1=head1 NAME 2 3perldebguts - Guts of Perl debugging 4 5=head1 DESCRIPTION 6 7This is not L<perldebug>, which tells you how to use 8the debugger. This manpage describes low-level details concerning 9the debugger's internals, which range from difficult to impossible 10to understand for anyone who isn't incredibly intimate with Perl's guts. 11Caveat lector. 12 13=head1 Debugger Internals 14 15Perl has special debugging hooks at compile-time and run-time used 16to create debugging environments. These hooks are not to be confused 17with the I<perl -Dxxx> command described in L<perlrun>, which is 18usable only if a special Perl is built per the instructions in the 19F<INSTALL> podpage in the Perl source tree. 20 21For example, whenever you call Perl's built-in C<caller> function 22from the package C<DB>, the arguments that the corresponding stack 23frame was called with are copied to the C<@DB::args> array. These 24mechanisms are enabled by calling Perl with the B<-d> switch. 25Specifically, the following additional features are enabled 26(cf. L<perlvar/$^P>): 27 28=over 4 29 30=item * 31 32Perl inserts the contents of C<$ENV{PERL5DB}> (or C<BEGIN {require 33'perl5db.pl'}> if not present) before the first line of your program. 34 35=item * 36 37Each array C<@{"_<$filename"}> holds the lines of $filename for a 38file compiled by Perl. The same is also true for C<eval>ed strings 39that contain subroutines, or which are currently being executed. 40The $filename for C<eval>ed strings looks like C<(eval 34)>. 41 42Values in this array are magical in numeric context: they compare 43equal to zero only if the line is not breakable. 44 45=item * 46 47Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed 48by line number. Individual entries (as opposed to the whole hash) 49are settable. Perl only cares about Boolean true here, although 50the values used by F<perl5db.pl> have the form 51C<"$break_condition\0$action">. 52 53The same holds for evaluated strings that contain subroutines, or 54which are currently being executed. The $filename for C<eval>ed strings 55looks like C<(eval 34)>. 56 57=item * 58 59Each scalar C<${"_<$filename"}> contains C<"_<$filename">. This is 60also the case for evaluated strings that contain subroutines, or 61which are currently being executed. The $filename for C<eval>ed 62strings looks like C<(eval 34)>. 63 64=item * 65 66After each C<require>d file is compiled, but before it is executed, 67C<DB::postponed(*{"_<$filename"})> is called if the subroutine 68C<DB::postponed> exists. Here, the $filename is the expanded name of 69the C<require>d file, as found in the values of %INC. 70 71=item * 72 73After each subroutine C<subname> is compiled, the existence of 74C<$DB::postponed{subname}> is checked. If this key exists, 75C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine 76also exists. 77 78=item * 79 80A hash C<%DB::sub> is maintained, whose keys are subroutine names 81and whose values have the form C<filename:startline-endline>. 82C<filename> has the form C<(eval 34)> for subroutines defined inside 83C<eval>s. 84 85=item * 86 87When the execution of your program reaches a point that can hold a 88breakpoint, the C<DB::DB()> subroutine is called if any of the variables 89C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true. These variables 90are not C<local>izable. This feature is disabled when executing 91inside C<DB::DB()>, including functions called from it 92unless C<< $^D & (1<<30) >> is true. 93 94=item * 95 96When execution of the program reaches a subroutine call, a call to 97C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> set to identify 98the called subroutine. (This doesn't happen if the calling subroutine 99was compiled in the C<DB> package.) C<$DB::sub> normally holds the name 100of the called subroutine, if it has a name by which it can be looked up. 101Failing that, C<$DB::sub> will hold a reference to the called subroutine. 102Either way, the C<&DB::sub> subroutine can use C<$DB::sub> as a reference 103by which to call the called subroutine, which it will normally want to do. 104 105X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub> 106is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling 107back to C<&DB::sub>(I<args>). 108 109=item * 110 111When execution of the program uses C<goto> to enter a non-XS subroutine 112and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto> is made, with 113C<$DB::sub> set to identify the subroutine being entered. The call to 114C<&DB::goto> does not replace the C<goto>; the requested subroutine will 115still be entered once C<&DB::goto> has returned. C<$DB::sub> normally 116holds the name of the subroutine being entered, if it has one. Failing 117that, C<$DB::sub> will hold a reference to the subroutine being entered. 118Unlike when C<&DB::sub> is called, it is not guaranteed that C<$DB::sub> 119can be used as a reference to operate on the subroutine being entered. 120 121=back 122 123Note that if C<&DB::sub> needs external data for it to work, no 124subroutine call is possible without it. As an example, the standard 125debugger's C<&DB::sub> depends on the C<$DB::deep> variable 126(it defines how many levels of recursion deep into the debugger you can go 127before a mandatory break). If C<$DB::deep> is not defined, subroutine 128calls are not possible, even though C<&DB::sub> exists. 129 130=head2 Writing Your Own Debugger 131 132=head3 Environment Variables 133 134The C<PERL5DB> environment variable can be used to define a debugger. 135For example, the minimal "working" debugger (it actually doesn't do anything) 136consists of one line: 137 138 sub DB::DB {} 139 140It can easily be defined like this: 141 142 $ PERL5DB="sub DB::DB {}" perl -d your-script 143 144Another brief debugger, slightly more useful, can be created 145with only the line: 146 147 sub DB::DB {print ++$i; scalar <STDIN>} 148 149This debugger prints a number which increments for each statement 150encountered and waits for you to hit a newline before continuing 151to the next statement. 152 153The following debugger is actually useful: 154 155 { 156 package DB; 157 sub DB {} 158 sub sub {print ++$i, " $sub\n"; &$sub} 159 } 160 161It prints the sequence number of each subroutine call and the name of the 162called subroutine. Note that C<&DB::sub> is being compiled into the 163package C<DB> through the use of the C<package> directive. 164 165When it starts, the debugger reads your rc file (F<./.perldb> or 166F<~/.perldb> under Unix), which can set important options. 167(A subroutine (C<&afterinit>) can be defined here as well; it is executed 168after the debugger completes its own initialization.) 169 170After the rc file is read, the debugger reads the PERLDB_OPTS 171environment variable and uses it to set debugger options. The 172contents of this variable are treated as if they were the argument 173of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">). 174 175=head3 Debugger Internal Variables 176 177In addition to the file and subroutine-related variables mentioned above, 178the debugger also maintains various magical internal variables. 179 180=over 4 181 182=item * 183 184C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which 185holds the lines of the currently-selected file (compiled by Perl), either 186explicitly chosen with the debugger's C<f> command, or implicitly by flow 187of execution. 188 189Values in this array are magical in numeric context: they compare 190equal to zero only if the line is not breakable. 191 192=item * 193 194C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which 195contains breakpoints and actions keyed by line number in 196the currently-selected file, either explicitly chosen with the 197debugger's C<f> command, or implicitly by flow of execution. 198 199As previously noted, individual entries (as opposed to the whole hash) 200are settable. Perl only cares about Boolean true here, although 201the values used by F<perl5db.pl> have the form 202C<"$break_condition\0$action">. 203 204=back 205 206=head3 Debugger Customization Functions 207 208Some functions are provided to simplify customization. 209 210=over 4 211 212=item * 213 214See L<perldebug/"Configurable Options"> for a description of options parsed by 215C<DB::parse_options(string)>. 216 217=item * 218 219C<DB::dump_trace(skip[,count])> skips the specified number of frames 220and returns a list containing information about the calling frames (all 221of them, if C<count> is missing). Each entry is reference to a hash 222with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine 223name, or info about C<eval>), C<args> (C<undef> or a reference to 224an array), C<file>, and C<line>. 225 226=item * 227 228C<DB::print_trace(FH, skip[, count[, short]])> prints 229formatted info about caller frames. The last two functions may be 230convenient as arguments to C<< < >>, C<< << >> commands. 231 232=back 233 234Note that any variables and functions that are not documented in 235this manpages (or in L<perldebug>) are considered for internal 236use only, and as such are subject to change without notice. 237 238=head1 Frame Listing Output Examples 239 240The C<frame> option can be used to control the output of frame 241information. For example, contrast this expression trace: 242 243 $ perl -de 42 244 Stack dump during die enabled outside of evals. 245 246 Loading DB routines from perl5db.pl patch level 0.94 247 Emacs support available. 248 249 Enter h or 'h h' for help. 250 251 main::(-e:1): 0 252 DB<1> sub foo { 14 } 253 254 DB<2> sub bar { 3 } 255 256 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): 260 42 261 262with this one, once the C<o>ption C<frame=2> has been set: 263 264 DB<4> o f=2 265 frame = '2' 266 DB<5> t print foo() * bar() 267 3: foo() * bar() 268 entering main::foo 269 2: sub foo { 14 }; 270 exited main::foo 271 entering main::bar 272 2: sub bar { 3 }; 273 exited main::bar 274 42 275 276By way of demonstration, we present below a laborious listing 277resulting from setting your C<PERLDB_OPTS> environment variable to 278the value C<f=n N>, and running I<perl -d -V> from the command line. 279Examples using various values of C<n> are shown to give you a feel 280for the difference between settings. Long though it may be, this 281is not a complete listing, but only excerpts. 282 283=over 4 284 285=item 1 286 287 entering main::BEGIN 288 entering Config::BEGIN 289 Package lib/Exporter.pm. 290 Package lib/Carp.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 300 301=item 2 302 303 entering main::BEGIN 304 entering Config::BEGIN 305 Package lib/Exporter.pm. 306 Package lib/Carp.pm. 307 exited Config::BEGIN 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 315 exited main::BEGIN 316 entering Config::myconfig 317 entering Config::FETCH 318 exited Config::FETCH 319 entering Config::FETCH 320 exited Config::FETCH 321 entering Config::FETCH 322 323=item 3 324 325 in $=main::BEGIN() from /dev/null:0 326 in $=Config::BEGIN() from lib/Config.pm:2 327 Package lib/Exporter.pm. 328 Package lib/Carp.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:574 340 341=item 4 342 343 in $=main::BEGIN() from /dev/null:0 344 in $=Config::BEGIN() from lib/Config.pm:2 345 Package lib/Exporter.pm. 346 Package lib/Carp.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:574 364 365=item 5 366 367 in $=main::BEGIN() from /dev/null:0 368 in $=Config::BEGIN() from lib/Config.pm:2 369 Package lib/Exporter.pm. 370 Package lib/Carp.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:574 385 386=item 6 387 388 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 $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 402 scalar context return from Exporter::import: '' 403 404=back 405 406In all cases shown above, the line indentation shows the call tree. 407If bit 2 of C<frame> is set, a line is printed on exit from a 408subroutine as well. If bit 4 is set, the arguments are printed 409along with the caller info. If bit 8 is set, the arguments are 410printed even if they are tied or references. If bit 16 is set, the 411return value is printed, too. 412 413When a package is compiled, a line like this 414 415 Package lib/Carp.pm. 416 417is printed with proper indentation. 418 419=head1 Debugging Regular Expressions 420 421There are two ways to enable debugging output for regular expressions. 422 423If your perl is compiled with C<-DDEBUGGING>, you may use the 424B<-Dr> flag on the command line, and C<-Drv> for more verbose 425information. 426 427Otherwise, one can C<use re 'debug'>, which has effects at both 428compile time and run time. Since Perl 5.9.5, this pragma is lexically 429scoped. 430 431=head2 Compile-time Output 432 433The debugging output at compile time looks like this: 434 435 Compiling REx '[bc]d(ef*g)+h[ij]k$' 436 size 45 Got 364 bytes for offset annotations. 437 first at 1 438 rarest char g at 0 439 rarest char d at 0 440 1: ANYOF[bc](12) 441 12: EXACT <d>(14) 442 14: CURLYX[0] {1,32767}(28) 443 16: OPEN1(18) 444 18: EXACT <e>(20) 445 20: STAR(23) 446 21: EXACT <f>(0) 447 23: EXACT <g>(25) 448 25: CLOSE1(27) 449 27: WHILEM[1/1](0) 450 28: NOTHING(29) 451 29: EXACT <h>(31) 452 31: ANYOF[ij](42) 453 42: EXACT <k>(44) 454 44: EOL(45) 455 45: END(0) 456 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 457 stclass 'ANYOF[bc]' minlen 7 458 Offsets: [45] 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. 464 465The first line shows the pre-compiled form of the regex. The second 466shows the size of the compiled form (in arbitrary units, usually 4674-byte words) and the total number of bytes allocated for the 468offset/length table, usually 4+C<size>*8. The next line shows the 469label I<id> of the first node that does a match. 470 471The 472 473 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 474 stclass 'ANYOF[bc]' minlen 7 475 476line (split into two lines above) contains optimizer 477information. In the example shown, the optimizer found that the match 478should contain a substring C<de> at offset 1, plus substring C<gh> 479at some offset between 3 and infinity. Moreover, when checking for 480these substrings (to abandon impossible matches quickly), Perl will check 481for the substring C<gh> before checking for the substring C<de>. The 482optimizer may also use the knowledge that the match starts (at the 483C<first> I<id>) with a character class, and no string 484shorter than 7 characters can possibly match. 485 486The fields of interest which may appear in this line are 487 488=over 4 489 490=item C<anchored> I<STRING> C<at> I<POS> 491 492=item C<floating> I<STRING> C<at> I<POS1..POS2> 493 494See above. 495 496=item C<matching floating/anchored> 497 498Which substring to check first. 499 500=item C<minlen> 501 502The minimal length of the match. 503 504=item C<stclass> I<TYPE> 505 506Type of first matching node. 507 508=item C<noscan> 509 510Don't scan for the found substrings. 511 512=item C<isall> 513 514Means that the optimizer information is all that the regular 515expression contains, and thus one does not need to enter the regex engine at 516all. 517 518=item C<GPOS> 519 520Set if the pattern contains C<\G>. 521 522=item C<plus> 523 524Set if the pattern starts with a repeated char (as in C<x+y>). 525 526=item C<implicit> 527 528Set if the pattern starts with C<.*>. 529 530=item C<with eval> 531 532Set if the pattern contain eval-groups, such as C<(?{ code })> and 533C<(??{ code })>. 534 535=item C<anchored(TYPE)> 536 537If the pattern may match only at a handful of places, with C<TYPE> 538being C<SBOL>, C<MBOL>, or C<GPOS>. See the table below. 539 540=back 541 542If a substring is known to match at end-of-line only, it may be 543followed by C<$>, as in C<floating 'k'$>. 544 545The optimizer-specific information is used to avoid entering (a slow) regex 546engine on strings that will not definitely match. If the C<isall> flag 547is set, a call to the regex engine may be avoided even when the optimizer 548found an appropriate place for the match. 549 550Above the optimizer section is the list of I<nodes> of the compiled 551form of the regex. Each line has format 552 553C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>) 554 555=head2 Types of Nodes 556 557Here are the current possible types, with short descriptions: 558 559=for comment 560This table is generated by regen/regcomp.pl. Any changes made here 561will be lost. 562 563=for regcomp.pl begin 564 565 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION 566 567 # Exit points 568 569 END no End of program. 570 SUCCEED no Return from a subroutine, basically. 571 572 # Line Start Anchors: 573 SBOL no Match "" at beginning of line: /^/, /\A/ 574 MBOL no Same, assuming multiline: /^/m 575 576 # Line End Anchors: 577 SEOL no Match "" at end of line: /$/ 578 MEOL no Same, assuming multiline: /$/m 579 EOS no Match "" at end of string: /\z/ 580 581 # Match Start Anchors: 582 GPOS no Matches where last m//g left off. 583 584 # Word Boundary Opcodes: 585 BOUND no Like BOUNDA for non-utf8, otherwise match 586 "" between any Unicode \w\W or \W\w 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 590 using /u rules. 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 match 594 "" between any Unicode \w\w or \W\W 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 600 \w is [_a-zA-Z0-9] 601 602 # [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 608 charclass 609 ANYOFL sv Like ANYOF, but /l is in effect 610 charclass 611 ANYOFPOSIXL sv Like ANYOFL, but matches [[:posix:]] 612 charclass_ classes 613 posixl 614 ANYOFH sv 1 Like ANYOF, but only has "High" matches, 615 none in the bitmap; non-zero flags "f" 616 means "f" is the first UTF-8 byte shared in 617 common by all code points matched 618 ANYOFM byte 1 Like ANYOF, but matches an invariant byte 619 as determined by the mask and arg 620 NANYOFM byte 1 complement of ANYOFM 621 622 # POSIX Character Classes: 623 POSIXD none Some [[:class:]] under /d; the FLAGS field 624 gives which one 625 POSIXL none Some [[:class:]] under /l; the FLAGS field 626 gives which one 627 POSIXU none Some [[:class:]] under /u; the FLAGS field 628 gives which one 629 POSIXA none Some [[:class:]] under /a; the FLAGS field 630 gives which one 631 NPOSIXD none complement of POSIXD, [[:^class:]] 632 NPOSIXL none complement of POSIXL, [[:^class:]] 633 NPOSIXU none complement of POSIXU, [[:^class:]] 634 NPOSIXA none complement of POSIXA, [[:^class:]] 635 636 CLUMP no Match any extended grapheme cluster 637 sequence 638 639 # Alternation 640 641 # BRANCH The set of branches constituting a single choice are 642 # hooked together with their "next" pointers, since 643 # precedence prevents anything being concatenated to 644 # any individual branch. The "next" pointer of the last 645 # BRANCH in a choice points to the thing following the 646 # whole choice. This is also where the final "next" 647 # pointer of each individual branch points; each branch 648 # starts with the operand node of a BRANCH node. 649 # 650 BRANCH node Match this alternative, or the next... 651 652 # Literals 653 654 EXACT str Match this string (preceded by length). 655 EXACTL str Like EXACT, but /l is in effect (used so 656 locale-related warnings can be checked 657 for). 658 EXACTF str Match this string using /id rules (w/len); 659 (string not UTF-8, not guaranteed to be 660 folded). 661 EXACTFL str Match this string using /il rules (w/len); 662 (string not guaranteed to be folded). 663 EXACTFU str Match this string using /iu rules (w/len); 664 (string folded iff in UTF-8; non-UTF8 665 folded length <= unfolded). 666 EXACTFAA str Match this string using /iaa rules (w/len) 667 (string folded iff in UTF-8; non-UTF8 668 folded length <= unfolded). 669 670 EXACTFUP str Match this string using /iu rules (w/len); 671 (string not UTF-8, not guaranteed to be 672 folded; and its Problematic). 673 674 EXACTFLU8 str Like EXACTFU, but use /il, UTF-8, folded, 675 and everything in it is above 255. 676 EXACTFAA_NO_TRIE str Match this string using /iaa rules (w/len) 677 (string not UTF-8, not guaranteed to be 678 folded, not currently trie-able). 679 680 EXACT_ONLY8 str Like EXACT, but only UTF-8 encoded targets 681 can match 682 EXACTFU_ONLY8 str Like EXACTFU, but only UTF-8 encoded 683 targets can match 684 685 EXACTFU_S_EDGE str /di rules, but nothing in it precludes /ui, 686 except begins and/or ends with [Ss]; 687 (string not UTF-8; compile-time only). 688 689 # Do nothing types 690 691 NOTHING no Match empty string. 692 # A variant of above which delimits a group, thus stops optimizations 693 TAIL no Match empty string. Can jump here from 694 outside. 695 696 # Loops 697 698 # STAR,PLUS '?', and complex '*' and '+', are implemented as 699 # circular BRANCH structures. Simple cases 700 # (one character per match) are implemented with STAR 701 # and PLUS for speed and to minimize recursive plunges. 702 # 703 STAR node Match this (simple) thing 0 or more times. 704 PLUS node Match this (simple) thing 1 or more times. 705 706 CURLY sv 2 Match this simple thing {n,m} times. 707 CURLYN no 2 Capture next-after-this simple thing 708 CURLYM no 2 Capture this medium-complex thing {n,m} 709 times. 710 CURLYX sv 2 Match this complex thing {n,m} times. 711 712 # This terminator creates a loop structure for CURLYX 713 WHILEM no Do curly processing and see if rest 714 matches. 715 716 # Buffer related 717 718 # OPEN,CLOSE,GROUPP ...are numbered at compile time. 719 OPEN num 1 Mark this point in input as start of #n. 720 CLOSE num 1 Close corresponding OPEN of #n. 721 SROPEN none Same as OPEN, but for script run 722 SRCLOSE none Close preceding SROPEN 723 724 REF num 1 Match some already matched string 725 REFF num 1 Match already matched string, using /di 726 rules. 727 REFFL num 1 Match already matched string, using /li 728 rules. 729 REFFU num 1 Match already matched string, usng /ui. 730 REFFA num 1 Match already matched string, using /aai 731 rules. 732 733 # Named references. Code in regcomp.c assumes that these all are after 734 # the numbered references 735 NREF no-sv 1 Match some already matched string 736 NREFF no-sv 1 Match already matched string, using /di 737 rules. 738 NREFFL no-sv 1 Match already matched string, using /li 739 rules. 740 NREFFU num 1 Match already matched string, using /ui 741 rules. 742 NREFFA num 1 Match already matched string, using /aai 743 rules. 744 745 # Support for long RE 746 LONGJMP off 1 1 Jump far away. 747 BRANCHJ off 1 1 BRANCH with long offset. 748 749 # Special Case Regops 750 IFMATCH off 1 1 Succeeds if the following matches; non-zero 751 flags "f", next_off "o" means lookbehind 752 assertion starting "f..(f-o)" characters 753 before current 754 UNLESSM off 1 1 Fails if the following matches; non-zero 755 flags "f", next_off "o" means lookbehind 756 assertion starting "f..(f-o)" characters 757 before current 758 SUSPEND off 1 1 "Independent" sub-RE. 759 IFTHEN off 1 1 Switch, should be preceded by switcher. 760 GROUPP num 1 Whether the group matched. 761 762 # The heavy worker 763 764 EVAL evl/flags Execute some Perl code. 765 2L 766 767 # Modifiers 768 769 MINMOD no Next operator is not greedy. 770 LOGICAL no Next opcode should set the flag only. 771 772 # This is not used yet 773 RENUM off 1 1 Group with independently numbered parens. 774 775 # Trie Related 776 777 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants 778 # have inline charclass data (ascii only), the 'C' store it in the 779 # structure. 780 781 TRIE trie 1 Match many EXACT(F[ALU]?)? at once. 782 flags==type 783 TRIEC trie Same as TRIE, but with embedded charclass 784 charclass data 785 786 AHOCORASICK trie 1 Aho Corasick stclass. flags==type 787 AHOCORASICKC trie Same as AHOCORASICK, but with embedded 788 charclass charclass data 789 790 # Regex Subroutines 791 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2 792 793 # Special conditionals 794 NGROUPP no-sv 1 Whether the group matched. 795 INSUBP num 1 Whether we are in a specific recurse. 796 DEFINEP none 1 Never execute directly. 797 798 # Backtracking Verbs 799 ENDLIKE none Used only for the type field of verbs 800 OPFAIL no-sv 1 Same as (?!), but with verb arg 801 ACCEPT no-sv/num Accepts the current matched string, with 802 2L verbar 803 804 # Verbs With Arguments 805 VERB no-sv 1 Used only for the type field of verbs 806 PRUNE no-sv 1 Pattern fails at this startpoint if no- 807 backtracking through this 808 MARKPOINT no-sv 1 Push the current location for rollback by 809 cut. 810 SKIP no-sv 1 On failure skip forward (to the mark) 811 before retrying 812 COMMIT no-sv 1 Pattern fails outright if backtracking 813 through this 814 CUTGROUP no-sv 1 On failure go to the next alternation in 815 the group 816 817 # Control what to keep in $&. 818 KEEPS no $& begins here. 819 820 # New charclass like patterns 821 LNBREAK none generic newline pattern 822 823 # SPECIAL REGOPS 824 825 # This is not really a node, but an optimized away piece of a "long" 826 # node. To simplify debugging output, we mark it as if it were a node 827 OPTIMIZED off Placeholder for dump. 828 829 # Special opcode with the property that no opcode in a compiled program 830 # will ever be of this type. Thus it can be used as a flag value that 831 # no other opcode has been seen. END is used similarly, in that an END 832 # node cant be optimized. So END implies "unoptimizable" and PSEUDO 833 # mean "not seen anything to optimize yet". 834 PSEUDO off Pseudo opcode for internal use. 835 836=for regcomp.pl end 837 838=for unprinted-credits 839Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421 840 841Following the optimizer information is a dump of the offset/length 842table, here split across several lines: 843 844 Offsets: [45] 845 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] 846 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] 847 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] 848 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] 849 850The first line here indicates that the offset/length table contains 45 851entries. Each entry is a pair of integers, denoted by C<offset[length]>. 852Entries are numbered starting with 1, so entry #1 here is C<1[4]> and 853entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:> 854(the C<1: ANYOF[bc]>) begins at character position 1 in the 855pre-compiled form of the regex, and has a length of 4 characters. 856C<5[1]> in position 12 857indicates that the node labeled C<12:> 858(the C<< 12: EXACT <d> >>) begins at character position 5 in the 859pre-compiled form of the regex, and has a length of 1 character. 860C<12[1]> in position 14 861indicates that the node labeled C<14:> 862(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the 863pre-compiled form of the regex, and has a length of 1 character---that 864is, it corresponds to the C<+> symbol in the precompiled regex. 865 866C<0[0]> items indicate that there is no corresponding node. 867 868=head2 Run-time Output 869 870First of all, when doing a match, one may get no run-time output even 871if debugging is enabled. This means that the regex engine was never 872entered and that all of the job was therefore done by the optimizer. 873 874If the regex engine was entered, the output may look like this: 875 876 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__' 877 Setting an EVAL scope, savestack=3 878 2 <ab> <cdefg__gh_> | 1: ANYOF 879 3 <abc> <defg__gh_> | 11: EXACT <d> 880 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767} 881 4 <abcd> <efg__gh_> | 26: WHILEM 882 0 out of 1..32767 cc=effff31c 883 4 <abcd> <efg__gh_> | 15: OPEN1 884 4 <abcd> <efg__gh_> | 17: EXACT <e> 885 5 <abcde> <fg__gh_> | 19: STAR 886 EXACT <f> can match 1 times out of 32767... 887 Setting an EVAL scope, savestack=3 888 6 <bcdef> <g__gh__> | 22: EXACT <g> 889 7 <bcdefg> <__gh__> | 24: CLOSE1 890 7 <bcdefg> <__gh__> | 26: WHILEM 891 1 out of 1..32767 cc=effff31c 892 Setting an EVAL scope, savestack=12 893 7 <bcdefg> <__gh__> | 15: OPEN1 894 7 <bcdefg> <__gh__> | 17: EXACT <e> 895 restoring \1 to 4(4)..7 896 failed, try continuation... 897 7 <bcdefg> <__gh__> | 27: NOTHING 898 7 <bcdefg> <__gh__> | 28: EXACT <h> 899 failed... 900 failed... 901 902The most significant information in the output is about the particular I<node> 903of the compiled regex that is currently being tested against the target string. 904The format of these lines is 905 906C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE> 907 908The I<TYPE> info is indented with respect to the backtracking level. 909Other incidental information appears interspersed within. 910 911=head1 Debugging Perl Memory Usage 912 913Perl is a profligate wastrel when it comes to memory use. There 914is a saying that to estimate memory usage of Perl, assume a reasonable 915algorithm for memory allocation, multiply that estimate by 10, and 916while you still may miss the mark, at least you won't be quite so 917astonished. This is not absolutely true, but may provide a good 918grasp of what happens. 919 920Assume that an integer cannot take less than 20 bytes of memory, a 921float cannot take less than 24 bytes, a string cannot take less 922than 32 bytes (all these examples assume 32-bit architectures, the 923result are quite a bit worse on 64-bit architectures). If a variable 924is accessed in two of three different ways (which require an integer, 925a float, or a string), the memory footprint may increase yet another 92620 bytes. A sloppy malloc(3) implementation can inflate these 927numbers dramatically. 928 929On the opposite end of the scale, a declaration like 930 931 sub foo; 932 933may take up to 500 bytes of memory, depending on which release of Perl 934you're running. 935 936Anecdotal estimates of source-to-compiled code bloat suggest an 937eightfold increase. This means that the compiled form of reasonable 938(normally commented, properly indented etc.) code will take 939about eight times more space in memory than the code took 940on disk. 941 942The B<-DL> command-line switch is obsolete since circa Perl 5.6.0 943(it was available only if Perl was built with C<-DDEBUGGING>). 944The switch was used to track Perl's memory allocations and possible 945memory leaks. These days the use of malloc debugging tools like 946F<Purify> or F<valgrind> is suggested instead. See also 947L<perlhacktips/PERL_MEM_LOG>. 948 949One way to find out how much memory is being used by Perl data 950structures is to install the Devel::Size module from CPAN: it gives 951you the minimum number of bytes required to store a particular data 952structure. Please be mindful of the difference between the size() 953and total_size(). 954 955If Perl has been compiled using Perl's malloc you can analyze Perl 956memory usage by setting $ENV{PERL_DEBUG_MSTATS}. 957 958=head2 Using C<$ENV{PERL_DEBUG_MSTATS}> 959 960If your perl is using Perl's malloc() and was compiled with the 961necessary switches (this is the default), then it will print memory 962usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS} 963> 1 >>, and before termination of the program when C<< 964$ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to 965the following example: 966 967 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp" 968 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192) 969 14216 free: 130 117 28 7 9 0 2 2 1 0 0 970 437 61 36 0 5 971 60924 used: 125 137 161 55 7 8 6 16 2 0 1 972 74 109 304 84 20 973 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048. 974 Memory allocation statistics after execution: (buckets 4(4)..8188(8192) 975 30888 free: 245 78 85 13 6 2 1 3 2 0 1 976 315 162 39 42 11 977 175816 used: 265 176 1112 111 26 22 11 27 2 1 1 978 196 178 1066 798 39 979 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144. 980 981It is possible to ask for such a statistic at arbitrary points in 982your execution using the mstat() function out of the standard 983Devel::Peek module. 984 985Here is some explanation of that format: 986 987=over 4 988 989=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)> 990 991Perl's malloc() uses bucketed allocations. Every request is rounded 992up to the closest bucket size available, and a bucket is taken from 993the pool of buckets of that size. 994 995The line above describes the limits of buckets currently in use. 996Each bucket has two sizes: memory footprint and the maximal size 997of user data that can fit into this bucket. Suppose in the above 998example that the smallest bucket were size 4. The biggest bucket 999would have usable size 8188, and the memory footprint would be 8192. 1000 1001In a Perl built for debugging, some buckets may have negative usable 1002size. This means that these buckets cannot (and will not) be used. 1003For larger buckets, the memory footprint may be one page greater 1004than a power of 2. If so, the corresponding power of two is 1005printed in the C<APPROX> field above. 1006 1007=item Free/Used 1008 1009The 1 or 2 rows of numbers following that correspond to the number 1010of buckets of each size between C<SMALLEST> and C<GREATEST>. In 1011the first row, the sizes (memory footprints) of buckets are powers 1012of two--or possibly one page greater. In the second row, if present, 1013the memory footprints of the buckets are between the memory footprints 1014of two buckets "above". 1015 1016For example, suppose under the previous example, the memory footprints 1017were 1018 1019 free: 8 16 32 64 128 256 512 1024 2048 4096 8192 1020 4 12 24 48 80 1021 1022With a non-C<DEBUGGING> perl, the buckets starting from C<128> have 1023a 4-byte overhead, and thus an 8192-long bucket may take up to 10248188-byte allocations. 1025 1026=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS> 1027 1028The first two fields give the total amount of memory perl sbrk(2)ed 1029(ess-broken? :-) and number of sbrk(2)s used. The third number is 1030what perl thinks about continuity of returned chunks. So long as 1031this number is positive, malloc() will assume that it is probable 1032that sbrk(2) will provide continuous memory. 1033 1034Memory allocated by external libraries is not counted. 1035 1036=item C<pad: 0> 1037 1038The amount of sbrk(2)ed memory needed to keep buckets aligned. 1039 1040=item C<heads: 2192> 1041 1042Although memory overhead of bigger buckets is kept inside the bucket, for 1043smaller buckets, it is kept in separate areas. This field gives the 1044total size of these areas. 1045 1046=item C<chain: 0> 1047 1048malloc() may want to subdivide a bigger bucket into smaller buckets. 1049If only a part of the deceased bucket is left unsubdivided, the rest 1050is kept as an element of a linked list. This field gives the total 1051size of these chunks. 1052 1053=item C<tail: 6144> 1054 1055To minimize the number of sbrk(2)s, malloc() asks for more memory. This 1056field gives the size of the yet unused part, which is sbrk(2)ed, but 1057never touched. 1058 1059=back 1060 1061=head1 SEE ALSO 1062 1063L<perldebug>, 1064L<perlguts>, 1065L<perlrun> 1066L<re>, 1067and 1068L<Devel::DProf>. 1069