1=head1 NAME 2 3perlreguts - Description of the Perl regular expression engine. 4 5=head1 DESCRIPTION 6 7This document is an attempt to shine some light on the guts of the regex 8engine and how it works. The regex engine represents a significant chunk 9of the perl codebase, but is relatively poorly understood. This document 10is a meagre attempt at addressing this situation. It is derived from the 11author's experience, comments in the source code, other papers on the 12regex engine, feedback on the perl5-porters mail list, and no doubt other 13places as well. 14 15B<NOTICE!> It should be clearly understood that the behavior and 16structures discussed in this represents the state of the engine as the 17author understood it at the time of writing. It is B<NOT> an API 18definition, it is purely an internals guide for those who want to hack 19the regex engine, or understand how the regex engine works. Readers of 20this document are expected to understand perl's regex syntax and its 21usage in detail. If you want to learn about the basics of Perl's 22regular expressions, see L<perlre>. And if you want to replace the 23regex engine with your own, see L<perlreapi>. 24 25=head1 OVERVIEW 26 27=head2 A quick note on terms 28 29There is some debate as to whether to say "regexp" or "regex". In this 30document we will use the term "regex" unless there is a special reason 31not to, in which case we will explain why. 32 33When speaking about regexes we need to distinguish between their source 34code form and their internal form. In this document we will use the term 35"pattern" when we speak of their textual, source code form, and the term 36"program" when we speak of their internal representation. These 37correspond to the terms I<S-regex> and I<B-regex> that Mark Jason 38Dominus employs in his paper on "Rx" ([1] in L</REFERENCES>). 39 40=head2 What is a regular expression engine? 41 42A regular expression engine is a program that takes a set of constraints 43specified in a mini-language, and then applies those constraints to a 44target string, and determines whether or not the string satisfies the 45constraints. See L<perlre> for a full definition of the language. 46 47In less grandiose terms, the first part of the job is to turn a pattern into 48something the computer can efficiently use to find the matching point in 49the string, and the second part is performing the search itself. 50 51To do this we need to produce a program by parsing the text. We then 52need to execute the program to find the point in the string that 53matches. And we need to do the whole thing efficiently. 54 55=head2 Structure of a Regexp Program 56 57=head3 High Level 58 59Although it is a bit confusing and some people object to the terminology, it 60is worth taking a look at a comment that has 61been in F<regexp.h> for years: 62 63I<This is essentially a linear encoding of a nondeterministic 64finite-state machine (aka syntax charts or "railroad normal form" in 65parsing technology).> 66 67The term "railroad normal form" is a bit esoteric, with "syntax 68diagram/charts", or "railroad diagram/charts" being more common terms. 69Nevertheless it provides a useful mental image of a regex program: each 70node can be thought of as a unit of track, with a single entry and in 71most cases a single exit point (there are pieces of track that fork, but 72statistically not many), and the whole forms a layout with a 73single entry and single exit point. The matching process can be thought 74of as a car that moves along the track, with the particular route through 75the system being determined by the character read at each possible 76connector point. A car can fall off the track at any point but it may 77only proceed as long as it matches the track. 78 79Thus the pattern C</foo(?:\w+|\d+|\s+)bar/> can be thought of as the 80following chart: 81 82 [start] 83 | 84 <foo> 85 | 86 +-----+-----+ 87 | | | 88 <\w+> <\d+> <\s+> 89 | | | 90 +-----+-----+ 91 | 92 <bar> 93 | 94 [end] 95 96The truth of the matter is that perl's regular expressions these days are 97much more complex than this kind of structure, but visualising it this way 98can help when trying to get your bearings, and it matches the 99current implementation pretty closely. 100 101To be more precise, we will say that a regex program is an encoding 102of a graph. Each node in the graph corresponds to part of 103the original regex pattern, such as a literal string or a branch, 104and has a pointer to the nodes representing the next component 105to be matched. Since "node" and "opcode" already have other meanings in the 106perl source, we will call the nodes in a regex program "regops". 107 108The program is represented by an array of C<regnode> structures, one or 109more of which represent a single regop of the program. Struct 110C<regnode> is the smallest struct needed, and has a field structure which is 111shared with all the other larger structures. (Outside this document, the term 112"regnode" is sometimes used to mean "regop", which could be confusing.) 113 114=for apidoc Cyh||regnode 115 116The "next" pointers of all regops except C<BRANCH> implement concatenation; 117a "next" pointer with a C<BRANCH> on both ends of it is connecting two 118alternatives. [Here we have one of the subtle syntax dependencies: an 119individual C<BRANCH> (as opposed to a collection of them) is never 120concatenated with anything because of operator precedence.] 121 122The operand of some types of regop is a literal string; for others, 123it is a regop leading into a sub-program. In particular, the operand 124of a C<BRANCH> node is the first regop of the branch. 125 126B<NOTE>: As the railroad metaphor suggests, this is B<not> a tree 127structure: the tail of the branch connects to the thing following the 128set of C<BRANCH>es. It is a like a single line of railway track that 129splits as it goes into a station or railway yard and rejoins as it comes 130out the other side. 131 132=head3 Regops 133 134The base structure of a regop is defined in F<regexp.h> as follows: 135 136 struct regnode { 137 U8 flags; /* Various purposes, sometimes overridden */ 138 U8 type; /* Opcode value as specified by regnodes.h */ 139 U16 next_off; /* Offset in size regnode */ 140 }; 141 142Other larger C<regnode>-like structures are defined in F<regcomp.h>. They 143are almost like subclasses in that they have the same fields as 144C<regnode>, with possibly additional fields following in 145the structure, and in some cases the specific meaning (and name) 146of some of base fields are overridden. The following is a more 147complete description. 148 149=over 4 150 151=item C<regnode_1> 152 153=item C<regnode_2> 154 155C<regnode_1> structures have the same header, followed by a single 156four-byte argument; C<regnode_2> structures contain two two-byte 157arguments instead: 158 159 regnode_1 U32 arg1; 160 regnode_2 U16 arg1; U16 arg2; 161 162=item C<regnode_string> 163 164C<regnode_string> structures, used for literal strings, follow the header 165with a one-byte length and then the string data. Strings are padded on 166the tail end with zero bytes so that the total length of the node is a 167multiple of four bytes: 168 169 regnode_string char string[1]; 170 U8 str_len; /* overrides flags */ 171 172=item C<regnode_charclass> 173 174Bracketed character classes are represented by C<regnode_charclass> 175structures, which have a four-byte argument and then a 32-byte (256-bit) 176bitmap indicating which characters in the Latin1 range are included in 177the class. 178 179 regnode_charclass U32 arg1; 180 char bitmap[ANYOF_BITMAP_SIZE]; 181 182Various flags whose names begin with C<ANYOF_> are used for special 183situations. Above Latin1 matches and things not known until run-time 184are stored in L</Perl's pprivate structure>. 185 186=item C<regnode_charclass_posixl> 187 188There is also a larger form of a char class structure used to represent 189POSIX char classes under C</l> matching, 190called C<regnode_charclass_posixl> which has an 191additional 32-bit bitmap indicating which POSIX char classes 192have been included. 193 194 regnode_charclass_posixl U32 arg1; 195 char bitmap[ANYOF_BITMAP_SIZE]; 196 U32 classflags; 197 198=back 199 200F<regnodes.h> defines an array called C<regarglen[]> which gives the size 201of each opcode in units of C<size regnode> (4-byte). A macro is used 202to calculate the size of an C<EXACT> node based on its C<str_len> field. 203 204The regops are defined in F<regnodes.h> which is generated from 205F<regcomp.sym> by F<regcomp.pl>. Currently the maximum possible number 206of distinct regops is restricted to 256, with about a quarter already 207used. 208 209A set of macros makes accessing the fields 210easier and more consistent. These include C<OP()>, which is used to determine 211the type of a C<regnode>-like structure; C<NEXT_OFF()>, which is the offset to 212the next node (more on this later); C<ARG()>, C<ARG1()>, C<ARG2()>, C<ARG_SET()>, 213and equivalents for reading and setting the arguments; and C<STR_LEN()>, 214C<STRING()> and C<OPERAND()> for manipulating strings and regop bearing 215types. 216 217=head3 What regop is next? 218 219There are three distinct concepts of "next" in the regex engine, and 220it is important to keep them clear. 221 222=over 4 223 224=item * 225 226There is the "next regnode" from a given regnode, a value which is 227rarely useful except that sometimes it matches up in terms of value 228with one of the others, and that sometimes the code assumes this to 229always be so. 230 231=item * 232 233There is the "next regop" from a given regop/regnode. This is the 234regop physically located after the current one, as determined by 235the size of the current regop. This is often useful, such as when 236dumping the structure we use this order to traverse. Sometimes the code 237assumes that the "next regnode" is the same as the "next regop", or in 238other words assumes that the sizeof a given regop type is always going 239to be one regnode large. 240 241=item * 242 243There is the "regnext" from a given regop. This is the regop which 244is reached by jumping forward by the value of C<NEXT_OFF()>, 245or in a few cases for longer jumps by the C<arg1> field of the C<regnode_1> 246structure. The subroutine C<regnext()> handles this transparently. 247This is the logical successor of the node, which in some cases, like 248that of the C<BRANCH> regop, has special meaning. 249 250=back 251 252=head1 Process Overview 253 254Broadly speaking, performing a match of a string against a pattern 255involves the following steps: 256 257=over 5 258 259=item A. Compilation 260 261=over 5 262 263=item 1. Parsing 264 265=item 2. Peep-hole optimisation and analysis 266 267=back 268 269=item B. Execution 270 271=over 5 272 273=item 3. Start position and no-match optimisations 274 275=item 4. Program execution 276 277=back 278 279=back 280 281 282Where these steps occur in the actual execution of a perl program is 283determined by whether the pattern involves interpolating any string 284variables. If interpolation occurs, then compilation happens at run time. If it 285does not, then compilation is performed at compile time. (The C</o> modifier changes this, 286as does C<qr//> to a certain extent.) The engine doesn't really care that 287much. 288 289=head2 Compilation 290 291This code resides primarily in F<regcomp.c>, along with the header files 292F<regcomp.h>, F<regexp.h> and F<regnodes.h>. 293 294Compilation starts with C<pregcomp()>, which is mostly an initialisation 295wrapper which farms work out to two other routines for the heavy lifting: the 296first is C<reg()>, which is the start point for parsing; the second, 297C<study_chunk()>, is responsible for optimisation. 298 299Initialisation in C<pregcomp()> mostly involves the creation and data-filling 300of a special structure, C<RExC_state_t> (defined in F<regcomp.c>). 301Almost all internally-used routines in F<regcomp.h> take a pointer to one 302of these structures as their first argument, with the name C<pRExC_state>. 303This structure is used to store the compilation state and contains many 304fields. Likewise there are many macros which operate on this 305variable: anything that looks like C<RExC_xxxx> is a macro that operates on 306this pointer/structure. 307 308C<reg()> is the start of the parse process. It is responsible for 309parsing an arbitrary chunk of pattern up to either the end of the 310string, or the first closing parenthesis it encounters in the pattern. 311This means it can be used to parse the top-level regex, or any section 312inside of a grouping parenthesis. It also handles the "special parens" 313that perl's regexes have. For instance when parsing C</x(?:foo)y/>, 314C<reg()> will at one point be called to parse from the "?" symbol up to 315and including the ")". 316 317Additionally, C<reg()> is responsible for parsing the one or more 318branches from the pattern, and for "finishing them off" by correctly 319setting their next pointers. In order to do the parsing, it repeatedly 320calls out to C<regbranch()>, which is responsible for handling up to the 321first C<|> symbol it sees. 322 323C<regbranch()> in turn calls C<regpiece()> which 324handles "things" followed by a quantifier. In order to parse the 325"things", C<regatom()> is called. This is the lowest level routine, which 326parses out constant strings, character classes, and the 327various special symbols like C<$>. If C<regatom()> encounters a "(" 328character it in turn calls C<reg()>. 329 330There used to be two main passes involved in parsing, the first to 331calculate the size of the compiled program, and the second to actually 332compile it. But now there is only one main pass, with an initial crude 333guess based on the length of the input pattern, which is increased if 334necessary as parsing proceeds, and afterwards, trimmed to the actual 335amount used. 336 337However, it may happen that parsing must be restarted at the beginning 338when various circumstances occur along the way. An example is if the 339program turns out to be so large that there are jumps in it that won't 340fit in the normal 16 bits available. There are two special regops that 341can hold bigger jump destinations, BRANCHJ and LONGBRANCH. The parse is 342restarted, and these are used instead of the normal shorter ones. 343Whenever restarting the parse is required, the function returns failure 344and sets a flag as to what needs to be done. This is passed up to the 345top level routine which takes the appropriate action and restarts from 346scratch. In the case of needing longer jumps, the C<RExC_use_BRANCHJ> 347flag is set in the C<RExC_state_t> structure, which the functions know 348to inspect before deciding how to do branches. 349 350In most instances, the function that discovers the issue sets the causal 351flag and returns failure immediately. L</Parsing complications> 352contains an explicit example of how this works. In other cases, such as 353a forward reference to a numbered parenthetical grouping, we need to 354finish the parse to know if that numbered grouping actually appears in 355the pattern. In those cases, the parse is just redone at the end, with 356the knowledge of how many groupings occur in it. 357 358The routine C<regtail()> is called by both C<reg()> and C<regbranch()> 359in order to "set the tail pointer" correctly. When executing and 360we get to the end of a branch, we need to go to the node following the 361grouping parens. When parsing, however, we don't know where the end will 362be until we get there, so when we do we must go back and update the 363offsets as appropriate. C<regtail> is used to make this easier. 364 365A subtlety of the parsing process means that a regex like C</foo/> is 366originally parsed into an alternation with a single branch. It is only 367afterwards that the optimiser converts single branch alternations into the 368simpler form. 369 370=head3 Parse Call Graph and a Grammar 371 372The call graph looks like this: 373 374 reg() # parse a top level regex, or inside of 375 # parens 376 regbranch() # parse a single branch of an alternation 377 regpiece() # parse a pattern followed by a quantifier 378 regatom() # parse a simple pattern 379 regclass() # used to handle a class 380 reg() # used to handle a parenthesised 381 # subpattern 382 .... 383 ... 384 regtail() # finish off the branch 385 ... 386 regtail() # finish off the branch sequence. Tie each 387 # branch's tail to the tail of the 388 # sequence 389 # (NEW) In Debug mode this is 390 # regtail_study(). 391 392A grammar form might be something like this: 393 394 atom : constant | class 395 quant : '*' | '+' | '?' | '{min,max}' 396 _branch: piece 397 | piece _branch 398 | nothing 399 branch: _branch 400 | _branch '|' branch 401 group : '(' branch ')' 402 _piece: atom | group 403 piece : _piece 404 | _piece quant 405 406=head3 Parsing complications 407 408The implication of the above description is that a pattern containing nested 409parentheses will result in a call graph which cycles through C<reg()>, 410C<regbranch()>, C<regpiece()>, C<regatom()>, C<reg()>, C<regbranch()> I<etc> 411multiple times, until the deepest level of nesting is reached. All the above 412routines return a pointer to a C<regnode>, which is usually the last regnode 413added to the program. However, one complication is that reg() returns NULL 414for parsing C<(?:)> syntax for embedded modifiers, setting the flag 415C<TRYAGAIN>. The C<TRYAGAIN> propagates upwards until it is captured, in 416some cases by C<regatom()>, but otherwise unconditionally by 417C<regbranch()>. Hence it will never be returned by C<regbranch()> to 418C<reg()>. This flag permits patterns such as C<(?i)+> to be detected as 419errors (I<Quantifier follows nothing in regex; marked by <-- HERE in m/(?i)+ 420<-- HERE />). 421 422Another complication is that the representation used for the program differs 423if it needs to store Unicode, but it's not always possible to know for sure 424whether it does until midway through parsing. The Unicode representation for 425the program is larger, and cannot be matched as efficiently. (See L</Unicode 426and Localisation Support> below for more details as to why.) If the pattern 427contains literal Unicode, it's obvious that the program needs to store 428Unicode. Otherwise, the parser optimistically assumes that the more 429efficient representation can be used, and starts sizing on this basis. 430However, if it then encounters something in the pattern which must be stored 431as Unicode, such as an C<\x{...}> escape sequence representing a character 432literal, then this means that all previously calculated sizes need to be 433redone, using values appropriate for the Unicode representation. This 434is another instance where the parsing needs to be restarted, and it can 435and is done immediately. The function returns failure, and sets the 436flag C<RESTART_UTF8> (encapsulated by using the macro C<REQUIRE_UTF8>). 437This restart request is propagated up the call chain in a similar 438fashion, until it is "caught" in C<Perl_re_op_compile()>, which marks 439the pattern as containing Unicode, and restarts the sizing pass. It is 440also possible for constructions within run-time code blocks to turn out 441to need Unicode representation., which is signalled by 442C<S_compile_runtime_code()> returning false to C<Perl_re_op_compile()>. 443 444The restart was previously implemented using a C<longjmp> in C<regatom()> 445back to a C<setjmp> in C<Perl_re_op_compile()>, but this proved to be 446problematic as the latter is a large function containing many automatic 447variables, which interact badly with the emergent control flow of C<setjmp>. 448 449=head3 Debug Output 450 451Starting in the 5.9.x development version of perl you can C<< use re 452Debug => 'PARSE' >> to see some trace information about the parse 453process. We will start with some simple patterns and build up to more 454complex patterns. 455 456So when we parse C</foo/> we see something like the following table. The 457left shows what is being parsed, and the number indicates where the next regop 458would go. The stuff on the right is the trace output of the graph. The 459names are chosen to be short to make it less dense on the screen. 'tsdy' 460is a special form of C<regtail()> which does some extra analysis. 461 462 >foo< 1 reg 463 brnc 464 piec 465 atom 466 >< 4 tsdy~ EXACT <foo> (EXACT) (1) 467 ~ attach to END (3) offset to 2 468 469The resulting program then looks like: 470 471 1: EXACT <foo>(3) 472 3: END(0) 473 474As you can see, even though we parsed out a branch and a piece, it was ultimately 475only an atom. The final program shows us how things work. We have an C<EXACT> regop, 476followed by an C<END> regop. The number in parens indicates where the C<regnext> of 477the node goes. The C<regnext> of an C<END> regop is unused, as C<END> regops mean 478we have successfully matched. The number on the left indicates the position of 479the regop in the regnode array. 480 481Now let's try a harder pattern. We will add a quantifier, so now we have the pattern 482C</foo+/>. We will see that C<regbranch()> calls C<regpiece()> twice. 483 484 >foo+< 1 reg 485 brnc 486 piec 487 atom 488 >o+< 3 piec 489 atom 490 >< 6 tail~ EXACT <fo> (1) 491 7 tsdy~ EXACT <fo> (EXACT) (1) 492 ~ PLUS (END) (3) 493 ~ attach to END (6) offset to 3 494 495And we end up with the program: 496 497 1: EXACT <fo>(3) 498 3: PLUS(6) 499 4: EXACT <o>(0) 500 6: END(0) 501 502Now we have a special case. The C<EXACT> regop has a C<regnext> of 0. This is 503because if it matches it should try to match itself again. The C<PLUS> regop 504handles the actual failure of the C<EXACT> regop and acts appropriately (going 505to regnode 6 if the C<EXACT> matched at least once, or failing if it didn't). 506 507Now for something much more complex: C</x(?:foo*|b[a][rR])(foo|bar)$/> 508 509 >x(?:foo*|b... 1 reg 510 brnc 511 piec 512 atom 513 >(?:foo*|b[... 3 piec 514 atom 515 >?:foo*|b[a... reg 516 >foo*|b[a][... brnc 517 piec 518 atom 519 >o*|b[a][rR... 5 piec 520 atom 521 >|b[a][rR])... 8 tail~ EXACT <fo> (3) 522 >b[a][rR])(... 9 brnc 523 10 piec 524 atom 525 >[a][rR])(f... 12 piec 526 atom 527 >a][rR])(fo... clas 528 >[rR])(foo|... 14 tail~ EXACT <b> (10) 529 piec 530 atom 531 >rR])(foo|b... clas 532 >)(foo|bar)... 25 tail~ EXACT <a> (12) 533 tail~ BRANCH (3) 534 26 tsdy~ BRANCH (END) (9) 535 ~ attach to TAIL (25) offset to 16 536 tsdy~ EXACT <fo> (EXACT) (4) 537 ~ STAR (END) (6) 538 ~ attach to TAIL (25) offset to 19 539 tsdy~ EXACT <b> (EXACT) (10) 540 ~ EXACT <a> (EXACT) (12) 541 ~ ANYOF[Rr] (END) (14) 542 ~ attach to TAIL (25) offset to 11 543 >(foo|bar)$< tail~ EXACT <x> (1) 544 piec 545 atom 546 >foo|bar)$< reg 547 28 brnc 548 piec 549 atom 550 >|bar)$< 31 tail~ OPEN1 (26) 551 >bar)$< brnc 552 32 piec 553 atom 554 >)$< 34 tail~ BRANCH (28) 555 36 tsdy~ BRANCH (END) (31) 556 ~ attach to CLOSE1 (34) offset to 3 557 tsdy~ EXACT <foo> (EXACT) (29) 558 ~ attach to CLOSE1 (34) offset to 5 559 tsdy~ EXACT <bar> (EXACT) (32) 560 ~ attach to CLOSE1 (34) offset to 2 561 >$< tail~ BRANCH (3) 562 ~ BRANCH (9) 563 ~ TAIL (25) 564 piec 565 atom 566 >< 37 tail~ OPEN1 (26) 567 ~ BRANCH (28) 568 ~ BRANCH (31) 569 ~ CLOSE1 (34) 570 38 tsdy~ EXACT <x> (EXACT) (1) 571 ~ BRANCH (END) (3) 572 ~ BRANCH (END) (9) 573 ~ TAIL (END) (25) 574 ~ OPEN1 (END) (26) 575 ~ BRANCH (END) (28) 576 ~ BRANCH (END) (31) 577 ~ CLOSE1 (END) (34) 578 ~ EOL (END) (36) 579 ~ attach to END (37) offset to 1 580 581Resulting in the program 582 583 1: EXACT <x>(3) 584 3: BRANCH(9) 585 4: EXACT <fo>(6) 586 6: STAR(26) 587 7: EXACT <o>(0) 588 9: BRANCH(25) 589 10: EXACT <ba>(14) 590 12: OPTIMIZED (2 nodes) 591 14: ANYOF[Rr](26) 592 25: TAIL(26) 593 26: OPEN1(28) 594 28: TRIE-EXACT(34) 595 [StS:1 Wds:2 Cs:6 Uq:5 #Sts:7 Mn:3 Mx:3 Stcls:bf] 596 <foo> 597 <bar> 598 30: OPTIMIZED (4 nodes) 599 34: CLOSE1(36) 600 36: EOL(37) 601 37: END(0) 602 603Here we can see a much more complex program, with various optimisations in 604play. At regnode 10 we see an example where a character class with only 605one character in it was turned into an C<EXACT> node. We can also see where 606an entire alternation was turned into a C<TRIE-EXACT> node. As a consequence, 607some of the regnodes have been marked as optimised away. We can see that 608the C<$> symbol has been converted into an C<EOL> regop, a special piece of 609code that looks for C<\n> or the end of the string. 610 611The next pointer for C<BRANCH>es is interesting in that it points at where 612execution should go if the branch fails. When executing, if the engine 613tries to traverse from a branch to a C<regnext> that isn't a branch then 614the engine will know that the entire set of branches has failed. 615 616=head3 Peep-hole Optimisation and Analysis 617 618The regular expression engine can be a weighty tool to wield. On long 619strings and complex patterns it can end up having to do a lot of work 620to find a match, and even more to decide that no match is possible. 621Consider a situation like the following pattern. 622 623 'ababababababababababab' =~ /(a|b)*z/ 624 625The C<(a|b)*> part can match at every char in the string, and then fail 626every time because there is no C<z> in the string. So obviously we can 627avoid using the regex engine unless there is a C<z> in the string. 628Likewise in a pattern like: 629 630 /foo(\w+)bar/ 631 632In this case we know that the string must contain a C<foo> which must be 633followed by C<bar>. We can use Fast Boyer-Moore matching as implemented 634in C<fbm_instr()> to find the location of these strings. If they don't exist 635then we don't need to resort to the much more expensive regex engine. 636Even better, if they do exist then we can use their positions to 637reduce the search space that the regex engine needs to cover to determine 638if the entire pattern matches. 639 640There are various aspects of the pattern that can be used to facilitate 641optimisations along these lines: 642 643=over 5 644 645=item * anchored fixed strings 646 647=item * floating fixed strings 648 649=item * minimum and maximum length requirements 650 651=item * start class 652 653=item * Beginning/End of line positions 654 655=back 656 657Another form of optimisation that can occur is the post-parse "peep-hole" 658optimisation, where inefficient constructs are replaced by more efficient 659constructs. The C<TAIL> regops which are used during parsing to mark the end 660of branches and the end of groups are examples of this. These regops are used 661as place-holders during construction and "always match" so they can be 662"optimised away" by making the things that point to the C<TAIL> point to the 663thing that C<TAIL> points to, thus "skipping" the node. 664 665Another optimisation that can occur is that of "C<EXACT> merging" which is 666where two consecutive C<EXACT> nodes are merged into a single 667regop. An even more aggressive form of this is that a branch 668sequence of the form C<EXACT BRANCH ... EXACT> can be converted into a 669C<TRIE-EXACT> regop. 670 671All of this occurs in the routine C<study_chunk()> which uses a special 672structure C<scan_data_t> to store the analysis that it has performed, and 673does the "peep-hole" optimisations as it goes. 674 675The code involved in C<study_chunk()> is extremely cryptic. Be careful. :-) 676 677=head2 Execution 678 679Execution of a regex generally involves two phases, the first being 680finding the start point in the string where we should match from, 681and the second being running the regop interpreter. 682 683If we can tell that there is no valid start point then we don't bother running 684the interpreter at all. Likewise, if we know from the analysis phase that we 685cannot detect a short-cut to the start position, we go straight to the 686interpreter. 687 688The two entry points are C<re_intuit_start()> and C<pregexec()>. These routines 689have a somewhat incestuous relationship with overlap between their functions, 690and C<pregexec()> may even call C<re_intuit_start()> on its own. Nevertheless 691other parts of the perl source code may call into either, or both. 692 693Execution of the interpreter itself used to be recursive, but thanks to the 694efforts of Dave Mitchell in the 5.9.x development track, that has changed: now an 695internal stack is maintained on the heap and the routine is fully 696iterative. This can make it tricky as the code is quite conservative 697about what state it stores, with the result that two consecutive lines in the 698code can actually be running in totally different contexts due to the 699simulated recursion. 700 701=for apidoc pregcomp 702=for apidoc pregexec 703 704=head3 Start position and no-match optimisations 705 706C<re_intuit_start()> is responsible for handling start points and no-match 707optimisations as determined by the results of the analysis done by 708C<study_chunk()> (and described in L</Peep-hole Optimisation and Analysis>). 709 710The basic structure of this routine is to try to find the start- and/or 711end-points of where the pattern could match, and to ensure that the string 712is long enough to match the pattern. It tries to use more efficient 713methods over less efficient methods and may involve considerable 714cross-checking of constraints to find the place in the string that matches. 715For instance it may try to determine that a given fixed string must be 716not only present but a certain number of chars before the end of the 717string, or whatever. 718 719It calls several other routines, such as C<fbm_instr()> which does 720Fast Boyer Moore matching and C<find_byclass()> which is responsible for 721finding the start using the first mandatory regop in the program. 722 723When the optimisation criteria have been satisfied, C<reg_try()> is called 724to perform the match. 725 726=head3 Program execution 727 728C<pregexec()> is the main entry point for running a regex. It contains 729support for initialising the regex interpreter's state, running 730C<re_intuit_start()> if needed, and running the interpreter on the string 731from various start positions as needed. When it is necessary to use 732the regex interpreter C<pregexec()> calls C<regtry()>. 733 734C<regtry()> is the entry point into the regex interpreter. It expects 735as arguments a pointer to a C<regmatch_info> structure and a pointer to 736a string. It returns an integer 1 for success and a 0 for failure. 737It is basically a set-up wrapper around C<regmatch()>. 738 739C<regmatch> is the main "recursive loop" of the interpreter. It is 740basically a giant switch statement that implements a state machine, where 741the possible states are the regops themselves, plus a number of additional 742intermediate and failure states. A few of the states are implemented as 743subroutines but the bulk are inline code. 744 745=head1 MISCELLANEOUS 746 747=head2 Unicode and Localisation Support 748 749When dealing with strings containing characters that cannot be represented 750using an eight-bit character set, perl uses an internal representation 751that is a permissive version of Unicode's UTF-8 encoding[2]. This uses single 752bytes to represent characters from the ASCII character set, and sequences 753of two or more bytes for all other characters. (See L<perlunitut> 754for more information about the relationship between UTF-8 and perl's 755encoding, utf8. The difference isn't important for this discussion.) 756 757No matter how you look at it, Unicode support is going to be a pain in a 758regex engine. Tricks that might be fine when you have 256 possible 759characters often won't scale to handle the size of the UTF-8 character 760set. Things you can take for granted with ASCII may not be true with 761Unicode. For instance, in ASCII, it is safe to assume that 762C<sizeof(char1) == sizeof(char2)>, but in UTF-8 it isn't. Unicode case folding is 763vastly more complex than the simple rules of ASCII, and even when not 764using Unicode but only localised single byte encodings, things can get 765tricky (for example, B<LATIN SMALL LETTER SHARP S> (U+00DF, E<szlig>) 766should match 'SS' in localised case-insensitive matching). 767 768Making things worse is that UTF-8 support was a later addition to the 769regex engine (as it was to perl) and this necessarily made things a lot 770more complicated. Obviously it is easier to design a regex engine with 771Unicode support in mind from the beginning than it is to retrofit it to 772one that wasn't. 773 774Nearly all regops that involve looking at the input string have 775two cases, one for UTF-8, and one not. In fact, it's often more complex 776than that, as the pattern may be UTF-8 as well. 777 778Care must be taken when making changes to make sure that you handle 779UTF-8 properly, both at compile time and at execution time, including 780when the string and pattern are mismatched. 781 782=head2 Base Structures 783 784The C<regexp> structure described in L<perlreapi> is common to all 785regex engines. Two of its fields are intended for the private use 786of the regex engine that compiled the pattern. These are the 787C<intflags> and pprivate members. The C<pprivate> is a void pointer to 788an arbitrary structure whose use and management is the responsibility 789of the compiling engine. perl will never modify either of these 790values. In the case of the stock engine the structure pointed to by 791C<pprivate> is called C<regexp_internal>. 792 793Its C<pprivate> and C<intflags> fields contain data 794specific to each engine. 795 796There are two structures used to store a compiled regular expression. 797One, the C<regexp> structure described in L<perlreapi> is populated by 798the engine currently being. used and some of its fields read by perl to 799implement things such as the stringification of C<qr//>. 800 801 802The other structure is pointed to by the C<regexp> struct's 803C<pprivate> and is in addition to C<intflags> in the same struct 804considered to be the property of the regex engine which compiled the 805regular expression; 806 807The regexp structure contains all the data that perl needs to be aware of 808to properly work with the regular expression. It includes data about 809optimisations that perl can use to determine if the regex engine should 810really be used, and various other control info that is needed to properly 811execute patterns in various contexts such as is the pattern anchored in 812some way, or what flags were used during the compile, or whether the 813program contains special constructs that perl needs to be aware of. 814 815In addition it contains two fields that are intended for the private use 816of the regex engine that compiled the pattern. These are the C<intflags> 817and pprivate members. The C<pprivate> is a void pointer to an arbitrary 818structure whose use and management is the responsibility of the compiling 819engine. perl will never modify either of these values. 820 821As mentioned earlier, in the case of the default engines, the C<pprivate> 822will be a pointer to a regexp_internal structure which holds the compiled 823program and any additional data that is private to the regex engine 824implementation. 825 826=head3 Perl's C<pprivate> structure 827 828The following structure is used as the C<pprivate> struct by perl's 829regex engine. Since it is specific to perl it is only of curiosity 830value to other engine implementations. 831 832 typedef struct regexp_internal { 833 regnode *regstclass; 834 struct reg_data *data; 835 struct reg_code_blocks *code_blocks; 836 U32 proglen; 837 U32 name_list_idx; 838 regnode program[1]; 839 } regexp_internal; 840 841Description of the attributes is as follows: 842 843=over 5 844 845=item C<regstclass> 846 847Special regop that is used by C<re_intuit_start()> to check if a pattern 848can match at a certain position. For instance if the regex engine knows 849that the pattern must start with a 'Z' then it can scan the string until 850it finds one and then launch the regex engine from there. The routine 851that handles this is called C<find_by_class()>. Sometimes this field 852points at a regop embedded in the program, and sometimes it points at 853an independent synthetic regop that has been constructed by the optimiser. 854 855=item C<data> 856 857This field points at a C<reg_data> structure, which is defined as follows 858 859 struct reg_data { 860 U32 count; 861 U8 *what; 862 void* data[1]; 863 }; 864 865This structure is used for handling data structures that the regex engine 866needs to handle specially during a clone or free operation on the compiled 867product. Each element in the data array has a corresponding element in the 868what array. During compilation regops that need special structures stored 869will add an element to each array using the add_data() routine and then store 870the index in the regop. 871 872In modern perls the 0th element of this structure is reserved and is NEVER 873used to store anything of use. This is to allow things that need to index 874into this array to represent "no value". 875 876=item C<code_blocks> 877 878This optional structure is used to manage C<(?{})> constructs in the 879pattern. It is made up of the following structures. 880 881 /* record the position of a (?{...}) within a pattern */ 882 struct reg_code_block { 883 STRLEN start; 884 STRLEN end; 885 OP *block; 886 REGEXP *src_regex; 887 }; 888 889 /* array of reg_code_block's plus header info */ 890 struct reg_code_blocks { 891 int refcnt; /* we may be pointed to from a regex 892 and from the savestack */ 893 int count; /* how many code blocks */ 894 struct reg_code_block *cb; /* array of reg_code_block's */ 895 }; 896 897=item C<proglen> 898 899Stores the length of the compiled program in units of regops. 900 901=item C<name_list_idx> 902 903This is the index into the data array where an AV is stored that contains 904the names of any named capture buffers in the pattern, should there be 905any. This is only used in the debugging version of the regex engine and 906when RXp_PAREN_NAMES(prog) is true. It will be 0 if there is no such data. 907 908=item C<program> 909 910Compiled program. Inlined into the structure so the entire struct can be 911treated as a single blob. 912 913=back 914 915=head1 SEE ALSO 916 917L<perlreapi> 918 919L<perlre> 920 921L<perlunitut> 922 923=head1 AUTHOR 924 925by Yves Orton, 2006. 926 927With excerpts from Perl, and contributions and suggestions from 928Ronald J. Kimball, Dave Mitchell, Dominic Dunlop, Mark Jason Dominus, 929Stephen McCamant, and David Landgren. 930 931Now maintained by Perl 5 Porters. 932 933=head1 LICENCE 934 935Same terms as Perl. 936 937=head1 REFERENCES 938 939[1] L<https://perl.plover.com/Rx/paper/> 940 941[2] L<https://www.unicode.org/> 942 943=cut 944