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.) 100 101=back 102 103Note that if C<&DB::sub> needs external data for it to work, no 104subroutine call is possible without it. As an example, the standard 105debugger's C<&DB::sub> depends on the C<$DB::deep> variable 106(it defines how many levels of recursion deep into the debugger you can go 107before a mandatory break). If C<$DB::deep> is not defined, subroutine 108calls are not possible, even though C<&DB::sub> exists. 109 110=head2 Writing Your Own Debugger 111 112=head3 Environment Variables 113 114The C<PERL5DB> environment variable can be used to define a debugger. 115For example, the minimal "working" debugger (it actually doesn't do anything) 116consists of one line: 117 118 sub DB::DB {} 119 120It can easily be defined like this: 121 122 $ PERL5DB="sub DB::DB {}" perl -d your-script 123 124Another brief debugger, slightly more useful, can be created 125with only the line: 126 127 sub DB::DB {print ++$i; scalar <STDIN>} 128 129This debugger prints a number which increments for each statement 130encountered and waits for you to hit a newline before continuing 131to the next statement. 132 133The following debugger is actually useful: 134 135 { 136 package DB; 137 sub DB {} 138 sub sub {print ++$i, " $sub\n"; &$sub} 139 } 140 141It prints the sequence number of each subroutine call and the name of the 142called subroutine. Note that C<&DB::sub> is being compiled into the 143package C<DB> through the use of the C<package> directive. 144 145When it starts, the debugger reads your rc file (F<./.perldb> or 146F<~/.perldb> under Unix), which can set important options. 147(A subroutine (C<&afterinit>) can be defined here as well; it is executed 148after the debugger completes its own initialization.) 149 150After the rc file is read, the debugger reads the PERLDB_OPTS 151environment variable and uses it to set debugger options. The 152contents of this variable are treated as if they were the argument 153of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">). 154 155=head3 Debugger Internal Variables 156 157In addition to the file and subroutine-related variables mentioned above, 158the debugger also maintains various magical internal variables. 159 160=over 4 161 162=item * 163 164C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which 165holds the lines of the currently-selected file (compiled by Perl), either 166explicitly chosen with the debugger's C<f> command, or implicitly by flow 167of execution. 168 169Values in this array are magical in numeric context: they compare 170equal to zero only if the line is not breakable. 171 172=item * 173 174C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which 175contains breakpoints and actions keyed by line number in 176the currently-selected file, either explicitly chosen with the 177debugger's C<f> command, or implicitly by flow of execution. 178 179As previously noted, individual entries (as opposed to the whole hash) 180are settable. Perl only cares about Boolean true here, although 181the values used by F<perl5db.pl> have the form 182C<"$break_condition\0$action">. 183 184=back 185 186=head3 Debugger Customization Functions 187 188Some functions are provided to simplify customization. 189 190=over 4 191 192=item * 193 194See L<perldebug/"Configurable Options"> for a description of options parsed by 195C<DB::parse_options(string)>. 196 197=item * 198 199C<DB::dump_trace(skip[,count])> skips the specified number of frames 200and returns a list containing information about the calling frames (all 201of them, if C<count> is missing). Each entry is reference to a hash 202with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine 203name, or info about C<eval>), C<args> (C<undef> or a reference to 204an array), C<file>, and C<line>. 205 206=item * 207 208C<DB::print_trace(FH, skip[, count[, short]])> prints 209formatted info about caller frames. The last two functions may be 210convenient as arguments to C<< < >>, C<< << >> commands. 211 212=back 213 214Note that any variables and functions that are not documented in 215this manpages (or in L<perldebug>) are considered for internal 216use only, and as such are subject to change without notice. 217 218=head1 Frame Listing Output Examples 219 220The C<frame> option can be used to control the output of frame 221information. For example, contrast this expression trace: 222 223 $ perl -de 42 224 Stack dump during die enabled outside of evals. 225 226 Loading DB routines from perl5db.pl patch level 0.94 227 Emacs support available. 228 229 Enter h or 'h h' for help. 230 231 main::(-e:1): 0 232 DB<1> sub foo { 14 } 233 234 DB<2> sub bar { 3 } 235 236 DB<3> t print foo() * bar() 237 main::((eval 172):3): print foo() + bar(); 238 main::foo((eval 168):2): 239 main::bar((eval 170):2): 240 42 241 242with this one, once the C<o>ption C<frame=2> has been set: 243 244 DB<4> o f=2 245 frame = '2' 246 DB<5> t print foo() * bar() 247 3: foo() * bar() 248 entering main::foo 249 2: sub foo { 14 }; 250 exited main::foo 251 entering main::bar 252 2: sub bar { 3 }; 253 exited main::bar 254 42 255 256By way of demonstration, we present below a laborious listing 257resulting from setting your C<PERLDB_OPTS> environment variable to 258the value C<f=n N>, and running I<perl -d -V> from the command line. 259Examples using various values of C<n> are shown to give you a feel 260for the difference between settings. Long though it may be, this 261is not a complete listing, but only excerpts. 262 263=over 4 264 265=item 1 266 267 entering main::BEGIN 268 entering Config::BEGIN 269 Package lib/Exporter.pm. 270 Package lib/Carp.pm. 271 Package lib/Config.pm. 272 entering Config::TIEHASH 273 entering Exporter::import 274 entering Exporter::export 275 entering Config::myconfig 276 entering Config::FETCH 277 entering Config::FETCH 278 entering Config::FETCH 279 entering Config::FETCH 280 281=item 2 282 283 entering main::BEGIN 284 entering Config::BEGIN 285 Package lib/Exporter.pm. 286 Package lib/Carp.pm. 287 exited Config::BEGIN 288 Package lib/Config.pm. 289 entering Config::TIEHASH 290 exited Config::TIEHASH 291 entering Exporter::import 292 entering Exporter::export 293 exited Exporter::export 294 exited Exporter::import 295 exited main::BEGIN 296 entering Config::myconfig 297 entering Config::FETCH 298 exited Config::FETCH 299 entering Config::FETCH 300 exited Config::FETCH 301 entering Config::FETCH 302 303=item 3 304 305 in $=main::BEGIN() from /dev/null:0 306 in $=Config::BEGIN() from lib/Config.pm:2 307 Package lib/Exporter.pm. 308 Package lib/Carp.pm. 309 Package lib/Config.pm. 310 in $=Config::TIEHASH('Config') from lib/Config.pm:644 311 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 312 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li 313 in @=Config::myconfig() from /dev/null:0 314 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 315 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 316 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 317 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 318 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574 319 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574 320 321=item 4 322 323 in $=main::BEGIN() from /dev/null:0 324 in $=Config::BEGIN() from lib/Config.pm:2 325 Package lib/Exporter.pm. 326 Package lib/Carp.pm. 327 out $=Config::BEGIN() from lib/Config.pm:0 328 Package lib/Config.pm. 329 in $=Config::TIEHASH('Config') from lib/Config.pm:644 330 out $=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 lib/ 333 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/ 334 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 335 out $=main::BEGIN() from /dev/null:0 336 in @=Config::myconfig() from /dev/null:0 337 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 338 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 339 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 340 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 341 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 342 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 343 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 344 345=item 5 346 347 in $=main::BEGIN() from /dev/null:0 348 in $=Config::BEGIN() from lib/Config.pm:2 349 Package lib/Exporter.pm. 350 Package lib/Carp.pm. 351 out $=Config::BEGIN() from lib/Config.pm:0 352 Package lib/Config.pm. 353 in $=Config::TIEHASH('Config') from lib/Config.pm:644 354 out $=Config::TIEHASH('Config') from lib/Config.pm:644 355 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 356 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E 357 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E 358 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 359 out $=main::BEGIN() from /dev/null:0 360 in @=Config::myconfig() from /dev/null:0 361 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 362 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 363 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 364 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 365 366=item 6 367 368 in $=CODE(0x15eca4)() from /dev/null:0 369 in $=CODE(0x182528)() from lib/Config.pm:2 370 Package lib/Exporter.pm. 371 out $=CODE(0x182528)() from lib/Config.pm:0 372 scalar context return from CODE(0x182528): undef 373 Package lib/Config.pm. 374 in $=Config::TIEHASH('Config') from lib/Config.pm:628 375 out $=Config::TIEHASH('Config') from lib/Config.pm:628 376 scalar context return from Config::TIEHASH: empty hash 377 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 378 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 379 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 380 scalar context return from Exporter::export: '' 381 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 382 scalar context return from Exporter::import: '' 383 384=back 385 386In all cases shown above, the line indentation shows the call tree. 387If bit 2 of C<frame> is set, a line is printed on exit from a 388subroutine as well. If bit 4 is set, the arguments are printed 389along with the caller info. If bit 8 is set, the arguments are 390printed even if they are tied or references. If bit 16 is set, the 391return value is printed, too. 392 393When a package is compiled, a line like this 394 395 Package lib/Carp.pm. 396 397is printed with proper indentation. 398 399=head1 Debugging Regular Expressions 400 401There are two ways to enable debugging output for regular expressions. 402 403If your perl is compiled with C<-DDEBUGGING>, you may use the 404B<-Dr> flag on the command line. 405 406Otherwise, one can C<use re 'debug'>, which has effects at 407compile time and run time. Since Perl 5.9.5, this pragma is lexically 408scoped. 409 410=head2 Compile-time Output 411 412The debugging output at compile time looks like this: 413 414 Compiling REx '[bc]d(ef*g)+h[ij]k$' 415 size 45 Got 364 bytes for offset annotations. 416 first at 1 417 rarest char g at 0 418 rarest char d at 0 419 1: ANYOF[bc](12) 420 12: EXACT <d>(14) 421 14: CURLYX[0] {1,32767}(28) 422 16: OPEN1(18) 423 18: EXACT <e>(20) 424 20: STAR(23) 425 21: EXACT <f>(0) 426 23: EXACT <g>(25) 427 25: CLOSE1(27) 428 27: WHILEM[1/1](0) 429 28: NOTHING(29) 430 29: EXACT <h>(31) 431 31: ANYOF[ij](42) 432 42: EXACT <k>(44) 433 44: EOL(45) 434 45: END(0) 435 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 436 stclass 'ANYOF[bc]' minlen 7 437 Offsets: [45] 438 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] 439 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] 440 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] 441 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] 442 Omitting $` $& $' support. 443 444The first line shows the pre-compiled form of the regex. The second 445shows the size of the compiled form (in arbitrary units, usually 4464-byte words) and the total number of bytes allocated for the 447offset/length table, usually 4+C<size>*8. The next line shows the 448label I<id> of the first node that does a match. 449 450The 451 452 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) 453 stclass 'ANYOF[bc]' minlen 7 454 455line (split into two lines above) contains optimizer 456information. In the example shown, the optimizer found that the match 457should contain a substring C<de> at offset 1, plus substring C<gh> 458at some offset between 3 and infinity. Moreover, when checking for 459these substrings (to abandon impossible matches quickly), Perl will check 460for the substring C<gh> before checking for the substring C<de>. The 461optimizer may also use the knowledge that the match starts (at the 462C<first> I<id>) with a character class, and no string 463shorter than 7 characters can possibly match. 464 465The fields of interest which may appear in this line are 466 467=over 4 468 469=item C<anchored> I<STRING> C<at> I<POS> 470 471=item C<floating> I<STRING> C<at> I<POS1..POS2> 472 473See above. 474 475=item C<matching floating/anchored> 476 477Which substring to check first. 478 479=item C<minlen> 480 481The minimal length of the match. 482 483=item C<stclass> I<TYPE> 484 485Type of first matching node. 486 487=item C<noscan> 488 489Don't scan for the found substrings. 490 491=item C<isall> 492 493Means that the optimizer information is all that the regular 494expression contains, and thus one does not need to enter the regex engine at 495all. 496 497=item C<GPOS> 498 499Set if the pattern contains C<\G>. 500 501=item C<plus> 502 503Set if the pattern starts with a repeated char (as in C<x+y>). 504 505=item C<implicit> 506 507Set if the pattern starts with C<.*>. 508 509=item C<with eval> 510 511Set if the pattern contain eval-groups, such as C<(?{ code })> and 512C<(??{ code })>. 513 514=item C<anchored(TYPE)> 515 516If the pattern may match only at a handful of places, with C<TYPE> 517being C<BOL>, C<MBOL>, or C<GPOS>. See the table below. 518 519=back 520 521If a substring is known to match at end-of-line only, it may be 522followed by C<$>, as in C<floating 'k'$>. 523 524The optimizer-specific information is used to avoid entering (a slow) regex 525engine on strings that will not definitely match. If the C<isall> flag 526is set, a call to the regex engine may be avoided even when the optimizer 527found an appropriate place for the match. 528 529Above the optimizer section is the list of I<nodes> of the compiled 530form of the regex. Each line has format 531 532C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>) 533 534=head2 Types of Nodes 535 536Here are the possible types, with short descriptions: 537 538=for comment 539This table is generated by regen/regcomp.pl. Any changes made here 540will be lost. 541 542=for regcomp.pl begin 543 544 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION 545 546 # Exit points 547 548 END no End of program. 549 SUCCEED no Return from a subroutine, basically. 550 551 # Anchors: 552 553 BOL no Match "" at beginning of line. 554 MBOL no Same, assuming multiline. 555 SBOL no Same, assuming singleline. 556 EOS no Match "" at end of string. 557 EOL no Match "" at end of line. 558 MEOL no Same, assuming multiline. 559 SEOL no Same, assuming singleline. 560 BOUND no Match "" at any word boundary using 561 native charset semantics for non-utf8 562 BOUNDL no Match "" at any locale word boundary 563 BOUNDU no Match "" at any word boundary using 564 Unicode semantics 565 BOUNDA no Match "" at any word boundary using ASCII 566 semantics 567 NBOUND no Match "" at any word non-boundary using 568 native charset semantics for non-utf8 569 NBOUNDL no Match "" at any locale word non-boundary 570 NBOUNDU no Match "" at any word non-boundary using 571 Unicode semantics 572 NBOUNDA no Match "" at any word non-boundary using 573 ASCII semantics 574 GPOS no Matches where last m//g left off. 575 576 # [Special] alternatives: 577 578 REG_ANY no Match any one character (except newline). 579 SANY no Match any one character. 580 CANY no Match any one byte. 581 ANYOF sv Match character in (or not in) this 582 class, single char match only 583 ANYOF_WARN_SUPER sv Match character in (or not in) this 584 class, warn (if enabled) upon matching a 585 char above Unicode max; 586 ANYOF_SYNTHETIC sv Synthetic start class 587 588 POSIXD none Some [[:class:]] under /d; the FLAGS 589 field gives which one 590 POSIXL none Some [[:class:]] under /l; the FLAGS 591 field gives which one 592 POSIXU none Some [[:class:]] under /u; the FLAGS 593 field gives which one 594 POSIXA none Some [[:class:]] under /a; the FLAGS 595 field gives which one 596 NPOSIXD none complement of POSIXD, [[:^class:]] 597 NPOSIXL none complement of POSIXL, [[:^class:]] 598 NPOSIXU none complement of POSIXU, [[:^class:]] 599 NPOSIXA none complement of POSIXA, [[:^class:]] 600 601 CLUMP no Match any extended grapheme cluster 602 sequence 603 604 # Alternation 605 606 # BRANCH The set of branches constituting a single choice are 607 # hooked together with their "next" pointers, since 608 # precedence prevents anything being concatenated to 609 # any individual branch. The "next" pointer of the last 610 # BRANCH in a choice points to the thing following the 611 # whole choice. This is also where the final "next" 612 # pointer of each individual branch points; each branch 613 # starts with the operand node of a BRANCH node. 614 # 615 BRANCH node Match this alternative, or the next... 616 617 # Back pointer 618 619 # BACK Normal "next" pointers all implicitly point forward; 620 # BACK exists to make loop structures possible. 621 # not used 622 BACK no Match "", "next" ptr points backward. 623 624 # Literals 625 626 EXACT str Match this string (preceded by length). 627 EXACTF str Match this non-UTF-8 string (not 628 guaranteed to be folded) using /id rules 629 (w/len). 630 EXACTFL str Match this string (not guaranteed to be 631 folded) using /il rules (w/len). 632 EXACTFU str Match this string (folded iff in UTF-8, 633 length in folding doesn't change if not 634 in UTF-8) using /iu rules (w/len). 635 EXACTFA str Match this string (not guaranteed to be 636 folded) using /iaa rules (w/len). 637 EXACTFU_SS str Match this string (folded iff in UTF-8, 638 length in folding may change even if not 639 in UTF-8) using /iu rules (w/len). 640 EXACTFU_TRICKYFOLD str Match this folded UTF-8 string using /iu 641 rules 642 643 # Do nothing types 644 645 NOTHING no Match empty string. 646 # A variant of above which delimits a group, thus stops optimizations 647 TAIL no Match empty string. Can jump here from 648 outside. 649 650 # Loops 651 652 # STAR,PLUS '?', and complex '*' and '+', are implemented as 653 # circular BRANCH structures using BACK. Simple cases 654 # (one character per match) are implemented with STAR 655 # and PLUS for speed and to minimize recursive plunges. 656 # 657 STAR node Match this (simple) thing 0 or more 658 times. 659 PLUS node Match this (simple) thing 1 or more 660 times. 661 662 CURLY sv 2 Match this simple thing {n,m} times. 663 CURLYN no 2 Capture next-after-this simple thing 664 CURLYM no 2 Capture this medium-complex thing {n,m} 665 times. 666 CURLYX sv 2 Match this complex thing {n,m} times. 667 668 # This terminator creates a loop structure for CURLYX 669 WHILEM no Do curly processing and see if rest 670 matches. 671 672 # Buffer related 673 674 # OPEN,CLOSE,GROUPP ...are numbered at compile time. 675 OPEN num 1 Mark this point in input as start of #n. 676 CLOSE num 1 Analogous to OPEN. 677 678 REF num 1 Match some already matched string 679 REFF num 1 Match already matched string, folded 680 using native charset semantics for non- 681 utf8 682 REFFL num 1 Match already matched string, folded in 683 loc. 684 REFFU num 1 Match already matched string, folded 685 using unicode semantics for non-utf8 686 REFFA num 1 Match already matched string, folded 687 using unicode semantics for non-utf8, no 688 mixing ASCII, non-ASCII 689 690 # Named references. Code in regcomp.c assumes that these all are after 691 # the numbered references 692 NREF no-sv 1 Match some already matched string 693 NREFF no-sv 1 Match already matched string, folded 694 using native charset semantics for non- 695 utf8 696 NREFFL no-sv 1 Match already matched string, folded in 697 loc. 698 NREFFU num 1 Match already matched string, folded 699 using unicode semantics for non-utf8 700 NREFFA num 1 Match already matched string, folded 701 using unicode semantics for non-utf8, no 702 mixing ASCII, non-ASCII 703 704 IFMATCH off 1 2 Succeeds if the following matches. 705 UNLESSM off 1 2 Fails if the following matches. 706 SUSPEND off 1 1 "Independent" sub-RE. 707 IFTHEN off 1 1 Switch, should be preceded by switcher. 708 GROUPP num 1 Whether the group matched. 709 710 # Support for long RE 711 712 LONGJMP off 1 1 Jump far away. 713 BRANCHJ off 1 1 BRANCH with long offset. 714 715 # The heavy worker 716 717 EVAL evl 1 Execute some Perl code. 718 719 # Modifiers 720 721 MINMOD no Next operator is not greedy. 722 LOGICAL no Next opcode should set the flag only. 723 724 # This is not used yet 725 RENUM off 1 1 Group with independently numbered parens. 726 727 # Trie Related 728 729 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants 730 # have inline charclass data (ascii only), the 'C' store it in the 731 # structure. 732 733 TRIE trie 1 Match many EXACT(F[ALU]?)? at once. 734 flags==type 735 TRIEC trie Same as TRIE, but with embedded charclass 736 charclass data 737 738 AHOCORASICK trie 1 Aho Corasick stclass. flags==type 739 AHOCORASICKC trie Same as AHOCORASICK, but with embedded 740 charclass charclass data 741 742 # Regex Subroutines 743 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs 744 arg2 745 GOSTART no recurse to start of pattern 746 747 # Special conditionals 748 NGROUPP no-sv 1 Whether the group matched. 749 INSUBP num 1 Whether we are in a specific recurse. 750 DEFINEP none 1 Never execute directly. 751 752 # Backtracking Verbs 753 ENDLIKE none Used only for the type field of verbs 754 OPFAIL none Same as (?!) 755 ACCEPT parno 1 Accepts the current matched string. 756 757 # Verbs With Arguments 758 VERB no-sv 1 Used only for the type field of verbs 759 PRUNE no-sv 1 Pattern fails at this startpoint if no- 760 backtracking through this 761 MARKPOINT no-sv 1 Push the current location for rollback by 762 cut. 763 SKIP no-sv 1 On failure skip forward (to the mark) 764 before retrying 765 COMMIT no-sv 1 Pattern fails outright if backtracking 766 through this 767 CUTGROUP no-sv 1 On failure go to the next alternation in 768 the group 769 770 # Control what to keep in $&. 771 KEEPS no $& begins here. 772 773 # New charclass like patterns 774 LNBREAK none generic newline pattern 775 776 # SPECIAL REGOPS 777 778 # This is not really a node, but an optimized away piece of a "long" 779 # node. To simplify debugging output, we mark it as if it were a node 780 OPTIMIZED off Placeholder for dump. 781 782 # Special opcode with the property that no opcode in a compiled program 783 # will ever be of this type. Thus it can be used as a flag value that 784 # no other opcode has been seen. END is used similarly, in that an END 785 # node cant be optimized. So END implies "unoptimizable" and PSEUDO 786 # mean "not seen anything to optimize yet". 787 PSEUDO off Pseudo opcode for internal use. 788 789=for regcomp.pl end 790 791=for unprinted-credits 792Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421 793 794Following the optimizer information is a dump of the offset/length 795table, here split across several lines: 796 797 Offsets: [45] 798 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] 799 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] 800 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] 801 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] 802 803The first line here indicates that the offset/length table contains 45 804entries. Each entry is a pair of integers, denoted by C<offset[length]>. 805Entries are numbered starting with 1, so entry #1 here is C<1[4]> and 806entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:> 807(the C<1: ANYOF[bc]>) begins at character position 1 in the 808pre-compiled form of the regex, and has a length of 4 characters. 809C<5[1]> in position 12 810indicates that the node labeled C<12:> 811(the C<< 12: EXACT <d> >>) begins at character position 5 in the 812pre-compiled form of the regex, and has a length of 1 character. 813C<12[1]> in position 14 814indicates that the node labeled C<14:> 815(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the 816pre-compiled form of the regex, and has a length of 1 character---that 817is, it corresponds to the C<+> symbol in the precompiled regex. 818 819C<0[0]> items indicate that there is no corresponding node. 820 821=head2 Run-time Output 822 823First of all, when doing a match, one may get no run-time output even 824if debugging is enabled. This means that the regex engine was never 825entered and that all of the job was therefore done by the optimizer. 826 827If the regex engine was entered, the output may look like this: 828 829 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__' 830 Setting an EVAL scope, savestack=3 831 2 <ab> <cdefg__gh_> | 1: ANYOF 832 3 <abc> <defg__gh_> | 11: EXACT <d> 833 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767} 834 4 <abcd> <efg__gh_> | 26: WHILEM 835 0 out of 1..32767 cc=effff31c 836 4 <abcd> <efg__gh_> | 15: OPEN1 837 4 <abcd> <efg__gh_> | 17: EXACT <e> 838 5 <abcde> <fg__gh_> | 19: STAR 839 EXACT <f> can match 1 times out of 32767... 840 Setting an EVAL scope, savestack=3 841 6 <bcdef> <g__gh__> | 22: EXACT <g> 842 7 <bcdefg> <__gh__> | 24: CLOSE1 843 7 <bcdefg> <__gh__> | 26: WHILEM 844 1 out of 1..32767 cc=effff31c 845 Setting an EVAL scope, savestack=12 846 7 <bcdefg> <__gh__> | 15: OPEN1 847 7 <bcdefg> <__gh__> | 17: EXACT <e> 848 restoring \1 to 4(4)..7 849 failed, try continuation... 850 7 <bcdefg> <__gh__> | 27: NOTHING 851 7 <bcdefg> <__gh__> | 28: EXACT <h> 852 failed... 853 failed... 854 855The most significant information in the output is about the particular I<node> 856of the compiled regex that is currently being tested against the target string. 857The format of these lines is 858 859C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE> 860 861The I<TYPE> info is indented with respect to the backtracking level. 862Other incidental information appears interspersed within. 863 864=head1 Debugging Perl Memory Usage 865 866Perl is a profligate wastrel when it comes to memory use. There 867is a saying that to estimate memory usage of Perl, assume a reasonable 868algorithm for memory allocation, multiply that estimate by 10, and 869while you still may miss the mark, at least you won't be quite so 870astonished. This is not absolutely true, but may provide a good 871grasp of what happens. 872 873Assume that an integer cannot take less than 20 bytes of memory, a 874float cannot take less than 24 bytes, a string cannot take less 875than 32 bytes (all these examples assume 32-bit architectures, the 876result are quite a bit worse on 64-bit architectures). If a variable 877is accessed in two of three different ways (which require an integer, 878a float, or a string), the memory footprint may increase yet another 87920 bytes. A sloppy malloc(3) implementation can inflate these 880numbers dramatically. 881 882On the opposite end of the scale, a declaration like 883 884 sub foo; 885 886may take up to 500 bytes of memory, depending on which release of Perl 887you're running. 888 889Anecdotal estimates of source-to-compiled code bloat suggest an 890eightfold increase. This means that the compiled form of reasonable 891(normally commented, properly indented etc.) code will take 892about eight times more space in memory than the code took 893on disk. 894 895The B<-DL> command-line switch is obsolete since circa Perl 5.6.0 896(it was available only if Perl was built with C<-DDEBUGGING>). 897The switch was used to track Perl's memory allocations and possible 898memory leaks. These days the use of malloc debugging tools like 899F<Purify> or F<valgrind> is suggested instead. See also 900L<perlhacktips/PERL_MEM_LOG>. 901 902One way to find out how much memory is being used by Perl data 903structures is to install the Devel::Size module from CPAN: it gives 904you the minimum number of bytes required to store a particular data 905structure. Please be mindful of the difference between the size() 906and total_size(). 907 908If Perl has been compiled using Perl's malloc you can analyze Perl 909memory usage by setting $ENV{PERL_DEBUG_MSTATS}. 910 911=head2 Using C<$ENV{PERL_DEBUG_MSTATS}> 912 913If your perl is using Perl's malloc() and was compiled with the 914necessary switches (this is the default), then it will print memory 915usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS} 916> 1 >>, and before termination of the program when C<< 917$ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to 918the following example: 919 920 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp" 921 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192) 922 14216 free: 130 117 28 7 9 0 2 2 1 0 0 923 437 61 36 0 5 924 60924 used: 125 137 161 55 7 8 6 16 2 0 1 925 74 109 304 84 20 926 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048. 927 Memory allocation statistics after execution: (buckets 4(4)..8188(8192) 928 30888 free: 245 78 85 13 6 2 1 3 2 0 1 929 315 162 39 42 11 930 175816 used: 265 176 1112 111 26 22 11 27 2 1 1 931 196 178 1066 798 39 932 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144. 933 934It is possible to ask for such a statistic at arbitrary points in 935your execution using the mstat() function out of the standard 936Devel::Peek module. 937 938Here is some explanation of that format: 939 940=over 4 941 942=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)> 943 944Perl's malloc() uses bucketed allocations. Every request is rounded 945up to the closest bucket size available, and a bucket is taken from 946the pool of buckets of that size. 947 948The line above describes the limits of buckets currently in use. 949Each bucket has two sizes: memory footprint and the maximal size 950of user data that can fit into this bucket. Suppose in the above 951example that the smallest bucket were size 4. The biggest bucket 952would have usable size 8188, and the memory footprint would be 8192. 953 954In a Perl built for debugging, some buckets may have negative usable 955size. This means that these buckets cannot (and will not) be used. 956For larger buckets, the memory footprint may be one page greater 957than a power of 2. If so, the corresponding power of two is 958printed in the C<APPROX> field above. 959 960=item Free/Used 961 962The 1 or 2 rows of numbers following that correspond to the number 963of buckets of each size between C<SMALLEST> and C<GREATEST>. In 964the first row, the sizes (memory footprints) of buckets are powers 965of two--or possibly one page greater. In the second row, if present, 966the memory footprints of the buckets are between the memory footprints 967of two buckets "above". 968 969For example, suppose under the previous example, the memory footprints 970were 971 972 free: 8 16 32 64 128 256 512 1024 2048 4096 8192 973 4 12 24 48 80 974 975With a non-C<DEBUGGING> perl, the buckets starting from C<128> have 976a 4-byte overhead, and thus an 8192-long bucket may take up to 9778188-byte allocations. 978 979=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS> 980 981The first two fields give the total amount of memory perl sbrk(2)ed 982(ess-broken? :-) and number of sbrk(2)s used. The third number is 983what perl thinks about continuity of returned chunks. So long as 984this number is positive, malloc() will assume that it is probable 985that sbrk(2) will provide continuous memory. 986 987Memory allocated by external libraries is not counted. 988 989=item C<pad: 0> 990 991The amount of sbrk(2)ed memory needed to keep buckets aligned. 992 993=item C<heads: 2192> 994 995Although memory overhead of bigger buckets is kept inside the bucket, for 996smaller buckets, it is kept in separate areas. This field gives the 997total size of these areas. 998 999=item C<chain: 0> 1000 1001malloc() may want to subdivide a bigger bucket into smaller buckets. 1002If only a part of the deceased bucket is left unsubdivided, the rest 1003is kept as an element of a linked list. This field gives the total 1004size of these chunks. 1005 1006=item C<tail: 6144> 1007 1008To minimize the number of sbrk(2)s, malloc() asks for more memory. This 1009field gives the size of the yet unused part, which is sbrk(2)ed, but 1010never touched. 1011 1012=back 1013 1014=head1 SEE ALSO 1015 1016L<perldebug>, 1017L<perlguts>, 1018L<perlrun> 1019L<re>, 1020and 1021L<Devel::DProf>. 1022