1PERLDEBGUTS(1) Perl Programmers Reference Guide PERLDEBGUTS(1)
2
3
4
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 podpage in
19 the Perl source tree.
20
21 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
27 • Perl inserts the contents of $ENV{PERL5DB} (or "BEGIN {require
28 'perl5db.pl'}" if not present) before the first line of your
29 program.
30
31 • Each array "@{"_<$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 $filename for "eval"ed strings looks like "(eval 34)".
35
36 Values in this array are magical in numeric context: they compare
37 equal to zero only if the line is not breakable.
38
39 • Each hash "%{"_<$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
45 The same holds for evaluated strings that contain subroutines, or
46 which are currently being executed. The $filename for "eval"ed
47 strings looks like "(eval 34)".
48
49 • Each scalar "${"_<$filename"}" contains $filename. This is also
50 the case for evaluated strings that contain subroutines, or which
51 are currently being executed. The $filename for "eval"ed strings
52 looks like "(eval 34)".
53
54 • After each "require"d file is compiled, but before it is executed,
55 "DB::postponed(*{"_<$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
59 • 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"
62 subroutine also exists.
63
64 • 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
69 • 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
76 • 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
86 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
90 • 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
101 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
108 Writing Your Own Debugger
109 Environment Variables
110
111 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
115 sub DB::DB {}
116
117 It can easily be defined like this:
118
119 $ PERL5DB="sub DB::DB {}" perl -d your-script
120
121 Another brief debugger, slightly more useful, can be created with only
122 the line:
123
124 sub DB::DB {print ++$i; scalar <STDIN>}
125
126 This debugger prints a number which increments for each statement
127 encountered and waits for you to hit a newline before continuing to the
128 next statement.
129
130 The following debugger is actually useful:
131
132 {
133 package DB;
134 sub DB {}
135 sub sub {print ++$i, " $sub\n"; &$sub}
136 }
137
138 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
142 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
147 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
152 Debugger Internal Variables
153
154 In addition to the file and subroutine-related variables mentioned
155 above, the debugger also maintains various magical internal variables.
156
157 • @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
162 Values in this array are magical in numeric context: they compare
163 equal to zero only if the line is not breakable.
164
165 • %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
170 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
175 Debugger Customization Functions
176
177 Some functions are provided to simplify customization.
178
179 • See "Configurable Options" in perldebug for a description of
180 options parsed by "DB::parse_options(string)".
181
182 • "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
189 • "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
193 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
201 $ perl -de 42
202 Stack dump during die enabled outside of evals.
203
204 Loading DB routines from perl5db.pl patch level 0.94
205 Emacs support available.
206
207 Enter h or 'h h' for help.
208
209 main::(-e:1): 0
210 DB<1> sub foo { 14 }
211
212 DB<2> sub bar { 3 }
213
214 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
220 with this one, once the "o"ption "frame=2" has been set:
221
222 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
234 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
241 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
256 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
277 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
294 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
317 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
337 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 $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
352 scalar context return from Exporter::import: ''
353
354 In all cases shown above, the line indentation shows the call tree. If
355 bit 2 of "frame" is set, a line is printed on exit from a subroutine as
356 well. If bit 4 is set, the arguments are printed along with the caller
357 info. If bit 8 is set, the arguments are printed even if they are tied
358 or references. If bit 16 is set, the return value is printed, too.
359
360 When a package is compiled, a line like this
361
362 Package lib/Carp.pm.
363
364 is printed with proper indentation.
365
367 There are two ways to enable debugging output for regular expressions.
368
369 If your perl is compiled with "-DDEBUGGING", you may use the -Dr flag
370 on the command line, and "-Drv" for more verbose information.
371
372 Otherwise, one can "use re 'debug'", which has effects at both compile
373 time and run time. Since Perl 5.9.5, this pragma is lexically scoped.
374
375 Compile-time Output
376 The debugging output at compile time looks like this:
377
378 Compiling REx '[bc]d(ef*g)+h[ij]k$'
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
408 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
414 The
415
416 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
417 stclass 'ANYOF[bc]' minlen 7
418
419 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
429 The fields of interest which may appear in this line are
430
431 "anchored" STRING "at" POS
432 "floating" STRING "at" POS1..POS2
433 See above.
434
435 "matching floating/anchored"
436 Which substring to check first.
437
438 "minlen"
439 The minimal length of the match.
440
441 "stclass" TYPE
442 Type of first matching node.
443
444 "noscan"
445 Don't scan for the found substrings.
446
447 "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
452 "GPOS"
453 Set if the pattern contains "\G".
454
455 "plus"
456 Set if the pattern starts with a repeated char (as in "x+y").
457
458 "implicit"
459 Set if the pattern starts with ".*".
460
461 "with eval"
462 Set if the pattern contain eval-groups, such as "(?{ code })" and
463 "(??{ code })".
464
465 "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
469 If a substring is known to match at end-of-line only, it may be
470 followed by "$", as in "floating 'k'$".
471
472 The optimizer-specific information is used to avoid entering (a slow)
473 regex engine on strings that will not definitely match. If the "isall"
474 flag is set, a call to the regex engine may be avoided even when the
475 optimizer found an appropriate place for the match.
476
477 Above the optimizer section is the list of nodes of the compiled form
478 of the regex. Each line has format
479
480 " "id: TYPE OPTIONAL-INFO (next-id)
481
482 Types of Nodes
483 Here are the current possible types, with short descriptions:
484
485 # TYPE arg-description [regnode-struct-suffix] [longjump-len] DESCRIPTION
486
487 # Exit points
488
489 END no End of program.
490 SUCCEED no Return from a subroutine, basically.
491
492 # Line Start Anchors:
493 SBOL no Match "" at beginning of line: /^/, /\A/
494 MBOL no Same, assuming multiline: /^/m
495
496 # Line End Anchors:
497 SEOL no Match "" at end of line: /$/
498 MEOL no Same, assuming multiline: /$/m
499 EOS no Match "" at end of string: /\z/
500
501 # Match Start Anchors:
502 GPOS no Matches where last m//g left off.
503
504 # 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
522 # [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
535 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 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
555 ANYOFM byte 1 Like ANYOF, but matches an invariant byte
556 as determined by the mask and arg
557 NANYOFM byte 1 complement of ANYOFM
558
559 # POSIX Character Classes:
560 POSIXD none Some [[:class:]] under /d; the FLAGS field
561 gives which one
562 POSIXL none Some [[:class:]] under /l; the FLAGS field
563 gives which one
564 POSIXU none Some [[:class:]] under /u; the FLAGS field
565 gives which one
566 POSIXA none Some [[:class:]] under /a; the FLAGS field
567 gives which one
568 NPOSIXD none complement of POSIXD, [[:^class:]]
569 NPOSIXL none complement of POSIXL, [[:^class:]]
570 NPOSIXU none complement of POSIXU, [[:^class:]]
571 NPOSIXA none complement of POSIXA, [[:^class:]]
572
573 CLUMP no Match any extended grapheme cluster
574 sequence
575
576 # Alternation
577
578 # BRANCH The set of branches constituting a single choice are
579 # hooked together with their "next" pointers, since
580 # precedence prevents anything being concatenated to
581 # any individual branch. The "next" pointer of the last
582 # BRANCH in a choice points to the thing following the
583 # whole choice. This is also where the final "next"
584 # pointer of each individual branch points; each branch
585 # starts with the operand node of a BRANCH node.
586 #
587 BRANCH node Match this alternative, or the next...
588
589 # Literals
590
591 EXACT str Match this string (flags field is the
592 length).
593
594 # In a long string node, the U32 argument is the length, and is
595 # immediately followed by the string.
596 LEXACT len:str 1 Match this long string (preceded by length;
597 flags unused).
598 EXACTL str Like EXACT, but /l is in effect (used so
599 locale-related warnings can be checked for)
600 EXACTF str Like EXACT, but match using /id rules;
601 (string not UTF-8, ASCII folded; non-ASCII
602 not)
603 EXACTFL str Like EXACT, but match using /il rules;
604 (string not likely to be folded)
605 EXACTFU str Like EXACT, but match using /iu rules;
606 (string folded)
607
608 EXACTFAA str Like EXACT, but match using /iaa rules;
609 (string folded except MICRO in non-UTF8
610 patterns; doesn't contain SHARP S unless
611 UTF-8; folded length <= unfolded)
612 EXACTFAA_NO_TRIE str Like EXACTFAA, (string not UTF-8, folded
613 except: MICRO, SHARP S; folded length <=
614 unfolded, not currently trie-able)
615
616 EXACTFUP str Like EXACT, but match using /iu rules;
617 (string not UTF-8, folded except MICRO:
618 hence Problematic)
619
620 EXACTFLU8 str Like EXACTFU, but use /il, UTF-8, (string
621 is folded, and everything in it is above
622 255
623 EXACT_REQ8 str Like EXACT, but only UTF-8 encoded targets
624 can match
625 LEXACT_REQ8 len:str 1 Like LEXACT, but only UTF-8 encoded targets
626 can match
627 EXACTFU_REQ8 str Like EXACTFU, but only UTF-8 encoded
628 targets can match
629
630 EXACTFU_S_EDGE str /di rules, but nothing in it precludes /ui,
631 except begins and/or ends with [Ss];
632 (string not UTF-8; compile-time only)
633
634 # New charclass like patterns
635 LNBREAK none generic newline pattern
636
637 # Trie Related
638
639 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
640 # have inline charclass data (ascii only), the 'C' store it in the
641 # structure.
642
643 TRIE trie 1 Match many EXACT(F[ALU]?)? at once.
644 flags==type
645 TRIEC trie Same as TRIE, but with embedded charclass
646 charclass data
647
648 AHOCORASICK trie 1 Aho Corasick stclass. flags==type
649 AHOCORASICKC trie Same as AHOCORASICK, but with embedded
650 charclass charclass data
651
652 # Do nothing types
653
654 NOTHING no Match empty string.
655 # A variant of above which delimits a group, thus stops optimizations
656 TAIL no Match empty string. Can jump here from
657 outside.
658
659 # Loops
660
661 # STAR,PLUS '?', and complex '*' and '+', are implemented as
662 # circular BRANCH structures. Simple cases
663 # (one character per match) are implemented with STAR
664 # and PLUS for speed and to minimize recursive plunges.
665 #
666 STAR node Match this (simple) thing 0 or more times.
667 PLUS node Match this (simple) thing 1 or more times.
668
669 CURLY sv 2 Match this simple thing {n,m} times.
670 CURLYN no 2 Capture next-after-this simple thing
671 CURLYM no 2 Capture this medium-complex thing {n,m}
672 times.
673 CURLYX sv 2 Match this complex thing {n,m} times.
674
675 # This terminator creates a loop structure for CURLYX
676 WHILEM no Do curly processing and see if rest
677 matches.
678
679 # Buffer related
680
681 # OPEN,CLOSE,GROUPP ...are numbered at compile time.
682 OPEN num 1 Mark this point in input as start of #n.
683 CLOSE num 1 Close corresponding OPEN of #n.
684 SROPEN none Same as OPEN, but for script run
685 SRCLOSE none Close preceding SROPEN
686
687 REF num 1 Match some already matched string
688 REFF num 1 Match already matched string, using /di
689 rules.
690 REFFL num 1 Match already matched string, using /li
691 rules.
692 REFFU num 1 Match already matched string, usng /ui.
693 REFFA num 1 Match already matched string, using /aai
694 rules.
695
696 # Named references. Code in regcomp.c assumes that these all are after
697 # the numbered references
698 REFN no-sv 1 Match some already matched string
699 REFFN no-sv 1 Match already matched string, using /di
700 rules.
701 REFFLN no-sv 1 Match already matched string, using /li
702 rules.
703 REFFUN num 1 Match already matched string, using /ui
704 rules.
705 REFFAN num 1 Match already matched string, using /aai
706 rules.
707
708 # Support for long RE
709 LONGJMP off 1 1 Jump far away.
710 BRANCHJ off 1 1 BRANCH with long offset.
711
712 # Special Case Regops
713 IFMATCH off 1 1 Succeeds if the following matches; non-zero
714 flags "f", next_off "o" means lookbehind
715 assertion starting "f..(f-o)" characters
716 before current
717 UNLESSM off 1 1 Fails if the following matches; non-zero
718 flags "f", next_off "o" means lookbehind
719 assertion starting "f..(f-o)" characters
720 before current
721 SUSPEND off 1 1 "Independent" sub-RE.
722 IFTHEN off 1 1 Switch, should be preceded by switcher.
723 GROUPP num 1 Whether the group matched.
724
725 # The heavy worker
726
727 EVAL evl/flags Execute some Perl code.
728 2L
729
730 # Modifiers
731
732 MINMOD no Next operator is not greedy.
733 LOGICAL no Next opcode should set the flag only.
734
735 # This is not used yet
736 RENUM off 1 1 Group with independently numbered parens.
737
738 # Regex Subroutines
739 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
740
741 # Special conditionals
742 GROUPPN no-sv 1 Whether the group matched.
743 INSUBP num 1 Whether we are in a specific recurse.
744 DEFINEP none 1 Never execute directly.
745
746 # Backtracking Verbs
747 ENDLIKE none Used only for the type field of verbs
748 OPFAIL no-sv 1 Same as (?!), but with verb arg
749 ACCEPT no-sv/num Accepts the current matched string, with
750 2L verbar
751
752 # Verbs With Arguments
753 VERB no-sv 1 Used only for the type field of verbs
754 PRUNE no-sv 1 Pattern fails at this startpoint if no-
755 backtracking through this
756 MARKPOINT no-sv 1 Push the current location for rollback by
757 cut.
758 SKIP no-sv 1 On failure skip forward (to the mark)
759 before retrying
760 COMMIT no-sv 1 Pattern fails outright if backtracking
761 through this
762 CUTGROUP no-sv 1 On failure go to the next alternation in
763 the group
764
765 # Control what to keep in $&.
766 KEEPS no $& begins here.
767
768 # SPECIAL REGOPS
769
770 # This is not really a node, but an optimized away piece of a "long"
771 # node. To simplify debugging output, we mark it as if it were a node
772 OPTIMIZED off Placeholder for dump.
773
774 # Special opcode with the property that no opcode in a compiled program
775 # will ever be of this type. Thus it can be used as a flag value that
776 # no other opcode has been seen. END is used similarly, in that an END
777 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
778 # mean "not seen anything to optimize yet".
779 PSEUDO off Pseudo opcode for internal use.
780
781 REGEX_SET depth p Regex set, temporary node used in pre-
782 optimization compilation
783
784 Following the optimizer information is a dump of the offset/length
785 table, here split across several lines:
786
787 Offsets: [45]
788 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
789 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
790 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
791 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
792
793 The first line here indicates that the offset/length table contains 45
794 entries. Each entry is a pair of integers, denoted by
795 "offset[length]". Entries are numbered starting with 1, so entry #1
796 here is "1[4]" and entry #12 is "5[1]". "1[4]" indicates that the node
797 labeled "1:" (the "1: ANYOF[bc]") begins at character position 1 in the
798 pre-compiled form of the regex, and has a length of 4 characters.
799 "5[1]" in position 12 indicates that the node labeled "12:" (the "12:
800 EXACT <d>") begins at character position 5 in the pre-compiled form of
801 the regex, and has a length of 1 character. "12[1]" in position 14
802 indicates that the node labeled "14:" (the "14: CURLYX[0] {1,32767}")
803 begins at character position 12 in the pre-compiled form of the regex,
804 and has a length of 1 character---that is, it corresponds to the "+"
805 symbol in the precompiled regex.
806
807 "0[0]" items indicate that there is no corresponding node.
808
809 Run-time Output
810 First of all, when doing a match, one may get no run-time output even
811 if debugging is enabled. This means that the regex engine was never
812 entered and that all of the job was therefore done by the optimizer.
813
814 If the regex engine was entered, the output may look like this:
815
816 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
817 Setting an EVAL scope, savestack=3
818 2 <ab> <cdefg__gh_> | 1: ANYOF
819 3 <abc> <defg__gh_> | 11: EXACT <d>
820 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
821 4 <abcd> <efg__gh_> | 26: WHILEM
822 0 out of 1..32767 cc=effff31c
823 4 <abcd> <efg__gh_> | 15: OPEN1
824 4 <abcd> <efg__gh_> | 17: EXACT <e>
825 5 <abcde> <fg__gh_> | 19: STAR
826 EXACT <f> can match 1 times out of 32767...
827 Setting an EVAL scope, savestack=3
828 6 <bcdef> <g__gh__> | 22: EXACT <g>
829 7 <bcdefg> <__gh__> | 24: CLOSE1
830 7 <bcdefg> <__gh__> | 26: WHILEM
831 1 out of 1..32767 cc=effff31c
832 Setting an EVAL scope, savestack=12
833 7 <bcdefg> <__gh__> | 15: OPEN1
834 7 <bcdefg> <__gh__> | 17: EXACT <e>
835 restoring \1 to 4(4)..7
836 failed, try continuation...
837 7 <bcdefg> <__gh__> | 27: NOTHING
838 7 <bcdefg> <__gh__> | 28: EXACT <h>
839 failed...
840 failed...
841
842 The most significant information in the output is about the particular
843 node of the compiled regex that is currently being tested against the
844 target string. The format of these lines is
845
846 " "STRING-OFFSET <PRE-STRING> <POST-STRING> |ID: TYPE
847
848 The TYPE info is indented with respect to the backtracking level.
849 Other incidental information appears interspersed within.
850
852 Perl is a profligate wastrel when it comes to memory use. There is a
853 saying that to estimate memory usage of Perl, assume a reasonable
854 algorithm for memory allocation, multiply that estimate by 10, and
855 while you still may miss the mark, at least you won't be quite so
856 astonished. This is not absolutely true, but may provide a good grasp
857 of what happens.
858
859 Assume that an integer cannot take less than 20 bytes of memory, a
860 float cannot take less than 24 bytes, a string cannot take less than 32
861 bytes (all these examples assume 32-bit architectures, the result are
862 quite a bit worse on 64-bit architectures). If a variable is accessed
863 in two of three different ways (which require an integer, a float, or a
864 string), the memory footprint may increase yet another 20 bytes. A
865 sloppy malloc(3) implementation can inflate these numbers dramatically.
866
867 On the opposite end of the scale, a declaration like
868
869 sub foo;
870
871 may take up to 500 bytes of memory, depending on which release of Perl
872 you're running.
873
874 Anecdotal estimates of source-to-compiled code bloat suggest an
875 eightfold increase. This means that the compiled form of reasonable
876 (normally commented, properly indented etc.) code will take about eight
877 times more space in memory than the code took on disk.
878
879 The -DL command-line switch is obsolete since circa Perl 5.6.0 (it was
880 available only if Perl was built with "-DDEBUGGING"). The switch was
881 used to track Perl's memory allocations and possible memory leaks.
882 These days the use of malloc debugging tools like Purify or valgrind is
883 suggested instead. See also "PERL_MEM_LOG" in perlhacktips.
884
885 One way to find out how much memory is being used by Perl data
886 structures is to install the Devel::Size module from CPAN: it gives you
887 the minimum number of bytes required to store a particular data
888 structure. Please be mindful of the difference between the size() and
889 total_size().
890
891 If Perl has been compiled using Perl's malloc you can analyze Perl
892 memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
893
894 Using $ENV{PERL_DEBUG_MSTATS}
895 If your perl is using Perl's malloc() and was compiled with the
896 necessary switches (this is the default), then it will print memory
897 usage statistics after compiling your code when
898 "$ENV{PERL_DEBUG_MSTATS} > 1", and before termination of the program
899 when "$ENV{PERL_DEBUG_MSTATS} >= 1". The report format is similar to
900 the following example:
901
902 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
903 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
904 14216 free: 130 117 28 7 9 0 2 2 1 0 0
905 437 61 36 0 5
906 60924 used: 125 137 161 55 7 8 6 16 2 0 1
907 74 109 304 84 20
908 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
909 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
910 30888 free: 245 78 85 13 6 2 1 3 2 0 1
911 315 162 39 42 11
912 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
913 196 178 1066 798 39
914 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
915
916 It is possible to ask for such a statistic at arbitrary points in your
917 execution using the mstat() function out of the standard Devel::Peek
918 module.
919
920 Here is some explanation of that format:
921
922 "buckets SMALLEST(APPROX)..GREATEST(APPROX)"
923 Perl's malloc() uses bucketed allocations. Every request is
924 rounded up to the closest bucket size available, and a bucket is
925 taken from the pool of buckets of that size.
926
927 The line above describes the limits of buckets currently in use.
928 Each bucket has two sizes: memory footprint and the maximal size of
929 user data that can fit into this bucket. Suppose in the above
930 example that the smallest bucket were size 4. The biggest bucket
931 would have usable size 8188, and the memory footprint would be
932 8192.
933
934 In a Perl built for debugging, some buckets may have negative
935 usable size. This means that these buckets cannot (and will not)
936 be used. For larger buckets, the memory footprint may be one page
937 greater than a power of 2. If so, the corresponding power of two
938 is printed in the "APPROX" field above.
939
940 Free/Used
941 The 1 or 2 rows of numbers following that correspond to the number
942 of buckets of each size between "SMALLEST" and "GREATEST". In the
943 first row, the sizes (memory footprints) of buckets are powers of
944 two--or possibly one page greater. In the second row, if present,
945 the memory footprints of the buckets are between the memory
946 footprints of two buckets "above".
947
948 For example, suppose under the previous example, the memory
949 footprints were
950
951 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
952 4 12 24 48 80
953
954 With a non-"DEBUGGING" perl, the buckets starting from 128 have a
955 4-byte overhead, and thus an 8192-long bucket may take up to
956 8188-byte allocations.
957
958 "Total sbrk(): SBRKed/SBRKs:CONTINUOUS"
959 The first two fields give the total amount of memory perl sbrk(2)ed
960 (ess-broken? :-) and number of sbrk(2)s used. The third number is
961 what perl thinks about continuity of returned chunks. So long as
962 this number is positive, malloc() will assume that it is probable
963 that sbrk(2) will provide continuous memory.
964
965 Memory allocated by external libraries is not counted.
966
967 "pad: 0"
968 The amount of sbrk(2)ed memory needed to keep buckets aligned.
969
970 "heads: 2192"
971 Although memory overhead of bigger buckets is kept inside the
972 bucket, for smaller buckets, it is kept in separate areas. This
973 field gives the total size of these areas.
974
975 "chain: 0"
976 malloc() may want to subdivide a bigger bucket into smaller
977 buckets. If only a part of the deceased bucket is left
978 unsubdivided, the rest is kept as an element of a linked list.
979 This field gives the total size of these chunks.
980
981 "tail: 6144"
982 To minimize the number of sbrk(2)s, malloc() asks for more memory.
983 This field gives the size of the yet unused part, which is
984 sbrk(2)ed, but never touched.
985
987 perldebug, perl5db.pl, perlguts, perlrun, re, and Devel::DProf.
988
989
990
991perl v5.34.0 2021-10-18 PERLDEBGUTS(1)