' minlen 7
401 Offsets: [45]
402 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
403 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
404 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
405 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
406 Omitting $` $& $' support.
407 The first line shows the pre-compiled form of the regex. The second
409 shows the size of the compiled form (in arbitrary units, usually 4-byte
410 words) and the total number of bytes allocated for the offset/length
411 table, usually 4+"size"*8. The next line shows the label id of the
412 first node that does a match.
413 The
415 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
417 stclass 'ANYOF[bc]' minlen 7
418 line (split into two lines above) contains optimizer information. In
420 the example shown, the optimizer found that the match should contain a
421 substring "de" at offset 1, plus substring "gh" at some offset between
422 3 and infinity. Moreover, when checking for these substrings (to
423 abandon impossible matches quickly), Perl will check for the substring
424 "gh" before checking for the substring "de". The optimizer may also
425 use the knowledge that the match starts (at the "first" id) with a
426 character class, and no string shorter than 7 characters can possibly
427 match.
428 The fields of interest which may appear in this line are
430 "anchored" STRING "at" POS
432 "floating" STRING "at" POS1..POS2
433 See above.
434 "matching floating/anchored"
436 Which substring to check first.
437 "minlen"
439 The minimal length of the match.
440 "stclass" TYPE
442 Type of first matching node.
443 "noscan"
445 Don't scan for the found substrings.
446 "isall"
448 Means that the optimizer information is all that the regular
449 expression contains, and thus one does not need to enter the regex
450 engine at all.
451 "GPOS"
453 Set if the pattern contains "\G".
454 "plus"
456 Set if the pattern starts with a repeated char (as in "x+y").
457 "implicit"
459 Set if the pattern starts with ".*".
460 "with eval"
462 Set if the pattern contain eval-groups, such as "(?{ code })" and
463 "(??{ code })".
464 anchored(TYPE)
466 If the pattern may match only at a handful of places, with "TYPE"
467 being "SBOL", "MBOL", or "GPOS". See the table below.
468 If a substring is known to match at end-of-line only, it may be
470 followed by "$", as in "floating 'k'$".
471 The optimizer-specific information is used to avoid entering (a slow)
473 regex engine on strings that will not definitely match. If the "isall"
474 flag is set, a call to the regex engine may be avoided even when the
475 optimizer found an appropriate place for the match.
476 Above the optimizer section is the list of nodes of the compiled form
478 of the regex. Each line has format
479 " "id: TYPE OPTIONAL-INFO (next-id)
481 Types of Nodes
483 Here are the current possible types, with short descriptions:
484 # TYPE arg-description [regnode-struct-suffix] [longjump-len] DESCRIPTION
486 # Exit points
488 END no End of program.
490 SUCCEED no Return from a subroutine, basically.
491 # Line Start Anchors:
493 SBOL no Match "" at beginning of line: /^/, /\A/
494 MBOL no Same, assuming multiline: /^/m
495 # Line End Anchors:
497 SEOL no Match "" at end of line: /$/
498 MEOL no Same, assuming multiline: /$/m
499 EOS no Match "" at end of string: /\z/
500 # Match Start Anchors:
502 GPOS no Matches where last m//g left off.
503 # Word Boundary Opcodes:
505 BOUND no Like BOUNDA for non-utf8, otherwise like
506 BOUNDU
507 BOUNDL no Like BOUND/BOUNDU, but \w and \W are
508 defined by current locale
509 BOUNDU no Match "" at any boundary of a given type
510 using /u rules.
511 BOUNDA no Match "" at any boundary between \w\W or
512 \W\w, where \w is [_a-zA-Z0-9]
513 NBOUND no Like NBOUNDA for non-utf8, otherwise like
514 BOUNDU
515 NBOUNDL no Like NBOUND/NBOUNDU, but \w and \W are
516 defined by current locale
517 NBOUNDU no Match "" at any non-boundary of a given
518 type using using /u rules.
519 NBOUNDA no Match "" betweeen any \w\w or \W\W, where
520 \w is [_a-zA-Z0-9]
521 # [Special] alternatives:
523 REG_ANY no Match any one character (except newline).
524 SANY no Match any one character.
525 ANYOF sv Match character in (or not in) this class,
526 charclass single char match only
527 ANYOFD sv Like ANYOF, but /d is in effect
528 charclass
529 ANYOFL sv Like ANYOF, but /l is in effect
530 charclass
531 ANYOFPOSIXL sv Like ANYOFL, but matches [[:posix:]]
532 charclass_ classes
533 posixl
534 ANYOFH sv 1 Like ANYOF, but only has "High" matches,
536 none in the bitmap; the flags field
537 contains the lowest matchable UTF-8 start
538 byte
539 ANYOFHb sv 1 Like ANYOFH, but all matches share the same
540 UTF-8 start byte, given in the flags field
541 ANYOFHr sv 1 Like ANYOFH, but the flags field contains
542 packed bounds for all matchable UTF-8 start
543 bytes.
544 ANYOFHs sv:str 1 Like ANYOFHb, but has a string field that
545 gives the leading matchable UTF-8 bytes;
546 flags field is len
547 ANYOFR packed 1 Matches any character in the range given by
548 its packed args: upper 12 bits is the max
549 delta from the base lower 20; the flags
550 field contains the lowest matchable UTF-8
551 start byte
552 ANYOFRb packed 1 Like ANYOFR, but all matches share the same
553 UTF-8 start byte, given in the flags field
554 ANYOFHbbm none bbm Like ANYOFHb, but only for 2-byte UTF-8
556 characters; uses a bitmap to match the
557 continuation byte
558 ANYOFM byte 1 Like ANYOF, but matches an invariant byte
560 as determined by the mask and arg
561 NANYOFM byte 1 complement of ANYOFM
562 # POSIX Character Classes:
564 POSIXD none Some [[:class:]] under /d; the FLAGS field
565 gives which one
566 POSIXL none Some [[:class:]] under /l; the FLAGS field
567 gives which one
568 POSIXU none Some [[:class:]] under /u; the FLAGS field
569 gives which one
570 POSIXA none Some [[:class:]] under /a; the FLAGS field
571 gives which one
572 NPOSIXD none complement of POSIXD, [[:^class:]]
573 NPOSIXL none complement of POSIXL, [[:^class:]]
574 NPOSIXU none complement of POSIXU, [[:^class:]]
575 NPOSIXA none complement of POSIXA, [[:^class:]]
576 CLUMP no Match any extended grapheme cluster
578 sequence
579 # Alternation
581 # BRANCH The set of branches constituting a single choice are
583 # hooked together with their "next" pointers, since
584 # precedence prevents anything being concatenated to
585 # any individual branch. The "next" pointer of the last
586 # BRANCH in a choice points to the thing following the
587 # whole choice. This is also where the final "next"
588 # pointer of each individual branch points; each branch
589 # starts with the operand node of a BRANCH node.
590 #
591 BRANCH node 1 Match this alternative, or the next...
592 # Literals
594 EXACT str Match this string (flags field is the
596 length).
597 # In a long string node, the U32 argument is the length, and is
599 # immediately followed by the string.
600 LEXACT len:str 1 Match this long string (preceded by length;
601 flags unused).
602 EXACTL str Like EXACT, but /l is in effect (used so
603 locale-related warnings can be checked for)
604 EXACTF str Like EXACT, but match using /id rules;
605 (string not UTF-8, ASCII folded; non-ASCII
606 not)
607 EXACTFL str Like EXACT, but match using /il rules;
608 (string not likely to be folded)
609 EXACTFU str Like EXACT, but match using /iu rules;
610 (string folded)
611 EXACTFAA str Like EXACT, but match using /iaa rules;
613 (string folded except MICRO in non-UTF8
614 patterns; doesn't contain SHARP S unless
615 UTF-8; folded length <= unfolded)
616 EXACTFAA_NO_TRIE str Like EXACTFAA, (string not UTF-8, folded
617 except: MICRO, SHARP S; folded length <=
618 unfolded, not currently trie-able)
619 EXACTFUP str Like EXACT, but match using /iu rules;
621 (string not UTF-8, folded except MICRO:
622 hence Problematic)
623 EXACTFLU8 str Like EXACTFU, but use /il, UTF-8, (string
625 is folded, and everything in it is above
626 255
627 EXACT_REQ8 str Like EXACT, but only UTF-8 encoded targets
628 can match
629 LEXACT_REQ8 len:str 1 Like LEXACT, but only UTF-8 encoded targets
630 can match
631 EXACTFU_REQ8 str Like EXACTFU, but only UTF-8 encoded
632 targets can match
633 EXACTFU_S_EDGE str /di rules, but nothing in it precludes /ui,
635 except begins and/or ends with [Ss];
636 (string not UTF-8; compile-time only)
637 # New charclass like patterns
639 LNBREAK none generic newline pattern
640 # Trie Related
642 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
644 # have inline charclass data (ascii only), the 'C' store it in the
645 # structure.
646 TRIE trie 1 Match many EXACT(F[ALU]?)? at once.
648 flags==type
649 TRIEC trie Same as TRIE, but with embedded charclass
650 charclass data
651 AHOCORASICK trie 1 Aho Corasick stclass. flags==type
653 AHOCORASICKC trie Same as AHOCORASICK, but with embedded
654 charclass charclass data
655 # Do nothing types
657 NOTHING no Match empty string.
659 # A variant of above which delimits a group, thus stops optimizations
660 TAIL no Match empty string. Can jump here from
661 outside.
662 # Loops
664 # STAR,PLUS '?', and complex '*' and '+', are implemented as
666 # circular BRANCH structures. Simple cases
667 # (one character per match) are implemented with STAR
668 # and PLUS for speed and to minimize recursive plunges.
669 #
670 STAR node Match this (simple) thing 0 or more times:
671 /A{0,}B/ where A is width 1 char
672 PLUS node Match this (simple) thing 1 or more times:
673 /A{1,}B/ where A is width 1 char
674 CURLY sv 3 Match this (simple) thing {n,m} times:
676 /A{m,n}B/ where A is width 1 char
677 CURLYN no 3 Capture next-after-this simple thing:
678 /(A){m,n}B/ where A is width 1 char
679 CURLYM no 3 Capture this medium-complex thing {n,m}
680 times: /(A){m,n}B/ where A is fixed-length
681 CURLYX sv 3 Match/Capture this complex thing {n,m}
682 times.
683 # This terminator creates a loop structure for CURLYX
685 WHILEM no Do curly processing and see if rest
686 matches.
687 # Buffer related
689 # OPEN,CLOSE,GROUPP ...are numbered at compile time.
691 OPEN num 1 Mark this point in input as start of #n.
692 CLOSE num 1 Close corresponding OPEN of #n.
693 SROPEN none Same as OPEN, but for script run
694 SRCLOSE none Close preceding SROPEN
695 REF num 2 Match some already matched string
697 REFF num 2 Match already matched string, using /di
698 rules.
699 REFFL num 2 Match already matched string, using /li
700 rules.
701 REFFU num 2 Match already matched string, usng /ui.
702 REFFA num 2 Match already matched string, using /aai
703 rules.
704 # Named references. Code in regcomp.c assumes that these all are after
706 # the numbered references
707 REFN no-sv 2 Match some already matched string
708 REFFN no-sv 2 Match already matched string, using /di
709 rules.
710 REFFLN no-sv 2 Match already matched string, using /li
711 rules.
712 REFFUN num 2 Match already matched string, using /ui
713 rules.
714 REFFAN num 2 Match already matched string, using /aai
715 rules.
716 # Support for long RE
718 LONGJMP off 1 1 Jump far away.
719 BRANCHJ off 2 1 BRANCH with long offset.
720 # Special Case Regops
722 IFMATCH off 1 1 Succeeds if the following matches; non-zero
723 flags "f", next_off "o" means lookbehind
724 assertion starting "f..(f-o)" characters
725 before current
726 UNLESSM off 1 1 Fails if the following matches; non-zero
727 flags "f", next_off "o" means lookbehind
728 assertion starting "f..(f-o)" characters
729 before current
730 SUSPEND off 1 1 "Independent" sub-RE.
731 IFTHEN off 1 1 Switch, should be preceded by switcher.
732 GROUPP num 1 Whether the group matched.
733 # The heavy worker
735 EVAL evl/flags Execute some Perl code.
737 2
738 # Modifiers
740 MINMOD no Next operator is not greedy.
742 LOGICAL no Next opcode should set the flag only.
743 # This is not used yet
745 RENUM off 1 1 Group with independently numbered parens.
746 # Regex Subroutines
748 GOSUB num/ofs 2 recurse to paren arg1 at (signed) ofs arg2
749 # Special conditionals
751 GROUPPN 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.
754 # Backtracking Verbs
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
759 2 verbar
760 # 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
766 cut.
767 SKIP no-sv 1 On failure skip forward (to the mark)
768 before retrying
769 COMMIT no-sv 1 Pattern fails outright if backtracking
770 through this
771 CUTGROUP no-sv 1 On failure go to the next alternation in
772 the group
773 # Control what to keep in $&.
775 KEEPS no $& begins here.
776 # Validate that lookbehind IFMATCH and UNLESSM end at the right place
778 LOOKBEHIND_END no Return from lookbehind (IFMATCH/UNLESSM)
779 and validate position
780 # SPECIAL REGOPS
782 # This is not really a node, but an optimized away piece of a "long"
784 # node. To simplify debugging output, we mark it as if it were a node
785 OPTIMIZED off Placeholder for dump.
786 # Special opcode with the property that no opcode in a compiled program
788 # will ever be of this type. Thus it can be used as a flag value that
789 # no other opcode has been seen. END is used similarly, in that an END
790 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
791 # mean "not seen anything to optimize yet".
792 PSEUDO off Pseudo opcode for internal use.
793 REGEX_SET depth p Regex set, temporary node used in pre-
795 optimization compilation
796 Following the optimizer information is a dump of the offset/length
798 table, here split across several lines:
799 Offsets: [45]
801 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
802 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
803 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
804 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
805 The first line here indicates that the offset/length table contains 45
807 entries. Each entry is a pair of integers, denoted by
808 "offset[length]". Entries are numbered starting with 1, so entry #1
809 here is "1[4]" and entry #12 is "5[1]". "1[4]" indicates that the node
810 labeled "1:" (the "1: ANYOF[bc]") begins at character position 1 in the
811 pre-compiled form of the regex, and has a length of 4 characters.
812 "5[1]" in position 12 indicates that the node labeled "12:" (the "12:
813 EXACT <d>") begins at character position 5 in the pre-compiled form of
814 the regex, and has a length of 1 character. "12[1]" in position 14
815 indicates that the node labeled "14:" (the "14: CURLYX[0] {1,32767}")
816 begins at character position 12 in the pre-compiled form of the regex,
817 and has a length of 1 character---that is, it corresponds to the "+"
818 symbol in the precompiled regex.
819 "0[0]" items indicate that there is no corresponding node.
821 Run-time Output
823 First of all, when doing a match, one may get no run-time output even
824 if debugging is enabled. This means that the regex engine was never
825 entered and that all of the job was therefore done by the optimizer.
826 If the regex engine was entered, the output may look like this:
828 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
830 Setting an EVAL scope, savestack=3
831 2 <ab> <cdefg__gh_> | 1: ANYOF
832 3 <abc> <defg__gh_> | 11: EXACT <d>
833 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
834 4 <abcd> <efg__gh_> | 26: WHILEM
835 0 out of 1..32767 cc=effff31c
836 4 <abcd> <efg__gh_> | 15: OPEN1
837 4 <abcd> <efg__gh_> | 17: EXACT <e>
838 5 <abcde> <fg__gh_> | 19: STAR
839 EXACT <f> can match 1 times out of 32767...
840 Setting an EVAL scope, savestack=3
841 6 <bcdef> <g__gh__> | 22: EXACT <g>
842 7 <bcdefg> <__gh__> | 24: CLOSE1
843 7 <bcdefg> <__gh__> | 26: WHILEM
844 1 out of 1..32767 cc=effff31c
845 Setting an EVAL scope, savestack=12
846 7 <bcdefg> <__gh__> | 15: OPEN1
847 7 <bcdefg> <__gh__> | 17: EXACT <e>
848 restoring \1 to 4(4)..7
849 failed, try continuation...
850 7 <bcdefg> <__gh__> | 27: NOTHING
851 7 <bcdefg> <__gh__> | 28: EXACT <h>
852 failed...
853 failed...
854 The most significant information in the output is about the particular
856 node of the compiled regex that is currently being tested against the
857 target string. The format of these lines is
858 " "STRING-OFFSET <PRE-STRING> <POST-STRING> |ID: TYPE
860 The TYPE info is indented with respect to the backtracking level.
862 Other incidental information appears interspersed within.
863
865 Perl is a profligate wastrel when it comes to memory use. There is a
866 saying that to estimate memory usage of Perl, assume a reasonable
867 algorithm for memory allocation, multiply that estimate by 10, and
868 while you still may miss the mark, at least you won't be quite so
869 astonished. This is not absolutely true, but may provide a good grasp
870 of what happens.
871 Assume that an integer cannot take less than 20 bytes of memory, a
873 float cannot take less than 24 bytes, a string cannot take less than 32
874 bytes (all these examples assume 32-bit architectures, the result are
875 quite a bit worse on 64-bit architectures). If a variable is accessed
876 in two of three different ways (which require an integer, a float, or a
877 string), the memory footprint may increase yet another 20 bytes. A
878 sloppy malloc(3) implementation can inflate these numbers dramatically.
879 On the opposite end of the scale, a declaration like
881 sub foo;
883 may take up to 500 bytes of memory, depending on which release of Perl
885 you're running.
886 Anecdotal estimates of source-to-compiled code bloat suggest an
888 eightfold increase. This means that the compiled form of reasonable
889 (normally commented, properly indented etc.) code will take about eight
890 times more space in memory than the code took on disk.
891 The -DL command-line switch is obsolete since circa Perl 5.6.0 (it was
893 available only if Perl was built with "-DDEBUGGING"). The switch was
894 used to track Perl's memory allocations and possible memory leaks.
895 These days the use of malloc debugging tools like Purify or valgrind is
896 suggested instead. See also "PERL_MEM_LOG" in perlhacktips.
897 One way to find out how much memory is being used by Perl data
899 structures is to install the Devel::Size module from CPAN: it gives you
900 the minimum number of bytes required to store a particular data
901 structure. Please be mindful of the difference between the size() and
902 total_size().
903 If Perl has been compiled using Perl's malloc you can analyze Perl
905 memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
906 Using $ENV{PERL_DEBUG_MSTATS}
908 If your perl is using Perl's malloc() and was compiled with the
909 necessary switches (this is the default), then it will print memory
910 usage statistics after compiling your code when
911 "$ENV{PERL_DEBUG_MSTATS} > 1", and before termination of the program
912 when "$ENV{PERL_DEBUG_MSTATS} >= 1". The report format is similar to
913 the following example:
914 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
916 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
917 14216 free: 130 117 28 7 9 0 2 2 1 0 0
918 437 61 36 0 5
919 60924 used: 125 137 161 55 7 8 6 16 2 0 1
920 74 109 304 84 20
921 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
922 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
923 30888 free: 245 78 85 13 6 2 1 3 2 0 1
924 315 162 39 42 11
925 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
926 196 178 1066 798 39
927 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
928 It is possible to ask for such a statistic at arbitrary points in your
930 execution using the mstat() function out of the standard Devel::Peek
931 module.
932 Here is some explanation of that format:
934 "buckets SMALLEST(APPROX)..GREATEST(APPROX)"
936 Perl's malloc() uses bucketed allocations. Every request is
937 rounded up to the closest bucket size available, and a bucket is
938 taken from the pool of buckets of that size.
939 The line above describes the limits of buckets currently in use.
941 Each bucket has two sizes: memory footprint and the maximal size of
942 user data that can fit into this bucket. Suppose in the above
943 example that the smallest bucket were size 4. The biggest bucket
944 would have usable size 8188, and the memory footprint would be
945 8192.
946 In a Perl built for debugging, some buckets may have negative
948 usable size. This means that these buckets cannot (and will not)
949 be used. For larger buckets, the memory footprint may be one page
950 greater than a power of 2. If so, the corresponding power of two
951 is printed in the "APPROX" field above.
952 Free/Used
954 The 1 or 2 rows of numbers following that correspond to the number
955 of buckets of each size between "SMALLEST" and "GREATEST". In the
956 first row, the sizes (memory footprints) of buckets are powers of
957 two--or possibly one page greater. In the second row, if present,
958 the memory footprints of the buckets are between the memory
959 footprints of two buckets "above".
960 For example, suppose under the previous example, the memory
962 footprints were
963 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
965 4 12 24 48 80
966 With a non-"DEBUGGING" perl, the buckets starting from 128 have a
968 4-byte overhead, and thus an 8192-long bucket may take up to
969 8188-byte allocations.
970 "Total sbrk(): SBRKed/SBRKs:CONTINUOUS"
972 The first two fields give the total amount of memory perl sbrk(2)ed
973 (ess-broken? :-) and number of sbrk(2)s used. The third number is
974 what perl thinks about continuity of returned chunks. So long as
975 this number is positive, malloc() will assume that it is probable
976 that sbrk(2) will provide continuous memory.
977 Memory allocated by external libraries is not counted.
979 "pad: 0"
981 The amount of sbrk(2)ed memory needed to keep buckets aligned.
982 "heads: 2192"
984 Although memory overhead of bigger buckets is kept inside the
985 bucket, for smaller buckets, it is kept in separate areas. This
986 field gives the total size of these areas.
987 "chain: 0"
989 malloc() may want to subdivide a bigger bucket into smaller
990 buckets. If only a part of the deceased bucket is left
991 unsubdivided, the rest is kept as an element of a linked list.
992 This field gives the total size of these chunks.
993 "tail: 6144"
995 To minimize the number of sbrk(2)s, malloc() asks for more memory.
996 This field gives the size of the yet unused part, which is
997 sbrk(2)ed, but never touched.
998
1000 perldebug, perl5db.pl, perlguts, perlrun, re, and Devel::DProf.
1001perl v5.38.2 2023-11-30 PERLDEBGUTS(1)