From: Yves Orton Date: Sun, 11 Jun 2006 23:01:14 +0000 (+0200) Subject: Re: [PATCH] regexec/regcomp.c cleanups X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=commitdiff_plain;h=be8e71aa96b3d344129617fadcaca7d958882caa;p=p5sagit%2Fp5-mst-13.2.git Re: [PATCH] regexec/regcomp.c cleanups Message-ID: <9b18b3110606111401o143b2f57rd17bf117979853e7@mail.gmail.com> p4raw-id: //depot/perl@28380 --- diff --git a/ext/re/re.pm b/ext/re/re.pm index c44994e..08abfa1 100644 --- a/ext/re/re.pm +++ b/ext/re/re.pm @@ -73,26 +73,120 @@ See L for additional info. Similarly C produces debugging output, the difference being that it allows the fine tuning of what debugging output will be -emitted. Following the 'Debug' keyword one of several options may be -provided: COMPILE, EXECUTE, TRIE_COMPILE, TRIE_EXECUTE, TRIE_MORE, -OPTIMISE, OFFSETS and ALL. Additionally the special keywords 'All' and -'More' may be provided. 'All' represents everything but OPTIMISE and -OFFSETS and TRIE_MORE, and 'More' is similar but include TRIE_MORE. -Saying C<< no re Debug => 'EXECUTE' >> will disable executing debug -statements and saying C<< use re Debug => 'EXECUTE' >> will turn it on. Note -that these flags can be set directly via ${^RE_DEBUG_FLAGS} by using the -following flag values: - - - RE_DEBUG_COMPILE 0x001 - RE_DEBUG_EXECUTE 0x002 - RE_DEBUG_TRIE_COMPILE 0x004 - RE_DEBUG_TRIE_EXECUTE 0x008 - RE_DEBUG_TRIE_MORE 0x010 - RE_DEBUG_OPTIMISE 0x020 - RE_DEBUG_OFFSETS 0x040 - RE_DEBUG_PARSE 0x080 - RE_DEBUG_OFFSETS_DEBUG 0x100 +emitted. Options are divided into three groups, those related to +compilation, those related to execution and those related to special +purposes. The options are as follows: + +=over 4 + +=item Compile related options + +=over 4 + +=item COMPILE + +Turns on all compile related debug options. + +=item PARSE + +Turns on debug output related to the process of parsing the pattern. + +=item OPTIMISE + +Enables output related to the optimisation phase of compilation. + +=item TRIE_COMPILE + +Detailed info about trie compilation. + +=item DUMP + +Dump the final program out after it is compiled and optimised. + +=item OFFSETS + +Dump offset information. This can be used to see how regops correlate +to the pattern. Output format is + + NODENUM:POSITION[LENGTH] + +Where 1 is the position of the first char in the string. Note that position +can be 0, or larger than the actual length of the pattern, likewise length +can be zero. + +=back + +=item Execute related options + +=over 4 + +=item EXECUTE + +Turns on all execute related debug options. + +=item MATCH + +Turns on debugging of the main matching loop. + +=item TRIE_EXECUTE + +Extra debugging of how tries execute. + +=item INTUIT + +Enable debugging of start point optimisations. + +=back + +=item Extra debugging options + +=over 4 + +=item EXTRA + +Turns on all "extra" debugging options. + +=item TRIE_MORE + +Enable enhanced TRIE debugging. Enhances both TRIE_EXECUTE +and TRIE_COMPILE. + +=item OFFSETS_DEBUG + +Enable debugging of offsets information. This emits copious +amounts of trace information and doesnt mesh well with other +debug options. + +Almost definately only useful to people hacking +on the offsets part of the debug engine. + +=back + +=item Other useful flags + +These are useful shortcuts to save on the typing. + +=over 4 + +=item ALL + +Enable all compile and execute options at once. + +=item All + +Enable DUMP and all execute options. Equivelent to: + + use re 'debug'; + +=item MORE + +=item More + +Enable TRIE_MORE and all execute compile and execute options. + +=back 4 + +=back 4 The directive C and its equivalents are I lexically scoped, as the other directives are. They have both compile-time and run-time @@ -124,21 +218,25 @@ sub setcolor { } my %flags = ( - COMPILE => 1, - EXECUTE => 2, - TRIE_COMPILE => 4, - TRIE_EXECUTE => 8, - TRIE_MORE => 16, - OPTIMISE => 32, - OPTIMIZE => 32, # alias - OFFSETS => 64, - PARSE => 128, - OFFSETS_DEBUG => 256, - OFFSETS_OLD => 576, - ALL => 0xFFFF, - All => 15, - More => 31, + COMPILE => 0x0000FF, + PARSE => 0x000001, + OPTIMISE => 0x000002, + TRIE_COMPILE => 0x000004, + DUMP => 0x000008, + OFFSETS => 0x000010, + + EXECUTE => 0x00FF00, + INTUIT => 0x000100, + MATCH => 0x000200, + TRIE_EXECUTE => 0x000400, + + EXTRA => 0xFF0000, + TRIE_MORE => 0x010000, + OFFSETS_DEBUG => 0x020000, ); +$flags{ALL} = $flags{COMPILE} | $flags{EXECUTE}; +$flags{All} = $flags{all} = $flags{DUMP} | $flags{EXECUTE}; +$flags{More} = $flags{MORE} = $flags{ALL} | $flags{TRIE_MORE}; my $installed = 0; @@ -165,8 +263,6 @@ sub bits { if ($flags{$_[$idx]}) { if ($on) { ${^RE_DEBUG_FLAGS} |= $flags{$_[$idx]}; - ${^RE_DEBUG_FLAGS} |= 1 - if $flags{$_[$idx]}>2; } else { ${^RE_DEBUG_FLAGS} &= ~ $flags{$_[$idx]}; } diff --git a/pod/perlreguts.pod b/pod/perlreguts.pod index ce9f3c0..ef5370f 100644 --- a/pod/perlreguts.pod +++ b/pod/perlreguts.pod @@ -8,42 +8,44 @@ This document is an attempt to shine some light on the guts of the regex engine and how it works. The regex engine represents a signifigant chunk of the perl codebase, but is relatively poorly understood. This document is a meagre attempt at addressing this situation. It is derived from the -authors experience, comments in the source code, other papers on the -regex engine, feedback in p5p, and no doubt other places as well. +author's experience, comments in the source code, other papers on the +regex engine, feedback on the perl5-porters mail list, and no doubt other +places as well. B It should be clearly understood that this document represents the state of the regex engine as the author understands it at -the time of writing. It is B an API definition, it is purely an +the time of writing. It is B an API definition; it is purely an internals guide for those who want to hack the regex engine, or understand how the regex engine works. Readers of this document are -expected to understand perls regex syntax and its usage in detail, if -you are a beginner you are in the wrong the place. +expected to understand perl's regex syntax and its usage in detail. If +you want to learn about the basics of Perl's regular expressions, see +L. =head1 OVERVIEW =head2 A quick note on terms -There is some debate as to whether to say 'regexp' or 'regex'. In this +There is some debate as to whether to say "regexp" or "regex". In this document we will use the term "regex" unless there is a special reason -not to, and then we will explain why. +not to, in which case we will explain why. When speaking about regexes we need to distinguish between their source code form and their internal form. In this document we will use the term "pattern" when we speak of their textual, source code form, the term "program" when we speak of their internal representation. These -correspond to the terms C and C that Mark Jason -Dominus employs in his paper on "Rx"[1]. +correspond to the terms I and I that Mark Jason +Dominus employs in his paper on "Rx" ([1] in L). =head2 What is a regular expression engine? -A regular expression engine is a program whose job is to efficiently -find a section of a string that matches a set criteria of criteria. The -criteria is expressed in text using a formal language. See perlre for a -full definition of the language. +A regular expression engine is a program that takes a set of constraints +specified in a mini-language, and then applies those constraints to a +target string, and determines whether or not the string satisfies the +constraints. See L for a full definition of the language. -So the job in less grandiose terms is to some turn a pattern into +So in less grandiose terms the first part of the job is to turn a pattern into something the computer can efficiently use to find the matching point in -the string. +the string, and the second part is performing the search itself. To do this we need to produce a program by parsing the text. We then need to execute the program to find the point in the string that @@ -53,9 +55,9 @@ matches. And we need to do the whole thing efficiently. =head3 High Level -Although it is a bit confusing and some object to the terminology it +Although it is a bit confusing and some people object to the terminology, it is worth taking a look at a comment that has -been in regexp.h for years: +been in F for years: I can be thought of as the following chart: - [start] - | - - | - +---+---+ - | | | - <\w+> | <\s+> - | <\d+> | - | | | - +---+---+ - | - - | - [end] - -The truth of the matter is that perls regular expressions these days are -way beyond such models, but they can help when trying to get your -bearings, and they do match pretty closely with the current -implementation. - -To be more precise we will say that a regex program is an encoding -of a graph. Each node in the graph corresponds to part of + [start] + | + + | + +-----+-----+ + | | | + <\w+> <\d+> <\s+> + | | | + +-----+-----+ + | + + | + [end] + +The truth of the matter is that perl's regular expressions these days are +much more complex than this kind of structure, but visualizing it this way +can help when trying to get your bearings, and it pretty closely with the +current implementation. + +To be more precise, we will say that a regex program is an encoding +of a graph. Each node in the graph corresponds to part of the original regex pattern, such as a literal string or a branch, and has a pointer to the nodes representing the next component -to be matched. Since "node" and opcode are overloaded terms in the -perl source, we will call the nodes in a regex program 'regops'. +to be matched. Since "node" and "opcode" already have other meanings in the +perl source, we will call the nodes in a regex program "regops". -The program is represented by an array of regnode structures, one or -more of which together represent a single regop of the program. Struct -regnode is the smallest struct needed and has a field structure which is +The program is represented by an array of C structures, one or +more of which represent a single regop of the program. Struct +C is the smallest struct needed and has a field structure which is shared with all the other larger structures. -"Next" pointers of all regops except BRANCH implement concatenation; a -"next" pointer with a BRANCH on both ends of it is connecting two -alternatives. [Here we have one of the subtle syntax dependencies: an -individual BRANCH (as opposed to a collection of them) is never -concatenated with anything because of operator precedence. +The "next" pointers of all regops except C implement concatenation; +a "next" pointer with a C on both ends of it is connecting two +alternatives. [Here we have one of the subtle syntax dependencies: an +individual C (as opposed to a collection of them) is never +concatenated with anything because of operator precedence.] The operand of some types of regop is a literal string; for others, it is a regop leading into a sub-program. In particular, the operand -of a BRANCH node is the first regop of the branch. +of a C node is the first regop of the branch. B: As the railroad metaphor suggests this is B a tree structure: the tail of the branch connects to the thing following the -set of BRANCHes. It is a like a single line of railway track that +set of Ces. It is a like a single line of railway track that splits as it goes into a station or railway yard and rejoins as it comes out the other side. =head3 Regops -The base structure of a regop is defined in regexp.h as follows: +The base structure of a regop is defined in F as follows: struct regnode { - U8 flags; /* Various purposes, sometimes overriden */ + U8 flags; /* Various purposes, sometimes overridden */ U8 type; /* Opcode value as specified by regnodes.h */ U16 next_off; /* Offset in size regnode */ }; -Other larger regnode-like structures are defined in regcomp.h. They +Other larger C-like structures are defined in F. They are almost like subclasses in that they have the same fields as regnode, with possibly additional fields following in the structure, and in some cases the specific meaning (and name) @@ -143,42 +144,42 @@ complete description. =over 4 -=item regnode_1 +=item C -=item regnode_2 +=item C -regnode_1 structures have the same header, followed by a single -four-byte argument; regnode_2 structures contain two two-byte +C structures have the same header, followed by a single +four-byte argument; C structures contain two two-byte arguments instead: regnode_1 U32 arg1; regnode_2 U16 arg1; U16 arg2; -=item regnode_string +=item C -regnode_string structures, used for literal strings, follow the header +C structures, used for literal strings, follow the header with a one-byte length and then the string data. Strings are padded on the end with zero bytes so that the total length of the node is a multiple of four bytes: regnode_string char string[1]; - U8 str_len; (overides flags) + U8 str_len; /* overrides flags */ -=item regnode_charclass +=item C -character classes are represented by regnode_charclass structures, +Character classes are represented by C structures, which have a four-byte argument and then a 32-byte (256-bit) bitmap indicating which characters are included in the class. regnode_charclass U32 arg1; char bitmap[ANYOF_BITMAP_SIZE]; -=item regnode_charclass_class +=item C There is also a larger form of a char class structure used to represent -POSIX char classes called regnode_charclass_class which contains the -same fields plus an additional 4-byte (32-bit) bitmap indicating which -POSIX char class have been included. +POSIX char classes called C which has an +additional 4-byte (32-bit) bitmap indicating which POSIX char class +have been included. regnode_charclass_class U32 arg1; char bitmap[ANYOF_BITMAP_SIZE]; @@ -186,24 +187,24 @@ POSIX char class have been included. =back -regnodes.h defines an array called regarglen[] which gives the size -of each opcode in units of size regnode (4-byte). A macro is used -to calculate the size of an EXACT node based on its C field. +F defines an array called C which gives the size +of each opcode in units of C (4-byte). A macro is used +to calculate the size of an C node based on its C field. -The opcodes are defined in regnodes.h which is generated from -regcomp.sym by regcomp.pl. Currently the maximum possible number -of distinct opcodes is restricted to 256, with about 1/4 already +The regops are defined in F which is generated from +F by F. Currently the maximum possible number +of distinct regops is restricted to 256, with about a quarter already used. -There's a set of macros provided to make accessing the fields -easier and more consistent. These include C which is used to tell -the type of a regnode-like structure, NEXT_OFF() which is the offset to -the next node (more on this later), ARG(), ARG1(), ARG2(), ARG_SET(), -and equivelents for reading and setting the arguments, STR_LEN(), -STRING(), and OPERAND() for manipulating strings and regop bearing +A set of macros makes accessing the fields +easier and more consistent. These include C which is used to determine +the type of a C-like structure, C which is the offset to +the next node (more on this later), C, C, C, C, +and equivelents for reading and setting the arguments, C, +C and C for manipulating strings and regop bearing types. -=head3 What opcode is next? +=head3 What regop is next? There are three distinct concepts of "next" in the regex engine, and it is important to keep them clear. @@ -219,49 +220,68 @@ always be so. =item * -There is the "next opcode" from a given opcode/regnode. This is the -opcode physically located after the the current one, as determined by -the size of the current opcode. This is often useful, such as when +There is the "next regop" from a given regop/regnode. This is the +regop physically located after the the current one, as determined by +the size of the current regop. This is often useful, such as when dumping the structure we use this order to traverse. Sometimes the code -assumes that the "next regnode" is the same as the "next opcode", or in -other words assumes that the sizeof a given opcode type is always going -to be 1 regnode large. +assumes that the "next regnode" is the same as the "next regop", or in +other words assumes that the sizeof a given regop type is always going +to be one regnode large. =item * -There is the "regnext" from a given opcode. This is the opcode which -is reached by jumping forward by the value of NEXT_OFF(), -or in a few cases for longer jumps by the arg1 field of the regnode_1 -structure. The subroutine regnext() handles this transparently. +There is the "regnext" from a given regop. This is the regop which +is reached by jumping forward by the value of C, +or in a few cases for longer jumps by the C field of the C +structure. The subroutine C handles this transparently. This is the logical successor of the node, which in some cases, like -that of the BRANCH opcode, has special meaning. +that of the C regop, has special meaning. =back -=head1 PROCESS OVERVIEW +=head1 Process Overview -Broadly speaking performing a match of a string against a pattern -involves the following steps +Broadly speaking, performing a match of a string against a pattern +involves the following steps: + +=over 5 + +=item A. Compilation + +=over 5 + +=item 1. Parsing for size + +=item 2. Parsing for construction + +=item 3. Peep-hole optimisation and analysis + +=back + +=item B. Execution + +=over 5 + +=item 4. Start position and no-match optimisations + +=item 5. Program execution + +=back + +=back - A. Compilation - 1. Parsing for size - 2. Parsing for construction - 3. Peep-hole Optimisation and Analysis - B. Execution - 4. Start position and no-match optimisations - 5. Program execution Where these steps occur in the actual execution of a perl program is determined by whether the pattern involves interpolating any string -variables. If it does then compilation happens at run time. If it -doesn't then it happens at compile time. (The C modifier changes this, -as does C to a certain extent.) The engine doesn't really care that +variables. If interpolation occurs, then compilation happens at run time. If it +does not, then compilation is performed at compile time. (The C modifier changes this, +as does C to a certain extent). The engine doesn't really care that much. =head2 Compilation -This code exists primarily in regcomp.c, along with the header files -regcomp.h, regexp.h, regnodes.h. +This code resides primarily in F, along with the header files +F, F and F. Compilation starts with C, which is mostly an initialization wrapper which farms out two other routines for the heavy lifting. The @@ -269,30 +289,30 @@ first being C which is the start point for parsing, and C which is responsible for optimisation. Initialization in C mostly involves the creation and data -filling of a special structure RExC_state_t, (defined in regcomp.c). -Almost all internally used routines in regcomp.h take a pointer to one -of these structures as their first argument, with the name *pRExC_state. +filling of a special structure C, (defined in F). +Almost all internally used routines in F take a pointer to one +of these structures as their first argument, with the name C. This structure is used to store the compilation state and contains many -fields. Likewise their are many macros defined which operate on this -variable. Anything that looks like RExC_xxxx is a macro that operates on +fields. Likewise there are many macros which operate on this +variable. Anything that looks like C is a macro that operates on this pointer/structure. =head3 Parsing for size In this pass the input pattern is parsed in order to calculate how much -space is needed for each opcode we would need to emit. The size is also +space is needed for each regop we would need to emit. The size is also used to determine whether long jumps will be required in the program. -This stage is controlled by the macro SIZE_ONLY being set. +This stage is controlled by the macro C being set. The parse procedes pretty much exactly as it does during the -construction phase except that most routines are shortcircuited to -change the size field RExC_size and not actually do anything. +construction phase, except that most routines are short-circuited to +change the size field C and not do anything else. =head3 Parsing for construcution -Once the size of the program has been determine the pattern is parsed -again, but this time for real. Now SIZE_ONLY will be false, and the +Once the size of the program has been determined, the pattern is parsed +again, but this time for real. Now C will be false, and the actual construction can occur. C is the start of the parse process. It is responsible for @@ -300,31 +320,31 @@ parsing an arbitrary chunk of pattern up to either the end of the string, or the first closing parenthesis it encounters in the pattern. This means it can be used to parse the toplevel regex, or any section inside of a grouping parenthesis. It also handles the "special parens" -that perls regexes have. For instance when parsing C C -will at one point be called to parse from the '?' symbol up to and -including the ')'. +that perl's regexes have. For instance when parsing C C +will at one point be called to parse from the "?" symbol up to and +including the ")". -Additionally C is responsible for parsing the one or more +Additionally, C is responsible for parsing the one or more branches from the pattern, and for "finishing them off" by correctly -setting their next pointers. In order to do the parsing it repeatedly -calls out to C which is responsible for handling up to the +setting their next pointers. In order to do the parsing, it repeatedly +calls out to C, which is responsible for handling up to the first C<|> symbol it sees. -C in turn calls C which is responsible for -handling "things" followed by a quantifier. In order to parse the -"things" C is called. This is the lowest level routine which -is responsible for parsing out constant strings, char classes, and the -various special symbols like C<$>. If C encounters a '(' +C in turn calls C which +handles "things" followed by a quantifier. In order to parse the +"things", C is called. This is the lowest level routine which +parses out constant strings, character classes, and the +various special symbols like C<$>. If C encounters a "(" character it in turn calls C. The routine C is called by both C, C in order to "set the tail pointer" correctly. When executing and -we get to the end of a branch we need to go to node following the -grouping parens. When parsing however we don't know where the end will +we get to the end of a branch, we need to go to the node following the +grouping parens. When parsing, however, we don't know where the end will be until we get there, so when we do we must go back and update the offsets as appropriate. C is used to make this easier. -A subtlety of the parse process means that a regex like C is +A subtlety of the parsing process means that a regex like C is originally parsed into an alternation with a single branch. It is only afterwards that the optimizer converts single branch alternations into the simpler form. @@ -387,13 +407,13 @@ The resulting program then looks like: 3: END(0) As you can see, even though we parsed out a branch and a piece, it was ultimately -only an atom. The final program shows us how things work. We have an EXACT regop, -followed by an END regop. The number in parens indicates where the 'regnext' of -the node goes. The 'regnext' of an END regop is unused, as END regops mean +only an atom. The final program shows us how things work. We have an C regop, +followed by an C regop. The number in parens indicates where the C of +the node goes. The C of an C regop is unused, as C regops mean we have successfully matched. The number on the left indicates the position of the regop in the regnode array. -Now lets try a harder pattern. We will add a quantifier so we have the pattern +Now let's try a harder pattern. We will add a quantifier, so now we have the pattern C. We will see that C calls C regpiece twice. >foo+< 1 reg @@ -414,10 +434,10 @@ And we end up with the program: 4: EXACT (0) 6: END(0) -Now we have a special case. The EXACT regop has a regnext of 0. This is +Now we have a special case. The C regop has a C of 0. This is because if it matches it should try to match itself again. The PLUS regop -handles the actual failure of the EXACT regop and acts appropriately (going -to regnode 6 if the EXACT matched at least once, or failing if it didn't.) +handles the actual failure of the C regop and acts appropriately (going +to regnode 6 if the C matched at least once, or failing if it didn't.) Now for something much more complex: C @@ -516,17 +536,17 @@ Resulting in the program 37: END(0) Here we can see a much more complex program, with various optimisations in -play. At regnode 10 we can see an example where a char class with only -one character in it was turned into an EXACT node. We can also see where -an entire alternation was turned into a TRIE-EXACT node. As a consequence +play. At regnode 10 we see an example where a character class with only +one character in it was turned into an C node. We can also see where +an entire alternation was turned into a C node. As a consequence, some of the regnodes have been marked as optimised away. We can see that -the C<$> symbol has been converted into an EOL regop, a special piece of -code that looks for \n or the end of a string. +the C<$> symbol has been converted into an C regop, a special piece of +code that looks for C<\n> or the end of the string. -The next pointer for BRANCHes is interesting in that it points at where +The next pointer for Ces is interesting in that it points at where execution should go if the branch fails. When executing if the engine -tries to traverse from a branch to a regnext that isnt a branch then -the engine will know the overall series of branches have failed. +tries to traverse from a branch to a C that isn't a branch then +the engine will know that the entire set of branches have failed. =head3 Peep-hole Optimisation and Analysis @@ -539,18 +559,18 @@ Consider a situation like the following pattern. The C<(a|b)*> part can match at every char in the string, and then fail every time because there is no C in the string. So obviously we can -not bother to use the regex engine unless there is a 'z' in the string. +avoid using the regex engine unless there is a 'z' in the string. Likewise in a pattern like: /foo(\w+)bar/ In this case we know that the string must contain a C which must be followed by C. We can use Fast Boyer-More matching as implemented -in fbm_instr() to find the location of these strings. If they dont exist -then we dont need to resort to the much more expensive regex engine. -Even better if they do exist then we can use their positions to +in C to find the location of these strings. If they don't exist +then we don't need to resort to the much more expensive regex engine. +Even better, if they do exist then we can use their positions to reduce the search space that the regex engine needs to cover to determine -if the entire pattern does match. +if the entire pattern matches. There are various aspects of the pattern that can be used to facilitate optimisations along these lines: @@ -562,24 +582,24 @@ optimisations along these lines: * Beginning/End of line positions Another form of optimisation that can occur is post-parse "peep-hole" -optimisations, where inefficient constructs are modified so that they -are more efficient. An example of this is TAIL regops which are used +optimisations, where inefficient constructs are replaced by +more efficient constructs. An example of this are C regops which are used during parsing to mark the end of branches and the end of groups. These regops are used as place holders during construction and "always match" so they can be "optimised away" by making the things that point to the -TAIL point to thing the TAIL points to, in essence "skipping" the node. +TAIL point to thing that the C points to, thus "skipping" the node. -Another optimisation that can occur is that of "EXACT merging" which is -where two consecutive EXACT nodes are merged into a single more efficient -to execute regop. An even more agressive form of this is that a branch -sequence of the form EXACT BRANCH ... EXACT can be converted into a TRIE -regop. +Another optimisation that can occur is that of "C merging" which is +where two consecutive C nodes are merged into a single +regop. An even more agressive form of this is that a branch +sequence of the form CEXACT BRANCH ... EXACT> can be converted into a +C regop. -All of this occurs in the routine study_chunk() which uses a special -structure scan_data_t to store the analysis that it has performed, and -as it goes does the "peep-hole" optimisations. +All of this occurs in the routine C which uses a special +structure C to store the analysis that it has performed, and +does the "peep-hole" optimisations as it goes. -The code involved in study_chunk() is extremely cryptic. Be careful. :-) +The code involved in C is extremely cryptic. Be careful. :-) =head2 Execution @@ -587,18 +607,18 @@ Execution of a regex generally involves two phases, the first being finding the start point in the string where we should match from, and the second being running the regop interpreter. -If we can tell that there is no valid start point we don't bother running -interpreter at all. Likewise if we know from the analysis phase that we -can not optimise detection of the start position we go straight to the +If we can tell that there is no valid start point then we don't bother running +interpreter at all. Likewise, if we know from the analysis phase that we +cannot detect a short-cut to the start position, we go straight to the interpreter. -The two entry points are re_intuit_start() and pregexec(). These routines +The two entry points are C and C. These routines have a somewhat incestuous relationship with overlap between their functions, -and pregexec() may even call re_intuit_start() on its own. Nevertheless +and C may even call C on its own. Nevertheless the perl source code may call into either, or both. Execution of the interpreter itself used to be recursive. Due to the -efforts of Dave Mitchel in blead perl it no longer is. Instead an +efforts of Dave Mitchell in blead perl, it is now iterative. Now an internal stack is maintained on the heap and the routine is fully iterative. This can make it tricky as the code is quite conservative about what state it stores which means that two consecutive lines in the @@ -607,9 +627,9 @@ simulated recursion. =head3 Start position and no-match optimisations -re_intuit_start() is responsible for handling start point and no match +C is responsible for handling start points and no match optimisations as determined by the results of the analysis done by -study_chunk() (and described in L). +C (and described in L). The basic structure of this routine is to try to find the start and/or end points of where the pattern could match, and to ensure that the string @@ -620,23 +640,23 @@ For instance it may try to determine that a given fixed string must be not only present but a certain number of chars before the end of the string, or whatever. -It calls out into several other routines, like fbm_instr() which does -"Fast Boyer More" matching and find_byclass() which is responsible for +It calls several other routines, such as C which does +"Fast Boyer More" matching and C which is responsible for finding the start using the first mandatory regop in the program. -When the optimisation criteria have been satisfied reg_try() is called +When the optimisation criteria have been satisfied C is called to perform the match. =head3 Program execution C is the main entry point for running a regex. It contains support for initializing the regex interpreters state, running -re_intuit_start() if needed, and running the intepreter on the string +C if needed, and running the intepreter on the string from various start positions as needed. When its necessary to use the regex interpreter C calls C. C is the entry point into the regex interpreter. It expects -as arguments a pointer to a regmatch_info structure and a pointer to +as arguments a pointer to a C structure and a pointer to a string. It returns an integer 1 for success and a 0 for failure. It is basically a setup wrapper around C. @@ -649,32 +669,32 @@ the bulk are inline code. =head2 UNICODE and Localization Support -No matter how you look at it unicode support is going to be a pain in a +No matter how you look at it, Unicode support is going to be a pain in a regex engine. Tricks that might be fine when you have 256 possible -characters often won't scale to handle the size of the 'utf8' character +characters often won't scale to handle the size of the UTF-8 character set. Things you can take for granted with ASCII may not be true with -unicode. For instance in ASCII its safe to assume that -C, in utf8 it isn't. Unicode case folding is -vastly more complex than the simple rules of English, and even when not -using unicode but only localized single byte encodings things can get -tricky (technically GERMAN-SHARP-ESS should match 'ss' in localized case -insensitive matching.) - -Making things worse is that C support was a later addition to the -regex engine (as it was to perl) and necessarily this made things a lot -more complicated. Obviously its easier to design a regex engine with -unicode support from the beginning than it is to retrofit one that -wasn't designed with it in mind. - -Pretty well every regop that involves looking at the input string has -two cases, one for 'utf8' and one not. In fact its often more complex -than that, as the pattern may be 'utf8' as well. +unicode. For instance, in ASCII, it is safe to assume that +C, but in UTF-8 it isn't. Unicode case folding is +vastly more complex than the simple rules of ASCII, and even when not +using Unicode but only localized single byte encodings, things can get +tricky (for example, GERMAN-SHARP-ESS should match 'ss' in localized case +insensitive matching). + +Making things worse is that UTF-8 support was a later addition to the +regex engine (as it was to perl) and this necessarily made things a lot +more complicated. Obviously it is easier to design a regex engine with +Unicode support in mind from the beginning than it is to retrofit it to +one that wasn't. + +Nearly all regops that involves looking at the input string have +two cases, one for UTF-8, and one not. In fact, it's often more complex +than that, as the pattern may be UTF-8 as well. Care must be taken when making changes to make sure that you handle -utf8 properly both at compile time and at execution time, including +UTF-8 properly, both at compile time and at execution time, including when the string and pattern are mismatched. -The following comment in regcomp.h gives an example of exactly how +The following comment in F gives an example of exactly how tricky this can be: Two problematic code points in Unicode casefolding of EXACT nodes: @@ -703,13 +723,70 @@ tricky this can be: A sequence of valid UTF-8 bytes cannot be a subsequence of another valid sequence of UTF-8 bytes. +=head3 Base Struct + +regexp.h contains the base structure definition: + + typedef struct regexp { + I32 *startp; + I32 *endp; + regnode *regstclass; + struct reg_substr_data *substrs; + char *precomp; /* pre-compilation regular expression */ + struct reg_data *data; /* Additional data. */ + char *subbeg; /* saved or original string + so \digit works forever. */ + #ifdef PERL_OLD_COPY_ON_WRITE + SV *saved_copy; /* If non-NULL, SV which is COW from original */ + #endif + U32 *offsets; /* offset annotations 20001228 MJD */ + I32 sublen; /* Length of string pointed by subbeg */ + I32 refcnt; + I32 minlen; /* mininum possible length of $& */ + I32 prelen; /* length of precomp */ + U32 nparens; /* number of parentheses */ + U32 lastparen; /* last paren matched */ + U32 lastcloseparen; /* last paren matched */ + U32 reganch; /* Internal use only + + Tainted information used by regexec? */ + regnode program[1]; /* Unwarranted chumminess with compiler. */ + } regexp; + +C, and C are the primary fields of concern in terms of +program structure. program is the actual array of nodes, and data is +an array of "whatever", with each whatever being typed by letter, and +freed or cloned as needed based on this type. regops use the data +array to store reference data that isn't convenient to store in the regop +itself. It also means memory management code doesnt need to traverse the +program to find pointers. So for instance if a regop needs a pointer, the +normal procedure is use a regnode_arg1 store the data index in the ARG +field and look it up from the data array. + +startp,endp,nparens,lasparen,lastcloseparen are used to manage capture +buffers. + +subbeg and optional saved_copy are used during exectuion phase for managing +replacements. + +offsets and precomp are used for debugging purposes. + +And the rest are used for start point optimisations. + + +=head2 Deallocation and Cloning + +Any patch that adds data items to the regexp will need to include +changes to sv.c (Perl_re_dup) and regcomp.c (pregfree). This +involves freeing or cloning items in the regexes data array based +on the data items type. + =head1 AUTHOR by Yves Orton, 2006. With excerpts from Perl, and contributions and suggestions from Ronald J. Kimball, Dave Mitchell, Dominic Dunlop, Mark Jason Dominus, -and Stephen McCamant. +Stephen McCamant, and David Landgren. =head1 LICENSE diff --git a/regcomp.c b/regcomp.c index df5d890..6165317 100644 --- a/regcomp.c +++ b/regcomp.c @@ -123,7 +123,7 @@ typedef struct RExC_state_t { #define RExC_starttry (pRExC_state->starttry) #endif #ifdef DEBUGGING - char *lastparse; + const char *lastparse; I32 lastnum; #define RExC_lastparse (pRExC_state->lastparse) #define RExC_lastnum (pRExC_state->lastnum) @@ -859,7 +859,7 @@ S_dump_trie(pTHX_ const struct _reg_trie_data *trie,U32 depth) "Match","Base","Ofs" ); for( state = 0 ; state < trie->uniquecharcount ; state++ ) { - SV **tmp = av_fetch( trie->revcharmap, state, 0); + SV ** const tmp = av_fetch( trie->revcharmap, state, 0); if ( tmp ) { PerlIO_printf( Perl_debug_log, "%4.4s ", SvPV_nolen_const( *tmp ) ); } @@ -872,7 +872,7 @@ S_dump_trie(pTHX_ const struct _reg_trie_data *trie,U32 depth) PerlIO_printf( Perl_debug_log, "\n"); for( state = 1 ; state < TRIE_LASTSTATE(trie) ; state++ ) { - const U32 base = trie->states[ state ].trans.base; + const U32 base = trie->states[ state ].trans.base; PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state); @@ -942,7 +942,7 @@ S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie, U32 next_alloc ); } for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) { - SV **tmp = av_fetch( trie->revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0); + SV ** const tmp = av_fetch( trie->revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0); PerlIO_printf( Perl_debug_log, "%s:%3X=%4"UVXf" | ", SvPV_nolen_const( *tmp ), TRIE_LIST_ITEM(state,charid).forid, @@ -975,7 +975,7 @@ S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie, U32 next_allo PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" ); for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) { - SV **tmp = av_fetch( trie->revcharmap, charid, 0); + SV ** const tmp = av_fetch( trie->revcharmap, charid, 0); if ( tmp ) { PerlIO_printf( Perl_debug_log, "%4.4s ", SvPV_nolen_const( *tmp ) ); } @@ -1042,14 +1042,13 @@ S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode /* add a fail transition */ reg_trie_data *trie=(reg_trie_data *)RExC_rx->data->data[ARG(source)]; U32 *q; - U32 ucharcount = trie->uniquecharcount; - U32 numstates = trie->laststate; - U32 ubound = trie->lasttrans + ucharcount; + const U32 ucharcount = trie->uniquecharcount; + const U32 numstates = trie->laststate; + const U32 ubound = trie->lasttrans + ucharcount; U32 q_read = 0; U32 q_write = 0; U32 charid; U32 base = trie->states[ 1 ].trans.base; - U32 newstate; U32 *fail; reg_ac_data *aho; const U32 data_slot = add_data( pRExC_state, 1, "T" ); @@ -1067,21 +1066,20 @@ S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode fail[ 0 ] = fail[ 1 ] = 1; for ( charid = 0; charid < ucharcount ; charid++ ) { - newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 ); - if ( newstate ) - { + const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 ); + if ( newstate ) { q[ q_write ] = newstate; /* set to point at the root */ fail[ q[ q_write++ ] ]=1; } } while ( q_read < q_write) { - U32 cur = q[ q_read++ % numstates ]; - U32 ch_state; + const U32 cur = q[ q_read++ % numstates ]; base = trie->states[ cur ].trans.base; for ( charid = 0 ; charid < ucharcount ; charid++ ) { - if ( ( ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 ) ) ) { + const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 ); + if (ch_state) { U32 fail_state = cur; U32 fail_base; do { @@ -1257,7 +1255,8 @@ S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *firs PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n", (int)depth * 2 + 2,"", ( trie->widecharmap ? "UTF8" : "NATIVE" ), TRIE_WORDCOUNT(trie), - (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount, trie->minlen, trie->maxlen ) + (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount, + (int)trie->minlen, (int)trie->maxlen ) ); Newxz( trie->wordlen, TRIE_WORDCOUNT(trie), U32 ); @@ -1731,9 +1730,9 @@ S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *firs "%*sNew Start State=%"UVuf" Class: [", (int)depth * 2 + 2, "", state)); - if (idx>-1) { - SV **tmp = av_fetch( TRIE_REVCHARMAP(trie), idx, 0); - const U8 *ch = (U8*)SvPV_nolen_const( *tmp ); + if (idx >= 0) { + SV ** const tmp = av_fetch( TRIE_REVCHARMAP(trie), idx, 0); + const U8 * const ch = (U8*)SvPV_nolen_const( *tmp ); TRIE_BITMAP_SET(trie,*ch); if ( folder ) @@ -1857,7 +1856,7 @@ S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *firs if (PL_regkind[OP(scan)] == EXACT) \ join_exact(pRExC_state,(scan),(min),(flags),NULL,depth+1) -U32 +STATIC U32 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, I32 *min, U32 flags,regnode *val, U32 depth) { /* Merge several consecutive EXACTish nodes into one. */ regnode *n = regnext(scan); @@ -3053,12 +3052,13 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, I32 *deltap, } scan->flags = (U8)minnext; } - if (data && data_fake.flags & (SF_HAS_PAR|SF_IN_PAR)) - pars++; - if (data && (data_fake.flags & SF_HAS_EVAL)) - data->flags |= SF_HAS_EVAL; - if (data) + if (data) { + if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR)) + pars++; + if (data_fake.flags & SF_HAS_EVAL) + data->flags |= SF_HAS_EVAL; data->whilem_c = data_fake.whilem_c; + } if (f & SCF_DO_STCLASS_AND) { const int was = (data->start_class->flags & ANYOF_EOS); @@ -3197,7 +3197,7 @@ Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm) I32 sawopen = 0; scan_data_t data; RExC_state_t RExC_state; - RExC_state_t *pRExC_state = &RExC_state; + RExC_state_t * const pRExC_state = &RExC_state; #ifdef TRIE_STUDY_OPT int restudied= 0; RExC_state_t copyRExC_state; @@ -3434,8 +3434,20 @@ reStudy: r->reganch |= ROPT_SKIP; /* Scan is after the zeroth branch, first is atomic matcher. */ - DEBUG_COMPILE_r(PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n", - (IV)(first - scan + 1))); +#ifdef TRIE_STUDY_OPT + DEBUG_COMPILE_r( + if (!restudied) + PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n", + (IV)(first - scan + 1)) + ); +#else + DEBUG_COMPILE_r( + PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n", + (IV)(first - scan + 1)) + ); +#endif + + /* * If there's something expensive in the r.e., find the * longest literal string that must appear and make it the @@ -3640,9 +3652,8 @@ reStudy: Newxz(r->endp, RExC_npar, I32); DEBUG_r( RX_DEBUG_on(r) ); - DEBUG_COMPILE_r({ - if (SvIV(re_debug_flags)> (RE_DEBUG_COMPILE | RE_DEBUG_EXECUTE)) - PerlIO_printf(Perl_debug_log,"Final program:\n"); + DEBUG_DUMP_r({ + PerlIO_printf(Perl_debug_log,"Final program:\n"); regdump(r); }); return(r); @@ -3673,11 +3684,11 @@ reStudy: else \ num=REG_NODE_NUM(RExC_emit); \ if (RExC_lastnum!=num) \ - PerlIO_printf(Perl_debug_log,"%4d",num); \ + PerlIO_printf(Perl_debug_log,"|%4d",num); \ else \ - PerlIO_printf(Perl_debug_log,"%4s",""); \ - PerlIO_printf(Perl_debug_log,"%*s%-4s", \ - (int)(10+(depth*2)), "", \ + PerlIO_printf(Perl_debug_log,"|%4s",""); \ + PerlIO_printf(Perl_debug_log,"|%*s%-4s", \ + (int)((depth*2)), "", \ (funcname) \ ); \ RExC_lastnum=num; \ @@ -6351,14 +6362,8 @@ Perl_regdump(pTHX_ const regexp *r) U32 i; PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)r->offsets[0]); for (i = 1; i <= len; i++) { - if (!(SvIV(re_debug_flags) & RE_DEBUG_OLD_OFFSETS)) { - if (r->offsets[i*2-1] || r->offsets[i*2]) - PerlIO_printf(Perl_debug_log, "%"UVuf":",i); - else - continue; - } - PerlIO_printf(Perl_debug_log, "%"UVuf"[%"UVuf"] ", - (UV)r->offsets[i*2-1], (UV)r->offsets[i*2]); + PerlIO_printf(Perl_debug_log, "%"UVuf":%"UVuf"[%"UVuf"] ", + i, (UV)r->offsets[i*2-1], (UV)r->offsets[i*2]); } PerlIO_printf(Perl_debug_log, "\n"); }); @@ -6605,7 +6610,7 @@ Perl_pregfree(pTHX_ struct regexp *r) if (!r || (--r->refcnt > 0)) return; - DEBUG_r(if (re_debug_flags && (SvIV(re_debug_flags) & RE_DEBUG_COMPILE)) { + DEBUG_COMPILE_r(if (RX_DEBUG(r)){ const char * const s = (r->reganch & ROPT_UTF8) ? pv_uni_display(dsv, (U8*)r->precomp, r->prelen, 60, UNI_DISPLAY_REGEX) : pv_display(dsv, r->precomp, r->prelen, 0, 60); @@ -6678,7 +6683,8 @@ Perl_pregfree(pTHX_ struct regexp *r) case 'n': break; case 'T': - { + { /* Aho Corasick add-on structure for a trie node. + Used in stclass optimization only */ U32 refcount; reg_ac_data *aho=(reg_ac_data*)r->data->data[n]; OP_REFCNT_LOCK; @@ -6690,11 +6696,13 @@ Perl_pregfree(pTHX_ struct regexp *r) aho->trie=NULL; /* not necessary to free this as it is handled by the 't' case */ Safefree(r->data->data[n]); /* do this last!!!! */ + Safefree(r->regstclass); } } break; case 't': { + /* trie structure. */ U32 refcount; reg_trie_data *trie=(reg_trie_data*)r->data->data[n]; OP_REFCNT_LOCK; @@ -6711,10 +6719,12 @@ Perl_pregfree(pTHX_ struct regexp *r) if (trie->wordlen) Safefree(trie->wordlen); #ifdef DEBUGGING - if (trie->words) - SvREFCNT_dec((SV*)trie->words); - if (trie->revcharmap) - SvREFCNT_dec((SV*)trie->revcharmap); + if (RX_DEBUG(r)) { + if (trie->words) + SvREFCNT_dec((SV*)trie->words); + if (trie->revcharmap) + SvREFCNT_dec((SV*)trie->revcharmap); + } #endif Safefree(r->data->data[n]); /* do this last!!!! */ } @@ -6889,7 +6899,7 @@ S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node, /* Where, what. */ if (OP(node) == OPTIMIZED) { - if (!optstart && (SvIV(re_debug_flags) & RE_DEBUG_OPTIMISE)) + if (!optstart && (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_OPTIMISE)) optstart = node; else goto after_print; @@ -6910,14 +6920,18 @@ S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node, after_print: if (PL_regkind[(U8)op] == BRANCHJ) { - register const regnode *nnode = (OP(next) == LONGJMP + assert(next); + { + register const regnode *nnode = (OP(next) == LONGJMP ? regnext((regnode *)next) : next); - if (last && nnode > last) - nnode = last; - DUMPUNTIL(r, start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1); + if (last && nnode > last) + nnode = last; + DUMPUNTIL(r, start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1); + } } else if (PL_regkind[(U8)op] == BRANCH) { + assert(next); DUMPUNTIL(r, start, NEXTOPER(node), next, sv, l + 1); } else if ( PL_regkind[(U8)op] == TRIE ) { @@ -6934,7 +6948,8 @@ S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node, (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount, (IV)TRIE_LASTSTATE(trie)-1, - trie->minlen, trie->maxlen + (int)trie->minlen, + (int)trie->maxlen ); if (trie->bitmap) { int i; @@ -6981,6 +6996,7 @@ S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node, NEXTOPER(node) + EXTRA_STEP_2ARGS + 1, sv, l + 1); } else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) { + assert(next); DUMPUNTIL(r, start, NEXTOPER(node) + EXTRA_STEP_2ARGS, next, sv, l + 1); } diff --git a/regcomp.h b/regcomp.h index d06c767..8363637 100644 --- a/regcomp.h +++ b/regcomp.h @@ -325,7 +325,11 @@ struct regnode_charclass_class { /* has [[:blah:]] classes */ START_EXTERN_C +#ifdef PLUGGABLE_RE_EXTENSION +#include "re_nodes.h" +#else #include "regnodes.h" +#endif /* The following have no fixed length. U8 so we can do strchr() on it. */ #ifndef DOINIT @@ -513,55 +517,105 @@ typedef struct _reg_ac_data reg_ac_data; #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF" #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS" -/* If you change these be sure to update ext/re/re.pm as well */ -#define RE_DEBUG_COMPILE 0x0001 -#define RE_DEBUG_EXECUTE 0x0002 -#define RE_DEBUG_TRIE_COMPILE 0x0004 -#define RE_DEBUG_TRIE_EXECUTE 0x0008 -#define RE_DEBUG_TRIE_MORE 0x0010 -#define RE_DEBUG_OPTIMISE 0x0020 -#define RE_DEBUG_OFFSETS 0x0040 -#define RE_DEBUG_PARSE 0x0080 -#define RE_DEBUG_OFFSETS_DEBUG 0x0100 -#define RE_DEBUG_OLD_OFFSETS 0x0200 - - -#define DEBUG_PARSE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_PARSE) x ) -#define DEBUG_OPTIMISE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_OPTIMISE) x ) -#define DEBUG_EXECUTE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_EXECUTE) x ) -#define DEBUG_COMPILE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE) x ) -#define DEBUG_OFFSETS_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_OFFSETS) x ) -#define DEBUG_OLD_OFFSETS_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_OLD_OFFSETS) x ) +/* + +RE_DEBUG_FLAGS is used to control what debug output is emitted +its divided into three groups of options, some of which interact. +The three groups are: Compile, Execute, Extra. There is room for a +further group, as currently only the low three bytes are used. + + Compile Options: -#define DEBUG_TRIE_r(x) DEBUG_r( \ - if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_COMPILE \ - || SvIV(re_debug_flags) & RE_DEBUG_TRIE_EXECUTE ) \ - x \ -) -#define DEBUG_TRIE_EXECUTE_r(x) \ - DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_EXECUTE) x ) + PARSE + PEEP + TRIE + PROGRAM + OFFSETS -#define DEBUG_TRIE_COMPILE_r(x) \ - DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_COMPILE) x ) + Execute Options: -#define DEBUG_TRIE_EXECUTE_MORE_r(x) \ - DEBUG_TRIE_EXECUTE_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_MORE) x ) + INTUIT + MATCH + TRIE -#define DEBUG_TRIE_COMPILE_MORE_r(x) \ - DEBUG_TRIE_COMPILE_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_MORE) x ) + Extra Options -/* get_sv() can return NULL during global destruction. */ -#define GET_RE_DEBUG_FLAGS DEBUG_r( \ - re_debug_flags=get_sv(RE_DEBUG_FLAGS, 1); \ - if (re_debug_flags && !SvIOK(re_debug_flags)) { \ - sv_setiv(re_debug_flags, RE_DEBUG_COMPILE | RE_DEBUG_EXECUTE ); \ - } \ - ) + TRIE + OFFSETS + +If you modify any of these make sure you make corresponding changes to +re.pm, especially to the documentation. +*/ + + +/* Compile */ +#define RE_DEBUG_COMPILE_MASK 0x0000FF +#define RE_DEBUG_COMPILE_PARSE 0x000001 +#define RE_DEBUG_COMPILE_OPTIMISE 0x000002 +#define RE_DEBUG_COMPILE_TRIE 0x000004 +#define RE_DEBUG_COMPILE_DUMP 0x000008 +#define RE_DEBUG_COMPILE_OFFSETS 0x000010 + +/* Execute */ +#define RE_DEBUG_EXECUTE_MASK 0x00FF00 +#define RE_DEBUG_EXECUTE_INTUIT 0x000100 +#define RE_DEBUG_EXECUTE_MATCH 0x000200 +#define RE_DEBUG_EXECUTE_TRIE 0x000400 + +/* Extra */ +#define RE_DEBUG_EXTRA_MASK 0xFF0000 +#define RE_DEBUG_EXTRA_TRIE 0x010000 +#define RE_DEBUG_EXTRA_OFFSETS 0x020000 + +/* Compile */ +#define DEBUG_COMPILE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_MASK) x ) +#define DEBUG_PARSE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_PARSE) x ) +#define DEBUG_OPTIMISE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_OPTIMISE) x ) +#define DEBUG_PARSE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_PARSE) x ) +#define DEBUG_DUMP_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_DUMP) x ) +#define DEBUG_OFFSETS_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_OFFSETS) x ) +#define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE_TRIE) x ) + +/* Execute */ +#define DEBUG_EXECUTE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_EXECUTE_MASK) x ) +#define DEBUG_INTUIT_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_EXECUTE_INTUIT) x ) +#define DEBUG_MATCH_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_EXECUTE_MATCH) x ) +#define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_EXECUTE_TRIE) x ) + +/* Extra */ +#define DEBUG_EXTRA_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_EXTRA_MASK) x ) #define MJD_OFFSET_DEBUG(x) DEBUG_r( \ - if (SvIV(re_debug_flags) & RE_DEBUG_OFFSETS_DEBUG) \ - Perl_warn_nocontext x \ -) + if (SvIV(re_debug_flags) & RE_DEBUG_EXTRA_OFFSETS) \ + Perl_warn_nocontext x ) +#define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_EXTRA_TRIE) x ) +#define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \ + if (SvIV(re_debug_flags) & RE_DEBUG_EXTRA_TRIE) x ) + +#define DEBUG_TRIE_r(x) DEBUG_r( \ + if (SvIV(re_debug_flags) & (RE_DEBUG_COMPILE_TRIE \ + | RE_DEBUG_EXECUTE_TRIE )) x ) + +/* initialization */ +/* get_sv() can return NULL during global destruction. */ +#define GET_RE_DEBUG_FLAGS DEBUG_r( \ + re_debug_flags = get_sv(RE_DEBUG_FLAGS, 1); \ + if (re_debug_flags && !SvIOK(re_debug_flags)) { \ + sv_setiv(re_debug_flags, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \ + } ) #ifdef DEBUGGING #define GET_RE_DEBUG_FLAGS_DECL SV *re_debug_flags = NULL; GET_RE_DEBUG_FLAGS; @@ -569,4 +623,3 @@ typedef struct _reg_ac_data reg_ac_data; #define GET_RE_DEBUG_FLAGS_DECL #endif - diff --git a/regexec.c b/regexec.c index 5338e79..a853b85 100644 --- a/regexec.c +++ b/regexec.c @@ -569,7 +569,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, { char * const last = HOP3c(s, -start_shift, strbeg); char *last1, *last2; - char *s1 = s; + char * const saved_s = s; SV* must; t = s - prog->check_offset_max; @@ -617,7 +617,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, ", trying floating at offset %ld...\n", - (long)(HOP3c(s1, 1, strend) - i_strpos))); + (long)(HOP3c(saved_s, 1, strend) - i_strpos))); other_last = HOP3c(last1, prog->anchored_offset+1, strend); s = HOP3c(last, 1, strend); goto restart; @@ -627,7 +627,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, (long)(s - i_strpos))); t = HOP3c(s, -prog->anchored_offset, strbeg); other_last = HOP3c(s, 1, strend); - s = s1; + s = saved_s; if (t == strpos) goto try_at_start; goto try_at_offset; @@ -636,7 +636,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } else { /* Take into account the floating substring. */ char *last, *last1; - char *s1 = s; + char * const saved_s = s; SV* must; t = HOP3c(s, -start_shift, strbeg); @@ -675,7 +675,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, ", trying anchored starting at offset %ld...\n", - (long)(s1 + 1 - i_strpos))); + (long)(saved_s + 1 - i_strpos))); other_last = last; s = HOP3c(t, 1, strend); goto restart; @@ -684,7 +684,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", (long)(s - i_strpos))); other_last = s; /* Fix this later. --Hugo */ - s = s1; + s = saved_s; if (t == strpos) goto try_at_start; goto try_at_offset; @@ -1581,11 +1581,26 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, const char *last_start = strend - trie->minlen; const char *real_start = s; STRLEN maxlen = trie->maxlen; - U8 **points; + SV *sv_points; + U8 **points; /* map of where we were in the input string + when reading a given string. For ASCII this + is unnecessary overhead as the relationship + is always 1:1, but for unicode, especially + case folded unicode this is not true. */ GET_RE_DEBUG_FLAGS_DECL; - Newxz(points,maxlen,U8 *); + /* We can't just allocate points here. We need to wrap it in + * an SV so it gets freed properly if there is a croak while + * running the match */ + ENTER; + SAVETMPS; + sv_points=newSV(maxlen * sizeof(U8 *)); + SvCUR_set(sv_points, + maxlen * sizeof(U8 *)); + SvPOK_on(sv_points); + sv_2mortal(sv_points); + points=(U8**)SvPV_nolen(sv_points ); if (trie->bitmap && trie_type != trie_utf8_fold) { while (!TRIE_BITMAP_TEST(trie,*s) && s <= last_start ) { @@ -1720,13 +1735,18 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, ); if (leftmost) { s = (char*)leftmost; - if (!reginfo || regtry(reginfo, s)) + if (!reginfo || regtry(reginfo, s)) { + FREETMPS; + LEAVE; goto got_it; + } s = HOPc(s,1); } else { break; } } + FREETMPS; + LEAVE; } break; default: @@ -2059,13 +2079,13 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } else if ((c = prog->regstclass)) { if (minlen) { - U8 op = OP(prog->regstclass); + const OPCODE op = OP(prog->regstclass); /* don't bother with what can't match */ if (PL_regkind[op] != EXACT && op != CANY && op != TRIE) strend = HOPc(strend, -(minlen - 1)); } DEBUG_EXECUTE_r({ - SV *prop = sv_newmortal(); + SV * const prop = sv_newmortal(); const char *s0; const char *s1; int len0; @@ -2177,7 +2197,7 @@ got_it: if ( !(flags & REXEC_NOT_FIRST) ) { RX_MATCH_COPY_FREE(prog); if (flags & REXEC_COPY_STR) { - I32 i = PL_regeol - startpos + (stringarg - strbeg); + const I32 i = PL_regeol - startpos + (stringarg - strbeg); #ifdef PERL_OLD_COPY_ON_WRITE if ((SvIsCOW(sv) || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) { @@ -2272,7 +2292,7 @@ S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) Newxz(PL_reg_curpm, 1, PMOP); #ifdef USE_ITHREADS { - SV* repointer = newSViv(0); + SV* const repointer = newSViv(0); /* so we know which PL_regex_padav element is PL_reg_curpm */ SvFLAGS(repointer) |= SVf_BREAK; av_push(PL_regex_padav,repointer); @@ -2857,7 +2877,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) : trie_plain; /* what trie are we using right now */ - reg_trie_data *trie + reg_trie_data * const trie = (reg_trie_data*)rex->data->data[ ARG( scan ) ]; U32 state = trie->startstate; @@ -3021,7 +3041,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) if ( st->u.trie.accepted == 1 ) { DEBUG_EXECUTE_r({ - SV ** const tmp = RX_DEBUG(reginfo->prog) + SV ** const tmp = RX_DEBUG(reginfo->prog) ? av_fetch( trie->words, st->u.trie.accept_buff[ 0 ].wordnum-1, 0 ) : NULL; PerlIO_printf( Perl_debug_log, @@ -3064,7 +3084,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) DEBUG_EXECUTE_r({ reg_trie_data * const trie = (reg_trie_data*) rex->data->data[ARG(scan)]; - SV ** const tmp = RX_DEBUG(reginfo->prog) + SV ** const tmp = RX_DEBUG(reginfo->prog) ? av_fetch( trie->words, st->u.trie.accept_buff[ best ].wordnum - 1, 0 ) : NULL; PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at node #%d %s\n", @@ -3108,7 +3128,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) if (do_utf8 != UTF) { /* The target and the pattern have differing utf8ness. */ char *l = locinput; - const char *e = s + st->ln; + const char * const e = s + st->ln; if (do_utf8) { /* The target is utf8, the pattern is not utf8. */ @@ -3158,12 +3178,12 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case EXACTF: { - char *s = STRING(scan); + char * const s = STRING(scan); st->ln = STR_LEN(scan); if (do_utf8 || UTF) { /* Either target or the pattern are utf8. */ - char *l = locinput; + const char * const l = locinput; char *e = PL_regeol; if (ibcmp_utf8(s, 0, st->ln, (bool)UTF, @@ -3552,7 +3572,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) * necessary */ MAGIC *mg = NULL; - SV *sv; + const SV *sv; if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret))) mg = mg_find(sv, PERL_MAGIC_qr); else if (SvSMAGICAL(ret)) { @@ -4017,8 +4037,7 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) inner = NEXTOPER(scan); do_branch: { - I32 type; - type = OP(scan); + const I32 type = OP(scan); if (!next || OP(next) != type) /* No choice. */ next = inner; /* Avoid recursion. */ else { @@ -4253,7 +4272,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) st->u.plus.c1 = st->u.plus.c2 = CHRTEST_VOID; goto assume_ok_easy; } - else { s = (U8*)STRING(text_node); } + else + s = (U8*)STRING(text_node); if (!UTF) { st->u.plus.c2 = st->u.plus.c1 = *s; @@ -4389,16 +4409,14 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) else c = UCHARAT(PL_reginput); /* If it could work, try it. */ - if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) - { + if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) { TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2); /*** all unsaved local vars undefined at this point */ REGCP_UNWIND(st->u.plus.lastcp); } } /* If it could work, try it. */ - else if (st->u.plus.c1 == CHRTEST_VOID) - { + else if (st->u.plus.c1 == CHRTEST_VOID) { TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3); /*** all unsaved local vars undefined at this point */ REGCP_UNWIND(st->u.plus.lastcp); @@ -4428,8 +4446,8 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) } REGCP_SET(st->u.plus.lastcp); { - UV c = 0; while (n >= st->ln) { + UV c = 0; if (st->u.plus.c1 != CHRTEST_VOID) { if (do_utf8) c = utf8n_to_uvchr((U8*)PL_reginput, @@ -4439,12 +4457,11 @@ S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) c = UCHARAT(PL_reginput); } /* If it could work, try it. */ - if (st->u.plus.c1 == CHRTEST_VOID || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) - { - TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4); - /*** all unsaved local vars undefined at this point */ - REGCP_UNWIND(st->u.plus.lastcp); - } + if (st->u.plus.c1 == CHRTEST_VOID || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) { + TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4); + /*** all unsaved local vars undefined at this point */ + REGCP_UNWIND(st->u.plus.lastcp); + } /* Couldn't or didn't -- back up. */ n--; PL_reginput = locinput = HOPc(locinput, -1); @@ -5111,22 +5128,22 @@ S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max) PL_reginput = scan; DEBUG_r({ - SV *re_debug_flags = NULL; - SV * const prop = sv_newmortal(); - GET_RE_DEBUG_FLAGS; - DEBUG_EXECUTE_r({ - regprop(prog, prop, p); - PerlIO_printf(Perl_debug_log, - "%*s %s can match %"IVdf" times out of %"IVdf"...\n", - REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max); - }); + SV *re_debug_flags = NULL; + SV * const prop = sv_newmortal(); + GET_RE_DEBUG_FLAGS; + DEBUG_EXECUTE_r({ + regprop(prog, prop, p); + PerlIO_printf(Perl_debug_log, + "%*s %s can match %"IVdf" times out of %"IVdf"...\n", + REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max); }); + }); return(c); } -#ifndef PERL_IN_XSUB_RE +#if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) /* - regclass_swash - prepare the utf8 swash */ diff --git a/regexp.h b/regexp.h index 777d82f..4ea7c69 100644 --- a/regexp.h +++ b/regexp.h @@ -14,7 +14,9 @@ * Caveat: this is V8 regexp(3) [actually, a reimplementation thereof], * not the System V one. */ - +#ifndef PLUGGABLE_RE_EXTENSION +/* we don't want to include this stuff if we are inside Nicholas' + * pluggable regex engine code */ struct regnode { U8 flags; @@ -80,22 +82,17 @@ typedef struct regexp { #define RE_USE_INTUIT_NOML 0x00100000 /* Best to intuit before matching */ #define RE_USE_INTUIT_ML 0x00200000 #define REINT_AUTORITATIVE_NOML 0x00400000 /* Can trust a positive answer */ -#define REINT_AUTORITATIVE_ML 0x00800000 +#define REINT_AUTORITATIVE_ML 0x00800000 #define REINT_ONCE_NOML 0x01000000 /* Intuit can succed once only. */ #define REINT_ONCE_ML 0x02000000 #define RE_INTUIT_ONECHAR 0x04000000 #define RE_INTUIT_TAIL 0x08000000 -#define RE_DEBUG_BIT 0x20000000 #define RE_USE_INTUIT (RE_USE_INTUIT_NOML|RE_USE_INTUIT_ML) #define REINT_AUTORITATIVE (REINT_AUTORITATIVE_NOML|REINT_AUTORITATIVE_ML) #define REINT_ONCE (REINT_ONCE_NOML|REINT_ONCE_ML) -#define RX_DEBUG(prog) ((prog)->reganch & RE_DEBUG_BIT) -#define RX_DEBUG_on(prog) ((prog)->reganch |= RE_DEBUG_BIT) - - #define RX_MATCH_TAINTED(prog) ((prog)->reganch & ROPT_TAINTED_SEEN) #define RX_MATCH_TAINTED_on(prog) ((prog)->reganch |= ROPT_TAINTED_SEEN) #define RX_MATCH_TAINTED_off(prog) ((prog)->reganch &= ~ROPT_TAINTED_SEEN) @@ -109,6 +106,13 @@ typedef struct regexp { #define RX_MATCH_COPIED_set(prog,t) ((t) \ ? RX_MATCH_COPIED_on(prog) \ : RX_MATCH_COPIED_off(prog)) +#endif /* PLUGGABLE_RE_EXTENSION */ + +/* Stuff that needs to be included in the plugable extension goes below here */ + +#define RE_DEBUG_BIT 0x20000000 +#define RX_DEBUG(prog) ((prog)->reganch & RE_DEBUG_BIT) +#define RX_DEBUG_on(prog) ((prog)->reganch |= RE_DEBUG_BIT) #ifdef PERL_OLD_COPY_ON_WRITE #define RX_MATCH_COPY_FREE(rx) \ diff --git a/sv.c b/sv.c index 9f8ada1..029d0fa 100644 --- a/sv.c +++ b/sv.c @@ -9536,12 +9536,22 @@ Perl_re_dup(pTHX_ const REGEXP *r, CLONE_PARAMS *param) d->data[i] = r->data->data[i]; break; case 't': - case 'T': d->data[i] = r->data->data[i]; OP_REFCNT_LOCK; ((reg_trie_data*)d->data[i])->refcount++; OP_REFCNT_UNLOCK; break; + case 'T': + d->data[i] = r->data->data[i]; + OP_REFCNT_LOCK; + ((reg_ac_data*)d->data[i])->refcount++; + OP_REFCNT_UNLOCK; + /* Trie stclasses are readonly and can thus be shared + * without duplication. We free the stclass in pregfree + * when the corresponding reg_ac_data struct is freed. + */ + ret->regstclass= r->regstclass; + break; default: Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]); }