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 $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 402 scalar context return from Exporter::import: '' 403 404=back 405 406In all cases shown above, the line indentation shows the call tree. 407If bit 2 of C<frame> is set, a line is printed on exit from a 408subroutine as well. If bit 4 is set, the arguments are printed 409along with the caller info. If bit 8 is set, the arguments are 410printed even if they are tied or references. If bit 16 is set, the 411return value is printed, too. 412 413When a package is compiled, a line like this 414 415 Package lib/Carp.pm. 416 417is printed with proper indentation. 418 419=head1 Debugging Regular Expressions 420 421There are two ways to enable debugging output for regular expressions. 422 423If your perl is compiled with C<-DDEBUGGING>, you may use the 424B<-Dr> flag on the command line. 425 426Otherwise, one can C<use re 'debug'>, which has effects at 427compile time and run time. Since Perl 5.9.5, this pragma is lexically 428scoped. 429 430=head2 Compile-time Output 431 432The debugging output at compile time looks like this: 433 434 Compiling REx '[bc]d(ef*g)+h[ij]k$' 435 size 45 Got 364 bytes for offset annotations. 436 first at 1 437 rarest char g at 0 438 rarest char d at 0 439 1: ANYOF[bc](12) 440 12: EXACT <d>(14) 441 14: CURLYX[0] {1,32767}(28) 442 16: OPEN1(18) 443 18: EXACT <e>(20) 444 20: STAR(23) 445 21: EXACT <f>(0) 446 23: EXACT <g>(25) 447 25: CLOSE1(27) 448 27: WHILEM[1/1](0) 449 28: NOTHING(29) 450 29: EXACT <h>(31) 451 31: ANYOF[ij](42) 452 42: EXACT <k>(44) 453 44: EOL(45) 454 45: END(0) 455 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 456 stclass 'ANYOF[bc]' minlen 7 457 Offsets: [45] 458 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] 459 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] 460 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] 461 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] 462 Omitting $` $& $' support. 463 464The first line shows the pre-compiled form of the regex. The second 465shows the size of the compiled form (in arbitrary units, usually 4664-byte words) and the total number of bytes allocated for the 467offset/length table, usually 4+C<size>*8. The next line shows the 468label I<id> of the first node that does a match. 469 470The 471 472 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 473 stclass 'ANYOF[bc]' minlen 7 474 475line (split into two lines above) contains optimizer 476information. In the example shown, the optimizer found that the match 477should contain a substring C<de> at offset 1, plus substring C<gh> 478at some offset between 3 and infinity. Moreover, when checking for 479these substrings (to abandon impossible matches quickly), Perl will check 480for the substring C<gh> before checking for the substring C<de>. The 481optimizer may also use the knowledge that the match starts (at the 482C<first> I<id>) with a character class, and no string 483shorter than 7 characters can possibly match. 484 485The fields of interest which may appear in this line are 486 487=over 4 488 489=item C<anchored> I<STRING> C<at> I<POS> 490 491=item C<floating> I<STRING> C<at> I<POS1..POS2> 492 493See above. 494 495=item C<matching floating/anchored> 496 497Which substring to check first. 498 499=item C<minlen> 500 501The minimal length of the match. 502 503=item C<stclass> I<TYPE> 504 505Type of first matching node. 506 507=item C<noscan> 508 509Don't scan for the found substrings. 510 511=item C<isall> 512 513Means that the optimizer information is all that the regular 514expression contains, and thus one does not need to enter the regex engine at 515all. 516 517=item C<GPOS> 518 519Set if the pattern contains C<\G>. 520 521=item C<plus> 522 523Set if the pattern starts with a repeated char (as in C<x+y>). 524 525=item C<implicit> 526 527Set if the pattern starts with C<.*>. 528 529=item C<with eval> 530 531Set if the pattern contain eval-groups, such as C<(?{ code })> and 532C<(??{ code })>. 533 534=item C<anchored(TYPE)> 535 536If the pattern may match only at a handful of places, with C<TYPE> 537being C<SBOL>, C<MBOL>, or C<GPOS>. See the table below. 538 539=back 540 541If a substring is known to match at end-of-line only, it may be 542followed by C<$>, as in C<floating 'k'$>. 543 544The optimizer-specific information is used to avoid entering (a slow) regex 545engine on strings that will not definitely match. If the C<isall> flag 546is set, a call to the regex engine may be avoided even when the optimizer 547found an appropriate place for the match. 548 549Above the optimizer section is the list of I<nodes> of the compiled 550form of the regex. Each line has format 551 552C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>) 553 554=head2 Types of Nodes 555 556Here are the current possible types, with short descriptions: 557 558=for comment 559This table is generated by regen/regcomp.pl. Any changes made here 560will be lost. 561 562=for regcomp.pl begin 563 564 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION 565 566 # Exit points 567 568 END no End of program. 569 SUCCEED no Return from a subroutine, basically. 570 571 # Line Start Anchors: 572 SBOL no Match "" at beginning of line: /^/, /\A/ 573 MBOL no Same, assuming multiline: /^/m 574 575 # Line End Anchors: 576 SEOL no Match "" at end of line: /$/ 577 MEOL no Same, assuming multiline: /$/m 578 EOS no Match "" at end of string: /\z/ 579 580 # Match Start Anchors: 581 GPOS no Matches where last m//g left off. 582 583 # Word Boundary Opcodes: 584 BOUND no Like BOUNDA for non-utf8, otherwise match 585 "" between any Unicode \w\W or \W\w 586 BOUNDL no Like BOUND/BOUNDU, but \w and \W are 587 defined by current locale 588 BOUNDU no Match "" at any boundary of a given type 589 using Unicode rules 590 BOUNDA no Match "" at any boundary between \w\W or 591 \W\w, where \w is [_a-zA-Z0-9] 592 NBOUND no Like NBOUNDA for non-utf8, otherwise match 593 "" between any Unicode \w\w or \W\W 594 NBOUNDL no Like NBOUND/NBOUNDU, but \w and \W are 595 defined by current locale 596 NBOUNDU no Match "" at any non-boundary of a given 597 type using using Unicode rules 598 NBOUNDA no Match "" betweeen any \w\w or \W\W, where 599 \w is [_a-zA-Z0-9] 600 601 # [Special] alternatives: 602 REG_ANY no Match any one character (except newline). 603 SANY no Match any one character. 604 ANYOF sv 1 Match character in (or not in) this class, 605 single char match only 606 ANYOFD sv 1 Like ANYOF, but /d is in effect 607 ANYOFL sv 1 Like ANYOF, but /l is in effect 608 ANYOFM byte 1 Like ANYOF, but matches an invariant byte 609 as determined by the mask and arg 610 611 # POSIX Character Classes: 612 POSIXD none Some [[:class:]] under /d; the FLAGS field 613 gives which one 614 POSIXL none Some [[:class:]] under /l; the FLAGS field 615 gives which one 616 POSIXU none Some [[:class:]] under /u; the FLAGS field 617 gives which one 618 POSIXA none Some [[:class:]] under /a; the FLAGS field 619 gives which one 620 NPOSIXD none complement of POSIXD, [[:^class:]] 621 NPOSIXL none complement of POSIXL, [[:^class:]] 622 NPOSIXU none complement of POSIXU, [[:^class:]] 623 NPOSIXA none complement of POSIXA, [[:^class:]] 624 625 ASCII none [[:ascii:]] 626 NASCII none [[:^ascii:]] 627 628 CLUMP no Match any extended grapheme cluster 629 sequence 630 631 # Alternation 632 633 # BRANCH The set of branches constituting a single choice are 634 # hooked together with their "next" pointers, since 635 # precedence prevents anything being concatenated to 636 # any individual branch. The "next" pointer of the last 637 # BRANCH in a choice points to the thing following the 638 # whole choice. This is also where the final "next" 639 # pointer of each individual branch points; each branch 640 # starts with the operand node of a BRANCH node. 641 # 642 BRANCH node Match this alternative, or the next... 643 644 # Literals 645 646 EXACT str Match this string (preceded by length). 647 EXACTL str Like EXACT, but /l is in effect (used so 648 locale-related warnings can be checked 649 for). 650 EXACTF str Match this non-UTF-8 string (not guaranteed 651 to be folded) using /id rules (w/len). 652 EXACTFL str Match this string (not guaranteed to be 653 folded) using /il rules (w/len). 654 EXACTFU str Match this string (folded iff in UTF-8, 655 length in folding doesn't change if not in 656 UTF-8) using /iu rules (w/len). 657 EXACTFAA str Match this string (not guaranteed to be 658 folded) using /iaa rules (w/len). 659 660 EXACTFU_SS str Match this string (folded iff in UTF-8, 661 length in folding may change even if not in 662 UTF-8) using /iu rules (w/len). 663 EXACTFLU8 str Rare circumstances: like EXACTFU, but is 664 under /l, UTF-8, folded, and everything in 665 it is above 255. 666 EXACTFAA_NO_TRIE str Match this string (which is not trie-able; 667 not guaranteed to be folded) using /iaa 668 rules (w/len). 669 670 # Do nothing types 671 672 NOTHING no Match empty string. 673 # A variant of above which delimits a group, thus stops optimizations 674 TAIL no Match empty string. Can jump here from 675 outside. 676 677 # Loops 678 679 # STAR,PLUS '?', and complex '*' and '+', are implemented as 680 # circular BRANCH structures. Simple cases 681 # (one character per match) are implemented with STAR 682 # and PLUS for speed and to minimize recursive plunges. 683 # 684 STAR node Match this (simple) thing 0 or more times. 685 PLUS node Match this (simple) thing 1 or more times. 686 687 CURLY sv 2 Match this simple thing {n,m} times. 688 CURLYN no 2 Capture next-after-this simple thing 689 CURLYM no 2 Capture this medium-complex thing {n,m} 690 times. 691 CURLYX sv 2 Match this complex thing {n,m} times. 692 693 # This terminator creates a loop structure for CURLYX 694 WHILEM no Do curly processing and see if rest 695 matches. 696 697 # Buffer related 698 699 # OPEN,CLOSE,GROUPP ...are numbered at compile time. 700 OPEN num 1 Mark this point in input as start of #n. 701 CLOSE num 1 Close corresponding OPEN of #n. 702 SROPEN none Same as OPEN, but for script run 703 SRCLOSE none Close preceding SROPEN 704 705 REF num 1 Match some already matched string 706 REFF num 1 Match already matched string, folded using 707 native charset rules for non-utf8 708 REFFL num 1 Match already matched string, folded in 709 loc. 710 REFFU num 1 Match already matched string, folded using 711 unicode rules for non-utf8 712 REFFA num 1 Match already matched string, folded using 713 unicode rules for non-utf8, no mixing 714 ASCII, non-ASCII 715 716 # Named references. Code in regcomp.c assumes that these all are after 717 # the numbered references 718 NREF no-sv 1 Match some already matched string 719 NREFF no-sv 1 Match already matched string, folded using 720 native charset rules for non-utf8 721 NREFFL no-sv 1 Match already matched string, folded in 722 loc. 723 NREFFU num 1 Match already matched string, folded using 724 unicode rules for non-utf8 725 NREFFA num 1 Match already matched string, folded using 726 unicode rules for non-utf8, no mixing 727 ASCII, non-ASCII 728 729 # Support for long RE 730 LONGJMP off 1 1 Jump far away. 731 BRANCHJ off 1 1 BRANCH with long offset. 732 733 # Special Case Regops 734 IFMATCH off 1 2 Succeeds if the following matches. 735 UNLESSM off 1 2 Fails if the following matches. 736 SUSPEND off 1 1 "Independent" sub-RE. 737 IFTHEN off 1 1 Switch, should be preceded by switcher. 738 GROUPP num 1 Whether the group matched. 739 740 # The heavy worker 741 742 EVAL evl/flags Execute some Perl code. 743 2L 744 745 # Modifiers 746 747 MINMOD no Next operator is not greedy. 748 LOGICAL no Next opcode should set the flag only. 749 750 # This is not used yet 751 RENUM off 1 1 Group with independently numbered parens. 752 753 # Trie Related 754 755 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants 756 # have inline charclass data (ascii only), the 'C' store it in the 757 # structure. 758 759 TRIE trie 1 Match many EXACT(F[ALU]?)? at once. 760 flags==type 761 TRIEC trie Same as TRIE, but with embedded charclass 762 charclass data 763 764 AHOCORASICK trie 1 Aho Corasick stclass. flags==type 765 AHOCORASICKC trie Same as AHOCORASICK, but with embedded 766 charclass charclass data 767 768 # Regex Subroutines 769 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2 770 771 # Special conditionals 772 NGROUPP no-sv 1 Whether the group matched. 773 INSUBP num 1 Whether we are in a specific recurse. 774 DEFINEP none 1 Never execute directly. 775 776 # Backtracking Verbs 777 ENDLIKE none Used only for the type field of verbs 778 OPFAIL no-sv 1 Same as (?!), but with verb arg 779 ACCEPT no-sv/num Accepts the current matched string, with 780 2L verbar 781 782 # Verbs With Arguments 783 VERB no-sv 1 Used only for the type field of verbs 784 PRUNE no-sv 1 Pattern fails at this startpoint if no- 785 backtracking through this 786 MARKPOINT no-sv 1 Push the current location for rollback by 787 cut. 788 SKIP no-sv 1 On failure skip forward (to the mark) 789 before retrying 790 COMMIT no-sv 1 Pattern fails outright if backtracking 791 through this 792 CUTGROUP no-sv 1 On failure go to the next alternation in 793 the group 794 795 # Control what to keep in $&. 796 KEEPS no $& begins here. 797 798 # New charclass like patterns 799 LNBREAK none generic newline pattern 800 801 # SPECIAL REGOPS 802 803 # This is not really a node, but an optimized away piece of a "long" 804 # node. To simplify debugging output, we mark it as if it were a node 805 OPTIMIZED off Placeholder for dump. 806 807 # Special opcode with the property that no opcode in a compiled program 808 # will ever be of this type. Thus it can be used as a flag value that 809 # no other opcode has been seen. END is used similarly, in that an END 810 # node cant be optimized. So END implies "unoptimizable" and PSEUDO 811 # mean "not seen anything to optimize yet". 812 PSEUDO off Pseudo opcode for internal use. 813 814=for regcomp.pl end 815 816=for unprinted-credits 817Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421 818 819Following the optimizer information is a dump of the offset/length 820table, here split across several lines: 821 822 Offsets: [45] 823 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] 824 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] 825 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] 826 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] 827 828The first line here indicates that the offset/length table contains 45 829entries. Each entry is a pair of integers, denoted by C<offset[length]>. 830Entries are numbered starting with 1, so entry #1 here is C<1[4]> and 831entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:> 832(the C<1: ANYOF[bc]>) begins at character position 1 in the 833pre-compiled form of the regex, and has a length of 4 characters. 834C<5[1]> in position 12 835indicates that the node labeled C<12:> 836(the C<< 12: EXACT <d> >>) begins at character position 5 in the 837pre-compiled form of the regex, and has a length of 1 character. 838C<12[1]> in position 14 839indicates that the node labeled C<14:> 840(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the 841pre-compiled form of the regex, and has a length of 1 character---that 842is, it corresponds to the C<+> symbol in the precompiled regex. 843 844C<0[0]> items indicate that there is no corresponding node. 845 846=head2 Run-time Output 847 848First of all, when doing a match, one may get no run-time output even 849if debugging is enabled. This means that the regex engine was never 850entered and that all of the job was therefore done by the optimizer. 851 852If the regex engine was entered, the output may look like this: 853 854 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__' 855 Setting an EVAL scope, savestack=3 856 2 <ab> <cdefg__gh_> | 1: ANYOF 857 3 <abc> <defg__gh_> | 11: EXACT <d> 858 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767} 859 4 <abcd> <efg__gh_> | 26: WHILEM 860 0 out of 1..32767 cc=effff31c 861 4 <abcd> <efg__gh_> | 15: OPEN1 862 4 <abcd> <efg__gh_> | 17: EXACT <e> 863 5 <abcde> <fg__gh_> | 19: STAR 864 EXACT <f> can match 1 times out of 32767... 865 Setting an EVAL scope, savestack=3 866 6 <bcdef> <g__gh__> | 22: EXACT <g> 867 7 <bcdefg> <__gh__> | 24: CLOSE1 868 7 <bcdefg> <__gh__> | 26: WHILEM 869 1 out of 1..32767 cc=effff31c 870 Setting an EVAL scope, savestack=12 871 7 <bcdefg> <__gh__> | 15: OPEN1 872 7 <bcdefg> <__gh__> | 17: EXACT <e> 873 restoring \1 to 4(4)..7 874 failed, try continuation... 875 7 <bcdefg> <__gh__> | 27: NOTHING 876 7 <bcdefg> <__gh__> | 28: EXACT <h> 877 failed... 878 failed... 879 880The most significant information in the output is about the particular I<node> 881of the compiled regex that is currently being tested against the target string. 882The format of these lines is 883 884C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE> 885 886The I<TYPE> info is indented with respect to the backtracking level. 887Other incidental information appears interspersed within. 888 889=head1 Debugging Perl Memory Usage 890 891Perl is a profligate wastrel when it comes to memory use. There 892is a saying that to estimate memory usage of Perl, assume a reasonable 893algorithm for memory allocation, multiply that estimate by 10, and 894while you still may miss the mark, at least you won't be quite so 895astonished. This is not absolutely true, but may provide a good 896grasp of what happens. 897 898Assume that an integer cannot take less than 20 bytes of memory, a 899float cannot take less than 24 bytes, a string cannot take less 900than 32 bytes (all these examples assume 32-bit architectures, the 901result are quite a bit worse on 64-bit architectures). If a variable 902is accessed in two of three different ways (which require an integer, 903a float, or a string), the memory footprint may increase yet another 90420 bytes. A sloppy malloc(3) implementation can inflate these 905numbers dramatically. 906 907On the opposite end of the scale, a declaration like 908 909 sub foo; 910 911may take up to 500 bytes of memory, depending on which release of Perl 912you're running. 913 914Anecdotal estimates of source-to-compiled code bloat suggest an 915eightfold increase. This means that the compiled form of reasonable 916(normally commented, properly indented etc.) code will take 917about eight times more space in memory than the code took 918on disk. 919 920The B<-DL> command-line switch is obsolete since circa Perl 5.6.0 921(it was available only if Perl was built with C<-DDEBUGGING>). 922The switch was used to track Perl's memory allocations and possible 923memory leaks. These days the use of malloc debugging tools like 924F<Purify> or F<valgrind> is suggested instead. See also 925L<perlhacktips/PERL_MEM_LOG>. 926 927One way to find out how much memory is being used by Perl data 928structures is to install the Devel::Size module from CPAN: it gives 929you the minimum number of bytes required to store a particular data 930structure. Please be mindful of the difference between the size() 931and total_size(). 932 933If Perl has been compiled using Perl's malloc you can analyze Perl 934memory usage by setting $ENV{PERL_DEBUG_MSTATS}. 935 936=head2 Using C<$ENV{PERL_DEBUG_MSTATS}> 937 938If your perl is using Perl's malloc() and was compiled with the 939necessary switches (this is the default), then it will print memory 940usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS} 941> 1 >>, and before termination of the program when C<< 942$ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to 943the following example: 944 945 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp" 946 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192) 947 14216 free: 130 117 28 7 9 0 2 2 1 0 0 948 437 61 36 0 5 949 60924 used: 125 137 161 55 7 8 6 16 2 0 1 950 74 109 304 84 20 951 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048. 952 Memory allocation statistics after execution: (buckets 4(4)..8188(8192) 953 30888 free: 245 78 85 13 6 2 1 3 2 0 1 954 315 162 39 42 11 955 175816 used: 265 176 1112 111 26 22 11 27 2 1 1 956 196 178 1066 798 39 957 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144. 958 959It is possible to ask for such a statistic at arbitrary points in 960your execution using the mstat() function out of the standard 961Devel::Peek module. 962 963Here is some explanation of that format: 964 965=over 4 966 967=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)> 968 969Perl's malloc() uses bucketed allocations. Every request is rounded 970up to the closest bucket size available, and a bucket is taken from 971the pool of buckets of that size. 972 973The line above describes the limits of buckets currently in use. 974Each bucket has two sizes: memory footprint and the maximal size 975of user data that can fit into this bucket. Suppose in the above 976example that the smallest bucket were size 4. The biggest bucket 977would have usable size 8188, and the memory footprint would be 8192. 978 979In a Perl built for debugging, some buckets may have negative usable 980size. This means that these buckets cannot (and will not) be used. 981For larger buckets, the memory footprint may be one page greater 982than a power of 2. If so, the corresponding power of two is 983printed in the C<APPROX> field above. 984 985=item Free/Used 986 987The 1 or 2 rows of numbers following that correspond to the number 988of buckets of each size between C<SMALLEST> and C<GREATEST>. In 989the first row, the sizes (memory footprints) of buckets are powers 990of two--or possibly one page greater. In the second row, if present, 991the memory footprints of the buckets are between the memory footprints 992of two buckets "above". 993 994For example, suppose under the previous example, the memory footprints 995were 996 997 free: 8 16 32 64 128 256 512 1024 2048 4096 8192 998 4 12 24 48 80 999 1000With a non-C<DEBUGGING> perl, the buckets starting from C<128> have 1001a 4-byte overhead, and thus an 8192-long bucket may take up to 10028188-byte allocations. 1003 1004=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS> 1005 1006The first two fields give the total amount of memory perl sbrk(2)ed 1007(ess-broken? :-) and number of sbrk(2)s used. The third number is 1008what perl thinks about continuity of returned chunks. So long as 1009this number is positive, malloc() will assume that it is probable 1010that sbrk(2) will provide continuous memory. 1011 1012Memory allocated by external libraries is not counted. 1013 1014=item C<pad: 0> 1015 1016The amount of sbrk(2)ed memory needed to keep buckets aligned. 1017 1018=item C<heads: 2192> 1019 1020Although memory overhead of bigger buckets is kept inside the bucket, for 1021smaller buckets, it is kept in separate areas. This field gives the 1022total size of these areas. 1023 1024=item C<chain: 0> 1025 1026malloc() may want to subdivide a bigger bucket into smaller buckets. 1027If only a part of the deceased bucket is left unsubdivided, the rest 1028is kept as an element of a linked list. This field gives the total 1029size of these chunks. 1030 1031=item C<tail: 6144> 1032 1033To minimize the number of sbrk(2)s, malloc() asks for more memory. This 1034field gives the size of the yet unused part, which is sbrk(2)ed, but 1035never touched. 1036 1037=back 1038 1039=head1 SEE ALSO 1040 1041L<perldebug>, 1042L<perlguts>, 1043L<perlrun> 1044L<re>, 1045and 1046L<Devel::DProf>. 1047