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