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> holding the 98name of the called subroutine. (This doesn't happen if the subroutine 99was compiled in the C<DB> package.) 100 101X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub> 102is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling 103back to C<&DB::sub>(I<args>). 104 105=item * 106 107When execution of the program uses C<goto> to enter a non-XS 108subroutine and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto> 109is made, with C<$DB::sub> holding the name of the subroutine being 110entered. 111 112=back 113 114Note that if C<&DB::sub> needs external data for it to work, no 115subroutine call is possible without it. As an example, the standard 116debugger's C<&DB::sub> depends on the C<$DB::deep> variable 117(it defines how many levels of recursion deep into the debugger you can go 118before a mandatory break). If C<$DB::deep> is not defined, subroutine 119calls are not possible, even though C<&DB::sub> exists. 120 121=head2 Writing Your Own Debugger 122 123=head3 Environment Variables 124 125The C<PERL5DB> environment variable can be used to define a debugger. 126For example, the minimal "working" debugger (it actually doesn't do anything) 127consists of one line: 128 129 sub DB::DB {} 130 131It can easily be defined like this: 132 133 $ PERL5DB="sub DB::DB {}" perl -d your-script 134 135Another brief debugger, slightly more useful, can be created 136with only the line: 137 138 sub DB::DB {print ++$i; scalar <STDIN>} 139 140This debugger prints a number which increments for each statement 141encountered and waits for you to hit a newline before continuing 142to the next statement. 143 144The following debugger is actually useful: 145 146 { 147 package DB; 148 sub DB {} 149 sub sub {print ++$i, " $sub\n"; &$sub} 150 } 151 152It prints the sequence number of each subroutine call and the name of the 153called subroutine. Note that C<&DB::sub> is being compiled into the 154package C<DB> through the use of the C<package> directive. 155 156When it starts, the debugger reads your rc file (F<./.perldb> or 157F<~/.perldb> under Unix), which can set important options. 158(A subroutine (C<&afterinit>) can be defined here as well; it is executed 159after the debugger completes its own initialization.) 160 161After the rc file is read, the debugger reads the PERLDB_OPTS 162environment variable and uses it to set debugger options. The 163contents of this variable are treated as if they were the argument 164of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">). 165 166=head3 Debugger Internal Variables 167 168In addition to the file and subroutine-related variables mentioned above, 169the debugger also maintains various magical internal variables. 170 171=over 4 172 173=item * 174 175C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which 176holds the lines of the currently-selected file (compiled by Perl), either 177explicitly chosen with the debugger's C<f> command, or implicitly by flow 178of execution. 179 180Values in this array are magical in numeric context: they compare 181equal to zero only if the line is not breakable. 182 183=item * 184 185C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which 186contains breakpoints and actions keyed by line number in 187the currently-selected file, either explicitly chosen with the 188debugger's C<f> command, or implicitly by flow of execution. 189 190As previously noted, individual entries (as opposed to the whole hash) 191are settable. Perl only cares about Boolean true here, although 192the values used by F<perl5db.pl> have the form 193C<"$break_condition\0$action">. 194 195=back 196 197=head3 Debugger Customization Functions 198 199Some functions are provided to simplify customization. 200 201=over 4 202 203=item * 204 205See L<perldebug/"Configurable Options"> for a description of options parsed by 206C<DB::parse_options(string)>. 207 208=item * 209 210C<DB::dump_trace(skip[,count])> skips the specified number of frames 211and returns a list containing information about the calling frames (all 212of them, if C<count> is missing). Each entry is reference to a hash 213with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine 214name, or info about C<eval>), C<args> (C<undef> or a reference to 215an array), C<file>, and C<line>. 216 217=item * 218 219C<DB::print_trace(FH, skip[, count[, short]])> prints 220formatted info about caller frames. The last two functions may be 221convenient as arguments to C<< < >>, C<< << >> commands. 222 223=back 224 225Note that any variables and functions that are not documented in 226this manpages (or in L<perldebug>) are considered for internal 227use only, and as such are subject to change without notice. 228 229=head1 Frame Listing Output Examples 230 231The C<frame> option can be used to control the output of frame 232information. For example, contrast this expression trace: 233 234 $ perl -de 42 235 Stack dump during die enabled outside of evals. 236 237 Loading DB routines from perl5db.pl patch level 0.94 238 Emacs support available. 239 240 Enter h or 'h h' for help. 241 242 main::(-e:1): 0 243 DB<1> sub foo { 14 } 244 245 DB<2> sub bar { 3 } 246 247 DB<3> t print foo() * bar() 248 main::((eval 172):3): print foo() + bar(); 249 main::foo((eval 168):2): 250 main::bar((eval 170):2): 251 42 252 253with this one, once the C<o>ption C<frame=2> has been set: 254 255 DB<4> o f=2 256 frame = '2' 257 DB<5> t print foo() * bar() 258 3: foo() * bar() 259 entering main::foo 260 2: sub foo { 14 }; 261 exited main::foo 262 entering main::bar 263 2: sub bar { 3 }; 264 exited main::bar 265 42 266 267By way of demonstration, we present below a laborious listing 268resulting from setting your C<PERLDB_OPTS> environment variable to 269the value C<f=n N>, and running I<perl -d -V> from the command line. 270Examples using various values of C<n> are shown to give you a feel 271for the difference between settings. Long though it may be, this 272is not a complete listing, but only excerpts. 273 274=over 4 275 276=item 1 277 278 entering main::BEGIN 279 entering Config::BEGIN 280 Package lib/Exporter.pm. 281 Package lib/Carp.pm. 282 Package lib/Config.pm. 283 entering Config::TIEHASH 284 entering Exporter::import 285 entering Exporter::export 286 entering Config::myconfig 287 entering Config::FETCH 288 entering Config::FETCH 289 entering Config::FETCH 290 entering Config::FETCH 291 292=item 2 293 294 entering main::BEGIN 295 entering Config::BEGIN 296 Package lib/Exporter.pm. 297 Package lib/Carp.pm. 298 exited Config::BEGIN 299 Package lib/Config.pm. 300 entering Config::TIEHASH 301 exited Config::TIEHASH 302 entering Exporter::import 303 entering Exporter::export 304 exited Exporter::export 305 exited Exporter::import 306 exited main::BEGIN 307 entering Config::myconfig 308 entering Config::FETCH 309 exited Config::FETCH 310 entering Config::FETCH 311 exited Config::FETCH 312 entering Config::FETCH 313 314=item 3 315 316 in $=main::BEGIN() from /dev/null:0 317 in $=Config::BEGIN() from lib/Config.pm:2 318 Package lib/Exporter.pm. 319 Package lib/Carp.pm. 320 Package lib/Config.pm. 321 in $=Config::TIEHASH('Config') from lib/Config.pm:644 322 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 323 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li 324 in @=Config::myconfig() from /dev/null:0 325 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 326 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 327 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 328 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 329 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574 330 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574 331 332=item 4 333 334 in $=main::BEGIN() from /dev/null:0 335 in $=Config::BEGIN() from lib/Config.pm:2 336 Package lib/Exporter.pm. 337 Package lib/Carp.pm. 338 out $=Config::BEGIN() from lib/Config.pm:0 339 Package lib/Config.pm. 340 in $=Config::TIEHASH('Config') from lib/Config.pm:644 341 out $=Config::TIEHASH('Config') from lib/Config.pm:644 342 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 343 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/ 344 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/ 345 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 346 out $=main::BEGIN() from /dev/null:0 347 in @=Config::myconfig() from /dev/null:0 348 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 349 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 350 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 351 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 352 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 353 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 354 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 355 356=item 5 357 358 in $=main::BEGIN() from /dev/null:0 359 in $=Config::BEGIN() from lib/Config.pm:2 360 Package lib/Exporter.pm. 361 Package lib/Carp.pm. 362 out $=Config::BEGIN() from lib/Config.pm:0 363 Package lib/Config.pm. 364 in $=Config::TIEHASH('Config') from lib/Config.pm:644 365 out $=Config::TIEHASH('Config') from lib/Config.pm:644 366 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 367 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E 368 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E 369 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 370 out $=main::BEGIN() from /dev/null:0 371 in @=Config::myconfig() from /dev/null:0 372 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 373 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 374 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 375 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 376 377=item 6 378 379 in $=CODE(0x15eca4)() from /dev/null:0 380 in $=CODE(0x182528)() from lib/Config.pm:2 381 Package lib/Exporter.pm. 382 out $=CODE(0x182528)() from lib/Config.pm:0 383 scalar context return from CODE(0x182528): undef 384 Package lib/Config.pm. 385 in $=Config::TIEHASH('Config') from lib/Config.pm:628 386 out $=Config::TIEHASH('Config') from lib/Config.pm:628 387 scalar context return from Config::TIEHASH: empty hash 388 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 389 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 390 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 391 scalar context return from Exporter::export: '' 392 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 393 scalar context return from Exporter::import: '' 394 395=back 396 397In all cases shown above, the line indentation shows the call tree. 398If bit 2 of C<frame> is set, a line is printed on exit from a 399subroutine as well. If bit 4 is set, the arguments are printed 400along with the caller info. If bit 8 is set, the arguments are 401printed even if they are tied or references. If bit 16 is set, the 402return value is printed, too. 403 404When a package is compiled, a line like this 405 406 Package lib/Carp.pm. 407 408is printed with proper indentation. 409 410=head1 Debugging Regular Expressions 411 412There are two ways to enable debugging output for regular expressions. 413 414If your perl is compiled with C<-DDEBUGGING>, you may use the 415B<-Dr> flag on the command line. 416 417Otherwise, one can C<use re 'debug'>, which has effects at 418compile time and run time. Since Perl 5.9.5, this pragma is lexically 419scoped. 420 421=head2 Compile-time Output 422 423The debugging output at compile time looks like this: 424 425 Compiling REx '[bc]d(ef*g)+h[ij]k$' 426 size 45 Got 364 bytes for offset annotations. 427 first at 1 428 rarest char g at 0 429 rarest char d at 0 430 1: ANYOF[bc](12) 431 12: EXACT <d>(14) 432 14: CURLYX[0] {1,32767}(28) 433 16: OPEN1(18) 434 18: EXACT <e>(20) 435 20: STAR(23) 436 21: EXACT <f>(0) 437 23: EXACT <g>(25) 438 25: CLOSE1(27) 439 27: WHILEM[1/1](0) 440 28: NOTHING(29) 441 29: EXACT <h>(31) 442 31: ANYOF[ij](42) 443 42: EXACT <k>(44) 444 44: EOL(45) 445 45: END(0) 446 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 447 stclass 'ANYOF[bc]' minlen 7 448 Offsets: [45] 449 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] 450 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] 451 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] 452 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] 453 Omitting $` $& $' support. 454 455The first line shows the pre-compiled form of the regex. The second 456shows the size of the compiled form (in arbitrary units, usually 4574-byte words) and the total number of bytes allocated for the 458offset/length table, usually 4+C<size>*8. The next line shows the 459label I<id> of the first node that does a match. 460 461The 462 463 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 464 stclass 'ANYOF[bc]' minlen 7 465 466line (split into two lines above) contains optimizer 467information. In the example shown, the optimizer found that the match 468should contain a substring C<de> at offset 1, plus substring C<gh> 469at some offset between 3 and infinity. Moreover, when checking for 470these substrings (to abandon impossible matches quickly), Perl will check 471for the substring C<gh> before checking for the substring C<de>. The 472optimizer may also use the knowledge that the match starts (at the 473C<first> I<id>) with a character class, and no string 474shorter than 7 characters can possibly match. 475 476The fields of interest which may appear in this line are 477 478=over 4 479 480=item C<anchored> I<STRING> C<at> I<POS> 481 482=item C<floating> I<STRING> C<at> I<POS1..POS2> 483 484See above. 485 486=item C<matching floating/anchored> 487 488Which substring to check first. 489 490=item C<minlen> 491 492The minimal length of the match. 493 494=item C<stclass> I<TYPE> 495 496Type of first matching node. 497 498=item C<noscan> 499 500Don't scan for the found substrings. 501 502=item C<isall> 503 504Means that the optimizer information is all that the regular 505expression contains, and thus one does not need to enter the regex engine at 506all. 507 508=item C<GPOS> 509 510Set if the pattern contains C<\G>. 511 512=item C<plus> 513 514Set if the pattern starts with a repeated char (as in C<x+y>). 515 516=item C<implicit> 517 518Set if the pattern starts with C<.*>. 519 520=item C<with eval> 521 522Set if the pattern contain eval-groups, such as C<(?{ code })> and 523C<(??{ code })>. 524 525=item C<anchored(TYPE)> 526 527If the pattern may match only at a handful of places, with C<TYPE> 528being C<BOL>, C<MBOL>, or C<GPOS>. See the table below. 529 530=back 531 532If a substring is known to match at end-of-line only, it may be 533followed by C<$>, as in C<floating 'k'$>. 534 535The optimizer-specific information is used to avoid entering (a slow) regex 536engine on strings that will not definitely match. If the C<isall> flag 537is set, a call to the regex engine may be avoided even when the optimizer 538found an appropriate place for the match. 539 540Above the optimizer section is the list of I<nodes> of the compiled 541form of the regex. Each line has format 542 543C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>) 544 545=head2 Types of Nodes 546 547Here are the current possible types, with short descriptions: 548 549=for comment 550This table is generated by regen/regcomp.pl. Any changes made here 551will be lost. 552 553=for regcomp.pl begin 554 555 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION 556 557 # Exit points 558 559 END no End of program. 560 SUCCEED no Return from a subroutine, basically. 561 562 # Anchors: 563 564 BOL no Match "" at beginning of line. 565 MBOL no Same, assuming multiline. 566 SBOL no Same, assuming singleline. 567 EOS no Match "" at end of string. 568 EOL no Match "" at end of line. 569 MEOL no Same, assuming multiline. 570 SEOL no Same, assuming singleline. 571 BOUND no Match "" at any word boundary using native 572 charset rules for non-utf8 573 BOUNDL no Match "" at any locale word boundary 574 BOUNDU no Match "" at any word boundary using Unicode 575 rules 576 BOUNDA no Match "" at any word boundary using ASCII 577 rules 578 NBOUND no Match "" at any word non-boundary using 579 native charset rules for non-utf8 580 NBOUNDL no Match "" at any locale word non-boundary 581 NBOUNDU no Match "" at any word non-boundary using 582 Unicode rules 583 NBOUNDA no Match "" at any word non-boundary using 584 ASCII rules 585 GPOS no Matches where last m//g left off. 586 587 # [Special] alternatives: 588 589 REG_ANY no Match any one character (except newline). 590 SANY no Match any one character. 591 CANY no Match any one byte. 592 ANYOF sv Match character in (or not in) this class, 593 single char match only 594 595 POSIXD none Some [[:class:]] under /d; the FLAGS field 596 gives which one 597 POSIXL none Some [[:class:]] under /l; the FLAGS field 598 gives which one 599 POSIXU none Some [[:class:]] under /u; the FLAGS field 600 gives which one 601 POSIXA none Some [[:class:]] under /a; the FLAGS field 602 gives which one 603 NPOSIXD none complement of POSIXD, [[:^class:]] 604 NPOSIXL none complement of POSIXL, [[:^class:]] 605 NPOSIXU none complement of POSIXU, [[:^class:]] 606 NPOSIXA none complement of POSIXA, [[:^class:]] 607 608 CLUMP no Match any extended grapheme cluster sequence 609 610 # Alternation 611 612 # BRANCH The set of branches constituting a single choice are 613 # hooked together with their "next" pointers, since 614 # precedence prevents anything being concatenated to 615 # any individual branch. The "next" pointer of the last 616 # BRANCH in a choice points to the thing following the 617 # whole choice. This is also where the final "next" 618 # pointer of each individual branch points; each branch 619 # starts with the operand node of a BRANCH node. 620 # 621 BRANCH node Match this alternative, or the next... 622 623 # Back pointer 624 625 # BACK Normal "next" pointers all implicitly point forward; 626 # BACK exists to make loop structures possible. 627 # not used 628 BACK no Match "", "next" ptr points backward. 629 630 # Literals 631 632 EXACT str Match this string (preceded by length). 633 EXACTF str Match this non-UTF-8 string (not guaranteed 634 to be folded) using /id rules (w/len). 635 EXACTFL str Match this string (not guaranteed to be 636 folded) using /il rules (w/len). 637 EXACTFU str Match this string (folded iff in UTF-8, 638 length in folding doesn't change if not in 639 UTF-8) using /iu rules (w/len). 640 EXACTFA str Match this string (not guaranteed to be 641 folded) using /iaa rules (w/len). 642 EXACTFU_SS str Match this string (folded iff in UTF-8, 643 length in folding may change even if not in 644 UTF-8) using /iu rules (w/len). 645 EXACTFA_NO_TRIE str Match this string (which is not trie-able; 646 not guaranteed to be folded) using /iaa 647 rules (w/len). 648 649 # Do nothing types 650 651 NOTHING no Match empty string. 652 # A variant of above which delimits a group, thus stops optimizations 653 TAIL no Match empty string. Can jump here from 654 outside. 655 656 # Loops 657 658 # STAR,PLUS '?', and complex '*' and '+', are implemented as 659 # circular BRANCH structures using BACK. Simple cases 660 # (one character per match) are implemented with STAR 661 # and PLUS for speed and to minimize recursive plunges. 662 # 663 STAR node Match this (simple) thing 0 or more times. 664 PLUS node Match this (simple) thing 1 or more times. 665 666 CURLY sv 2 Match this simple thing {n,m} times. 667 CURLYN no 2 Capture next-after-this simple thing 668 CURLYM no 2 Capture this medium-complex thing {n,m} 669 times. 670 CURLYX sv 2 Match this complex thing {n,m} times. 671 672 # This terminator creates a loop structure for CURLYX 673 WHILEM no Do curly processing and see if rest matches. 674 675 # Buffer related 676 677 # OPEN,CLOSE,GROUPP ...are numbered at compile time. 678 OPEN num 1 Mark this point in input as start of #n. 679 CLOSE num 1 Analogous to OPEN. 680 681 REF num 1 Match some already matched string 682 REFF num 1 Match already matched string, folded using 683 native charset rules for non-utf8 684 REFFL num 1 Match already matched string, folded in loc. 685 REFFU num 1 Match already matched string, folded using 686 unicode rules for non-utf8 687 REFFA num 1 Match already matched string, folded using 688 unicode rules for non-utf8, no mixing ASCII, 689 non-ASCII 690 691 # Named references. Code in regcomp.c assumes that these all are after 692 # the numbered references 693 NREF no-sv 1 Match some already matched string 694 NREFF no-sv 1 Match already matched string, folded using 695 native charset rules for non-utf8 696 NREFFL no-sv 1 Match already matched string, folded in loc. 697 NREFFU num 1 Match already matched string, folded using 698 unicode rules for non-utf8 699 NREFFA num 1 Match already matched string, folded using 700 unicode rules for non-utf8, no mixing ASCII, 701 non-ASCII 702 703 IFMATCH off 1 2 Succeeds if the following matches. 704 UNLESSM off 1 2 Fails if the following matches. 705 SUSPEND off 1 1 "Independent" sub-RE. 706 IFTHEN off 1 1 Switch, should be preceded by switcher. 707 GROUPP num 1 Whether the group matched. 708 709 # Support for long RE 710 711 LONGJMP off 1 1 Jump far away. 712 BRANCHJ off 1 1 BRANCH with long offset. 713 714 # The heavy worker 715 716 EVAL evl 1 Execute some Perl code. 717 718 # Modifiers 719 720 MINMOD no Next operator is not greedy. 721 LOGICAL no Next opcode should set the flag only. 722 723 # This is not used yet 724 RENUM off 1 1 Group with independently numbered parens. 725 726 # Trie Related 727 728 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants 729 # have inline charclass data (ascii only), the 'C' store it in the 730 # structure. 731 732 TRIE trie 1 Match many EXACT(F[ALU]?)? at once. 733 flags==type 734 TRIEC trie Same as TRIE, but with embedded charclass 735 charclass data 736 737 AHOCORASICK trie 1 Aho Corasick stclass. flags==type 738 AHOCORASICKC trie Same as AHOCORASICK, but with embedded 739 charclass charclass data 740 741 # Regex Subroutines 742 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2 743 GOSTART no recurse to start of pattern 744 745 # Special conditionals 746 NGROUPP no-sv 1 Whether the group matched. 747 INSUBP num 1 Whether we are in a specific recurse. 748 DEFINEP none 1 Never execute directly. 749 750 # Backtracking Verbs 751 ENDLIKE none Used only for the type field of verbs 752 OPFAIL none Same as (?!) 753 ACCEPT parno 1 Accepts the current matched string. 754 755 # Verbs With Arguments 756 VERB no-sv 1 Used only for the type field of verbs 757 PRUNE no-sv 1 Pattern fails at this startpoint if no- 758 backtracking through this 759 MARKPOINT no-sv 1 Push the current location for rollback by 760 cut. 761 SKIP no-sv 1 On failure skip forward (to the mark) before 762 retrying 763 COMMIT no-sv 1 Pattern fails outright if backtracking 764 through this 765 CUTGROUP no-sv 1 On failure go to the next alternation in the 766 group 767 768 # Control what to keep in $&. 769 KEEPS no $& begins here. 770 771 # New charclass like patterns 772 LNBREAK none generic newline pattern 773 774 # SPECIAL REGOPS 775 776 # This is not really a node, but an optimized away piece of a "long" 777 # node. To simplify debugging output, we mark it as if it were a node 778 OPTIMIZED off Placeholder for dump. 779 780 # Special opcode with the property that no opcode in a compiled program 781 # will ever be of this type. Thus it can be used as a flag value that 782 # no other opcode has been seen. END is used similarly, in that an END 783 # node cant be optimized. So END implies "unoptimizable" and PSEUDO 784 # mean "not seen anything to optimize yet". 785 PSEUDO off Pseudo opcode for internal use. 786 787=for regcomp.pl end 788 789=for unprinted-credits 790Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421 791 792Following the optimizer information is a dump of the offset/length 793table, here split across several lines: 794 795 Offsets: [45] 796 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] 797 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] 798 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] 799 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] 800 801The first line here indicates that the offset/length table contains 45 802entries. Each entry is a pair of integers, denoted by C<offset[length]>. 803Entries are numbered starting with 1, so entry #1 here is C<1[4]> and 804entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:> 805(the C<1: ANYOF[bc]>) begins at character position 1 in the 806pre-compiled form of the regex, and has a length of 4 characters. 807C<5[1]> in position 12 808indicates that the node labeled C<12:> 809(the C<< 12: EXACT <d> >>) begins at character position 5 in the 810pre-compiled form of the regex, and has a length of 1 character. 811C<12[1]> in position 14 812indicates that the node labeled C<14:> 813(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the 814pre-compiled form of the regex, and has a length of 1 character---that 815is, it corresponds to the C<+> symbol in the precompiled regex. 816 817C<0[0]> items indicate that there is no corresponding node. 818 819=head2 Run-time Output 820 821First of all, when doing a match, one may get no run-time output even 822if debugging is enabled. This means that the regex engine was never 823entered and that all of the job was therefore done by the optimizer. 824 825If the regex engine was entered, the output may look like this: 826 827 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__' 828 Setting an EVAL scope, savestack=3 829 2 <ab> <cdefg__gh_> | 1: ANYOF 830 3 <abc> <defg__gh_> | 11: EXACT <d> 831 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767} 832 4 <abcd> <efg__gh_> | 26: WHILEM 833 0 out of 1..32767 cc=effff31c 834 4 <abcd> <efg__gh_> | 15: OPEN1 835 4 <abcd> <efg__gh_> | 17: EXACT <e> 836 5 <abcde> <fg__gh_> | 19: STAR 837 EXACT <f> can match 1 times out of 32767... 838 Setting an EVAL scope, savestack=3 839 6 <bcdef> <g__gh__> | 22: EXACT <g> 840 7 <bcdefg> <__gh__> | 24: CLOSE1 841 7 <bcdefg> <__gh__> | 26: WHILEM 842 1 out of 1..32767 cc=effff31c 843 Setting an EVAL scope, savestack=12 844 7 <bcdefg> <__gh__> | 15: OPEN1 845 7 <bcdefg> <__gh__> | 17: EXACT <e> 846 restoring \1 to 4(4)..7 847 failed, try continuation... 848 7 <bcdefg> <__gh__> | 27: NOTHING 849 7 <bcdefg> <__gh__> | 28: EXACT <h> 850 failed... 851 failed... 852 853The most significant information in the output is about the particular I<node> 854of the compiled regex that is currently being tested against the target string. 855The format of these lines is 856 857C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE> 858 859The I<TYPE> info is indented with respect to the backtracking level. 860Other incidental information appears interspersed within. 861 862=head1 Debugging Perl Memory Usage 863 864Perl is a profligate wastrel when it comes to memory use. There 865is a saying that to estimate memory usage of Perl, assume a reasonable 866algorithm for memory allocation, multiply that estimate by 10, and 867while you still may miss the mark, at least you won't be quite so 868astonished. This is not absolutely true, but may provide a good 869grasp of what happens. 870 871Assume that an integer cannot take less than 20 bytes of memory, a 872float cannot take less than 24 bytes, a string cannot take less 873than 32 bytes (all these examples assume 32-bit architectures, the 874result are quite a bit worse on 64-bit architectures). If a variable 875is accessed in two of three different ways (which require an integer, 876a float, or a string), the memory footprint may increase yet another 87720 bytes. A sloppy malloc(3) implementation can inflate these 878numbers dramatically. 879 880On the opposite end of the scale, a declaration like 881 882 sub foo; 883 884may take up to 500 bytes of memory, depending on which release of Perl 885you're running. 886 887Anecdotal estimates of source-to-compiled code bloat suggest an 888eightfold increase. This means that the compiled form of reasonable 889(normally commented, properly indented etc.) code will take 890about eight times more space in memory than the code took 891on disk. 892 893The B<-DL> command-line switch is obsolete since circa Perl 5.6.0 894(it was available only if Perl was built with C<-DDEBUGGING>). 895The switch was used to track Perl's memory allocations and possible 896memory leaks. These days the use of malloc debugging tools like 897F<Purify> or F<valgrind> is suggested instead. See also 898L<perlhacktips/PERL_MEM_LOG>. 899 900One way to find out how much memory is being used by Perl data 901structures is to install the Devel::Size module from CPAN: it gives 902you the minimum number of bytes required to store a particular data 903structure. Please be mindful of the difference between the size() 904and total_size(). 905 906If Perl has been compiled using Perl's malloc you can analyze Perl 907memory usage by setting $ENV{PERL_DEBUG_MSTATS}. 908 909=head2 Using C<$ENV{PERL_DEBUG_MSTATS}> 910 911If your perl is using Perl's malloc() and was compiled with the 912necessary switches (this is the default), then it will print memory 913usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS} 914> 1 >>, and before termination of the program when C<< 915$ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to 916the following example: 917 918 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp" 919 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192) 920 14216 free: 130 117 28 7 9 0 2 2 1 0 0 921 437 61 36 0 5 922 60924 used: 125 137 161 55 7 8 6 16 2 0 1 923 74 109 304 84 20 924 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048. 925 Memory allocation statistics after execution: (buckets 4(4)..8188(8192) 926 30888 free: 245 78 85 13 6 2 1 3 2 0 1 927 315 162 39 42 11 928 175816 used: 265 176 1112 111 26 22 11 27 2 1 1 929 196 178 1066 798 39 930 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144. 931 932It is possible to ask for such a statistic at arbitrary points in 933your execution using the mstat() function out of the standard 934Devel::Peek module. 935 936Here is some explanation of that format: 937 938=over 4 939 940=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)> 941 942Perl's malloc() uses bucketed allocations. Every request is rounded 943up to the closest bucket size available, and a bucket is taken from 944the pool of buckets of that size. 945 946The line above describes the limits of buckets currently in use. 947Each bucket has two sizes: memory footprint and the maximal size 948of user data that can fit into this bucket. Suppose in the above 949example that the smallest bucket were size 4. The biggest bucket 950would have usable size 8188, and the memory footprint would be 8192. 951 952In a Perl built for debugging, some buckets may have negative usable 953size. This means that these buckets cannot (and will not) be used. 954For larger buckets, the memory footprint may be one page greater 955than a power of 2. If so, the corresponding power of two is 956printed in the C<APPROX> field above. 957 958=item Free/Used 959 960The 1 or 2 rows of numbers following that correspond to the number 961of buckets of each size between C<SMALLEST> and C<GREATEST>. In 962the first row, the sizes (memory footprints) of buckets are powers 963of two--or possibly one page greater. In the second row, if present, 964the memory footprints of the buckets are between the memory footprints 965of two buckets "above". 966 967For example, suppose under the previous example, the memory footprints 968were 969 970 free: 8 16 32 64 128 256 512 1024 2048 4096 8192 971 4 12 24 48 80 972 973With a non-C<DEBUGGING> perl, the buckets starting from C<128> have 974a 4-byte overhead, and thus an 8192-long bucket may take up to 9758188-byte allocations. 976 977=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS> 978 979The first two fields give the total amount of memory perl sbrk(2)ed 980(ess-broken? :-) and number of sbrk(2)s used. The third number is 981what perl thinks about continuity of returned chunks. So long as 982this number is positive, malloc() will assume that it is probable 983that sbrk(2) will provide continuous memory. 984 985Memory allocated by external libraries is not counted. 986 987=item C<pad: 0> 988 989The amount of sbrk(2)ed memory needed to keep buckets aligned. 990 991=item C<heads: 2192> 992 993Although memory overhead of bigger buckets is kept inside the bucket, for 994smaller buckets, it is kept in separate areas. This field gives the 995total size of these areas. 996 997=item C<chain: 0> 998 999malloc() may want to subdivide a bigger bucket into smaller buckets. 1000If only a part of the deceased bucket is left unsubdivided, the rest 1001is kept as an element of a linked list. This field gives the total 1002size of these chunks. 1003 1004=item C<tail: 6144> 1005 1006To minimize the number of sbrk(2)s, malloc() asks for more memory. This 1007field gives the size of the yet unused part, which is sbrk(2)ed, but 1008never touched. 1009 1010=back 1011 1012=head1 SEE ALSO 1013 1014L<perldebug>, 1015L<perlguts>, 1016L<perlrun> 1017L<re>, 1018and 1019L<Devel::DProf>. 1020