"Fossies" - the Fresh Open Source Software Archive

Member "install-tl-20200916/tlpkg/tlperl/lib/CORE/regcomp.h" (7 Mar 2020, 48923 Bytes) of package /windows/misc/install-tl.zip:


As a special service "Fossies" has tried to format the requested source page into HTML format using (guessed) C and C++ source code syntax highlighting (style: standard) with prefixed line numbers and code folding option. Alternatively you can here view or download the uninterpreted source code file.

    1 /*    regcomp.h
    2  *
    3  *    Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
    4  *    2000, 2001, 2002, 2003, 2005, 2006, 2007, by Larry Wall and others
    5  *
    6  *    You may distribute under the terms of either the GNU General Public
    7  *    License or the Artistic License, as specified in the README file.
    8  *
    9  */
   10 
   11 #ifndef PERL_REGCOMP_H_
   12 #define PERL_REGCOMP_H_
   13 
   14 #include "regcharclass.h"
   15 
   16 /* Convert branch sequences to more efficient trie ops? */
   17 #define PERL_ENABLE_TRIE_OPTIMISATION 1
   18 
   19 /* Be really aggressive about optimising patterns with trie sequences? */
   20 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
   21 
   22 /* Should the optimiser take positive assertions into account? */
   23 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
   24 
   25 /* Not for production use: */
   26 #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
   27 
   28 /* Activate offsets code - set to if 1 to enable */
   29 #ifdef DEBUGGING
   30 #define RE_TRACK_PATTERN_OFFSETS
   31 #endif
   32 
   33 /*
   34  * The "internal use only" fields in regexp.h are present to pass info from
   35  * compile to execute that permits the execute phase to run lots faster on
   36  * simple cases.  They are:
   37  *
   38  * regstart sv that must begin a match; NULL if none obvious
   39  * reganch  is the match anchored (at beginning-of-line only)?
   40  * regmust  string (pointer into program) that match must include, or NULL
   41  *  [regmust changed to SV* for bminstr()--law]
   42  * regmlen  length of regmust string
   43  *  [regmlen not used currently]
   44  *
   45  * Regstart and reganch permit very fast decisions on suitable starting points
   46  * for a match, cutting down the work a lot.  Regmust permits fast rejection
   47  * of lines that cannot possibly match.  The regmust tests are costly enough
   48  * that pregcomp() supplies a regmust only if the r.e. contains something
   49  * potentially expensive (at present, the only such thing detected is * or +
   50  * at the start of the r.e., which can involve a lot of backup).  Regmlen is
   51  * supplied because the test in pregexec() needs it and pregcomp() is computing
   52  * it anyway.
   53  * [regmust is now supplied always.  The tests that use regmust have a
   54  * heuristic that disables the test if it usually matches.]
   55  *
   56  * [In fact, we now use regmust in many cases to locate where the search
   57  * starts in the string, so if regback is >= 0, the regmust search is never
   58  * wasted effort.  The regback variable says how many characters back from
   59  * where regmust matched is the earliest possible start of the match.
   60  * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.]
   61  */
   62 
   63 /*
   64  * Structure for regexp "program".  This is essentially a linear encoding
   65  * of a nondeterministic finite-state machine (aka syntax charts or
   66  * "railroad normal form" in parsing technology).  Each node is an opcode
   67  * plus a "next" pointer, possibly plus an operand.  "Next" pointers of
   68  * all nodes except BRANCH implement concatenation; a "next" pointer with
   69  * a BRANCH on both ends of it is connecting two alternatives.  (Here we
   70  * have one of the subtle syntax dependencies:  an individual BRANCH (as
   71  * opposed to a collection of them) is never concatenated with anything
   72  * because of operator precedence.)  The operand of some types of node is
   73  * a literal string; for others, it is a node leading into a sub-FSM.  In
   74  * particular, the operand of a BRANCH node is the first node of the branch.
   75  * (NB this is *not* a tree structure:  the tail of the branch connects
   76  * to the thing following the set of BRANCHes.)  The opcodes are defined
   77  * in regnodes.h which is generated from regcomp.sym by regcomp.pl.
   78  */
   79 
   80 /*
   81  * A node is one char of opcode followed by two chars of "next" pointer.
   82  * "Next" pointers are stored as two 8-bit pieces, high order first.  The
   83  * value is a positive offset from the opcode of the node containing it.
   84  * An operand, if any, simply follows the node.  (Note that much of the
   85  * code generation knows about this implicit relationship.)
   86  *
   87  * Using two bytes for the "next" pointer is vast overkill for most things,
   88  * but allows patterns to get big without disasters.
   89  *
   90  * [The "next" pointer is always aligned on an even
   91  * boundary, and reads the offset directly as a short.]
   92  */
   93 
   94 /* This is the stuff that used to live in regexp.h that was truly
   95    private to the engine itself. It now lives here. */
   96 
   97  typedef struct regexp_internal {
   98         int name_list_idx;  /* Optional data index of an array of paren names */
   99         union {
  100         U32 *offsets;           /* offset annotations 20001228 MJD
  101                                        data about mapping the program to the
  102                                        string -
  103                                        offsets[0] is proglen when this is used
  104                                        */
  105             U32 proglen;
  106         } u;
  107 
  108         regnode *regstclass;    /* Optional startclass as identified or constructed
  109                                    by the optimiser */
  110         struct reg_data *data;  /* Additional miscellaneous data used by the program.
  111                                    Used to make it easier to clone and free arbitrary
  112                                    data that the regops need. Often the ARG field of
  113                                    a regop is an index into this structure */
  114     struct reg_code_blocks *code_blocks;/* positions of literal (?{}) */
  115     regnode program[1]; /* Unwarranted chumminess with compiler. */
  116 } regexp_internal;
  117 
  118 #define RXi_SET(x,y) (x)->pprivate = (void*)(y)   
  119 #define RXi_GET(x)   ((regexp_internal *)((x)->pprivate))
  120 #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
  121 /*
  122  * Flags stored in regexp->intflags
  123  * These are used only internally to the regexp engine
  124  *
  125  * See regexp.h for flags used externally to the regexp engine
  126  */
  127 #define RXp_INTFLAGS(rx)        ((rx)->intflags)
  128 #define RX_INTFLAGS(prog)        RXp_INTFLAGS(ReANY(prog))
  129 
  130 #define PREGf_SKIP      0x00000001
  131 #define PREGf_IMPLICIT      0x00000002 /* Converted .* to ^.* */
  132 #define PREGf_NAUGHTY       0x00000004 /* how exponential is this pattern? */
  133 #define PREGf_VERBARG_SEEN  0x00000008
  134 #define PREGf_CUTGROUP_SEEN 0x00000010
  135 #define PREGf_USE_RE_EVAL   0x00000020 /* compiled with "use re 'eval'" */
  136 /* these used to be extflags, but are now intflags */
  137 #define PREGf_NOSCAN            0x00000040
  138                                 /* spare */
  139 #define PREGf_GPOS_SEEN         0x00000100
  140 #define PREGf_GPOS_FLOAT        0x00000200
  141 
  142 #define PREGf_ANCH_MBOL         0x00000400
  143 #define PREGf_ANCH_SBOL         0x00000800
  144 #define PREGf_ANCH_GPOS         0x00001000
  145 #define PREGf_RECURSE_SEEN      0x00002000
  146 
  147 #define PREGf_ANCH              \
  148     ( PREGf_ANCH_SBOL | PREGf_ANCH_GPOS | PREGf_ANCH_MBOL )
  149 
  150 /* this is where the old regcomp.h started */
  151 
  152 struct regnode_string {
  153     U8  str_len;
  154     U8  type;
  155     U16 next_off;
  156     char string[1];
  157 };
  158 
  159 /* Argument bearing node - workhorse, 
  160    arg1 is often for the data field */
  161 struct regnode_1 {
  162     U8  flags;
  163     U8  type;
  164     U16 next_off;
  165     U32 arg1;
  166 };
  167 
  168 /* Similar to a regnode_1 but with an extra signed argument */
  169 struct regnode_2L {
  170     U8  flags;
  171     U8  type;
  172     U16 next_off;
  173     U32 arg1;
  174     I32 arg2;
  175 };
  176 
  177 /* 'Two field' -- Two 16 bit unsigned args */
  178 struct regnode_2 {
  179     U8  flags;
  180     U8  type;
  181     U16 next_off;
  182     U16 arg1;
  183     U16 arg2;
  184 };
  185 
  186 /* This give the number of code points that can be in the bitmap of an ANYOF
  187  * node.  The shift number must currently be one of: 8..12.  It can't be less
  188  * than 8 (256) because some code relies on it being at least that.  Above 12
  189  * (4096), and you start running into warnings that some data structure widths
  190  * have been exceeded, though the test suite as of this writing still passes
  191  * for up through 16, which is as high as anyone would ever want to go,
  192  * encompassing all of the Unicode BMP, and thus including all the economically
  193  * important world scripts.  At 12 most of them are: including Arabic,
  194  * Cyrillic, Greek, Hebrew, Indian subcontinent, Latin, and Thai; but not Han,
  195  * Japanese, nor Korean.  (The regarglen structure in regnodes.h is a U8, and
  196  * the trie types TRIEC and AHOCORASICKC are larger than U8 for shift values
  197  * above 12.)  Be sure to benchmark before changing, as larger sizes do
  198  * significantly slow down the test suite */
  199 #define NUM_ANYOF_CODE_POINTS   (1 << 8)
  200 
  201 #define ANYOF_BITMAP_SIZE   (NUM_ANYOF_CODE_POINTS / 8)   /* 8 bits/Byte */
  202 
  203 /* Note that these form structs which are supersets of the next smaller one, by
  204  * appending fields.  Alignment problems can occur if one of those optional
  205  * fields requires stricter alignment than the base struct.  And formal
  206  * parameters that can really be two or more of the structs should be
  207  * declared as the smallest one it could be.  See commit message for
  208  * 7dcac5f6a5195002b55c935ee1d67f67e1df280b.  Regnode allocation is done
  209  * without regard to alignment, and changing it to would also require changing
  210  * the code that inserts and deletes regnodes.  The basic single-argument
  211  * regnode has a U32, which is what reganode() allocates as a unit.  Therefore
  212  * no field can require stricter alignment than U32. */
  213 
  214 /* also used by trie */
  215 struct regnode_charclass {
  216     U8  flags;
  217     U8  type;
  218     U16 next_off;
  219     U32 arg1;                           /* set by set_ANYOF_arg() */
  220     char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
  221 };
  222 
  223 /* has runtime (locale) \d, \w, ..., [:posix:] classes */
  224 struct regnode_charclass_posixl {
  225     U8  flags;                      /* ANYOF_MATCHES_POSIXL bit must go here */
  226     U8  type;
  227     U16 next_off;
  228     U32 arg1;
  229     char bitmap[ANYOF_BITMAP_SIZE];     /* both compile-time ... */
  230     U32 classflags;                         /* and run-time */
  231 };
  232 
  233 /* A synthetic start class (SSC); is a regnode_charclass_posixl_fold, plus an
  234  * extra SV*, used only during its construction and which is not used by
  235  * regexec.c.  Note that the 'next_off' field is unused, as the SSC stands
  236  * alone, so there is never a next node.  Also, there is no alignment issue,
  237  * because these are declared or allocated as a complete unit so the compiler
  238  * takes care of alignment.  This is unlike the other regnodes which are
  239  * allocated in terms of multiples of a single-argument regnode.  SSC nodes can
  240  * have a pointer field because there is no alignment issue, and because it is
  241  * set to NULL after construction, before any cloning of the pattern */
  242 struct regnode_ssc {
  243     U8  flags;                      /* ANYOF_MATCHES_POSIXL bit must go here */
  244     U8  type;
  245     U16 next_off;
  246     U32 arg1;
  247     char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
  248     U32 classflags;                 /* ... and run-time */
  249 
  250     /* Auxiliary, only used during construction; NULL afterwards: list of code
  251      * points matched */
  252     SV* invlist;
  253 };
  254 
  255 /*  We take advantage of 'next_off' not otherwise being used in the SSC by
  256  *  actually using it: by setting it to 1.  This allows us to test and
  257  *  distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
  258  *  otherwise be 1, because it is the offset to the next regnode expressed in
  259  *  units of regnodes.  Since an ANYOF node contains extra fields, it adds up
  260  *  to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
  261  *  not 0) is 11 there.  Even if things get tightly packed on a 64-bit system,
  262  *  it still would be more than 1. */
  263 #define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF;              \
  264                                            NEXT_OFF(n) = 1;            \
  265                                } STMT_END
  266 #define is_ANYOF_SYNTHETIC(n) (PL_regkind[OP(n)] == ANYOF && NEXT_OFF(n) == 1)
  267 
  268 /* XXX fix this description.
  269    Impose a limit of REG_INFTY on various pattern matching operations
  270    to limit stack growth and to avoid "infinite" recursions.
  271 */
  272 /* The default size for REG_INFTY is U16_MAX, which is the same as
  273    USHORT_MAX (see perl.h).  Unfortunately U16 isn't necessarily 16 bits
  274    (see handy.h).  On the Cray C90, sizeof(short)==4 and hence U16_MAX is
  275    ((1<<32)-1), while on the Cray T90, sizeof(short)==8 and U16_MAX is
  276    ((1<<64)-1).  To limit stack growth to reasonable sizes, supply a
  277    smaller default.
  278     --Andy Dougherty  11 June 1998
  279 */
  280 #if SHORTSIZE > 2
  281 #  ifndef REG_INFTY
  282 #    define REG_INFTY ((1<<16)-1)
  283 #  endif
  284 #endif
  285 
  286 #ifndef REG_INFTY
  287 #  define REG_INFTY U16_MAX
  288 #endif
  289 
  290 #define ARG_VALUE(arg) (arg)
  291 #define ARG__SET(arg,val) ((arg) = (val))
  292 
  293 #undef ARG
  294 #undef ARG1
  295 #undef ARG2
  296 
  297 #define ARG(p) ARG_VALUE(ARG_LOC(p))
  298 #define ARG1(p) ARG_VALUE(ARG1_LOC(p))
  299 #define ARG2(p) ARG_VALUE(ARG2_LOC(p))
  300 #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
  301 
  302 #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
  303 #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
  304 #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
  305 #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
  306 
  307 #undef NEXT_OFF
  308 #undef NODE_ALIGN
  309 
  310 #define NEXT_OFF(p) ((p)->next_off)
  311 #define NODE_ALIGN(node)
  312 /* the following define was set to 0xde in 075abff3
  313  * as part of some linting logic. I have set it to 0
  314  * as otherwise in every place where we /might/ set flags
  315  * we have to set it 0 explicitly, which duplicates
  316  * assignments and IMO adds an unacceptable level of
  317  * surprise to working in the regex engine. If this
  318  * is changed from 0 then at the very least make sure
  319  * that SBOL for /^/ sets the flags to 0 explicitly.
  320  * -- Yves */
  321 #define NODE_ALIGN_FILL(node) ((node)->flags = 0)
  322 
  323 #define SIZE_ALIGN NODE_ALIGN
  324 
  325 #undef OP
  326 #undef OPERAND
  327 #undef MASK
  328 #undef STRING
  329 
  330 #define OP(p)       ((p)->type)
  331 #define FLAGS(p)    ((p)->flags)    /* Caution: Doesn't apply to all      \
  332                        regnode types.  For some, it's the \
  333                        character set of the regnode */
  334 #define OPERAND(p)  (((struct regnode_string *)p)->string)
  335 #define MASK(p)     ((char*)OPERAND(p))
  336 #define STR_LEN(p)  (((struct regnode_string *)p)->str_len)
  337 #define STRING(p)   (((struct regnode_string *)p)->string)
  338 #define STR_SZ(l)   ((l + sizeof(regnode) - 1) / sizeof(regnode))
  339 #define NODE_SZ_STR(p)  (STR_SZ(STR_LEN(p))+1)
  340 
  341 #undef NODE_ALIGN
  342 #undef ARG_LOC
  343 #undef NEXTOPER
  344 #undef PREVOPER
  345 
  346 #define NODE_ALIGN(node)
  347 #define ARG_LOC(p)  (((struct regnode_1 *)p)->arg1)
  348 #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
  349 #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
  350 #define ARG2L_LOC(p)    (((struct regnode_2L *)p)->arg2)
  351 
  352 #define NODE_STEP_REGNODE   1   /* sizeof(regnode)/sizeof(regnode) */
  353 #define EXTRA_STEP_2ARGS    EXTRA_SIZE(struct regnode_2)
  354 
  355 #define NODE_STEP_B 4
  356 
  357 #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
  358 #define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
  359 
  360 #define FILL_NODE(offset, op)                                           \
  361     STMT_START {                                                        \
  362                     OP(REGNODE_p(offset)) = op;                         \
  363                     NEXT_OFF(REGNODE_p(offset)) = 0;                    \
  364     } STMT_END
  365 #define FILL_ADVANCE_NODE(offset, op)                                   \
  366     STMT_START {                                                        \
  367                     FILL_NODE(offset, op);                              \
  368                     (offset)++;                                         \
  369     } STMT_END
  370 #define FILL_ADVANCE_NODE_ARG(offset, op, arg)                          \
  371     STMT_START {                                                        \
  372                     ARG_SET(REGNODE_p(offset), arg);                    \
  373                     FILL_ADVANCE_NODE(offset, op);                      \
  374                     /* This is used generically for other operations    \
  375                      * that have a longer argument */                   \
  376                     (offset) += regarglen[op];                          \
  377     } STMT_END
  378 #define FILL_ADVANCE_NODE_2L_ARG(offset, op, arg1, arg2)                \
  379     STMT_START {                                                        \
  380                     ARG_SET(REGNODE_p(offset), arg1);                   \
  381                     ARG2L_SET(REGNODE_p(offset), arg2);                 \
  382                     FILL_ADVANCE_NODE(offset, op);                      \
  383                     (offset) += 2;                                      \
  384     } STMT_END
  385 
  386 #define REG_MAGIC 0234
  387 
  388 /* An ANYOF node is basically a bitmap with the index being a code point.  If
  389  * the bit for that code point is 1, the code point matches;  if 0, it doesn't
  390  * match (complemented if inverted).  There is an additional mechanism to deal
  391  * with cases where the bitmap is insufficient in and of itself.  This #define
  392  * indicates if the bitmap does fully represent what this ANYOF node can match.
  393  * The ARG is set to this special value (since 0, 1, ... are legal, but will
  394  * never reach this high). */
  395 #define ANYOF_ONLY_HAS_BITMAP   ((U32) -1)
  396 
  397 /* When the bitmap isn't completely sufficient for handling the ANYOF node,
  398  * flags (in node->flags of the ANYOF node) get set to indicate this.  These
  399  * are perennially in short supply.  Beyond several cases where warnings need
  400  * to be raised under certain circumstances, currently, there are six cases
  401  * where the bitmap alone isn't sufficient.  We could use six flags to
  402  * represent the 6 cases, but to save flags bits, we play some games.  The
  403  * cases are:
  404  *
  405  *  1)  The bitmap has a compiled-in very finite size.  So something else needs
  406  *      to be used to specify if a code point that is too large for the bitmap
  407  *      actually matches.  The mechanism currently is an inversion
  408  *      list.  ANYOF_ONLY_HAS_BITMAP, described above, being TRUE indicates
  409  *      there are no matches of too-large code points.  But if it is FALSE,
  410  *      then almost certainly there are matches too large for the bitmap.  (The
  411  *      other cases, described below, either imply this one or are extremely
  412  *      rare in practice.)  So we can just assume that a too-large code point
  413  *      will need something beyond the bitmap if ANYOF_ONLY_HAS_BITMAP is
  414  *      FALSE, instead of having a separate flag for this.
  415  *  2)  A subset of item 1) is if all possible code points outside the bitmap
  416  *      match.  This is a common occurrence when the class is complemented,
  417  *      like /[^ij]/.  Therefore a bit is reserved to indicate this,
  418  *      rather than having a more expensive inversion list created,
  419  *      ANYOF_MATCHES_ALL_ABOVE_BITMAP.
  420  *  3)  Under /d rules, it can happen that code points that are in the upper
  421  *      latin1 range (\x80-\xFF or their equivalents on EBCDIC platforms) match
  422  *      only if the runtime target string being matched against is UTF-8.  For
  423  *      example /[\w[:punct:]]/d.  This happens only for posix classes (with a
  424  *      couple of exceptions, like \d where it doesn't happen), and all such
  425  *      ones also have above-bitmap matches.  Thus, 3) implies 1) as well.
  426  *      Note that /d rules are no longer encouraged; 'use 5.14' or higher
  427  *      deselects them.  But a flag is required so that they can be properly
  428  *      handled.  But it can be a shared flag: see 5) below.
  429  *  4)  Also under /d rules, something like /[\Wfoo]/ will match everything in
  430  *      the \x80-\xFF range, unless the string being matched against is UTF-8.
  431  *      An inversion list could be created for this case, but this is
  432  *      relatively common, and it turns out that it's all or nothing:  if any
  433  *      one of these code points matches, they all do.  Hence a single bit
  434  *      suffices.  We use a shared flag that doesn't take up space by itself:
  435  *      ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER.  This
  436  *      also implies 1), with one exception: [:^cntrl:].
  437  *  5)  A user-defined \p{} property may not have been defined by the time the
  438  *      regex is compiled.  In this case, we don't know until runtime what it
  439  *      will match, so we have to assume it could match anything, including
  440  *      code points that ordinarily would be in the bitmap.  A flag bit is
  441  *      necessary to indicate this , though it can be shared with the item 3)
  442  *      flag, as that only occurs under /d, and this only occurs under non-d.
  443  *      This case is quite uncommon in the field, and the /(?[ ...])/ construct
  444  *      is a better way to accomplish what this feature does.  This case also
  445  *      implies 1).
  446  *      ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
  447  *      is the shared flag.
  448  *  6)  /[foo]/il may have folds that are only valid if the runtime locale is a
  449  *      UTF-8 one.  These are quite rare, so it would be good to avoid the
  450  *      expense of looking for them.  But /l matching is slow anyway, and we've
  451  *      traditionally not worried too much about its performance.  And this
  452  *      condition requires the ANYOFL_FOLD flag to be set, so testing for
  453  *      that flag would be sufficient to rule out most cases of this.  So it is
  454  *      unclear if this should have a flag or not.  But, this flag can be
  455  *      shared with another, so it doesn't occupy extra space.
  456  *
  457  * At the moment, there is one spare bit, but this could be increased by
  458  * various tricks:
  459  *
  460  * If just one more bit is needed, as of this writing it seems to khw that the
  461  * best choice would be to make ANYOF_MATCHES_ALL_ABOVE_BITMAP not a flag, but
  462  * something like
  463  *
  464  *      #define ANYOF_MATCHES_ALL_ABOVE_BITMAP      ((U32) -2)
  465  *
  466  * and access it through the ARG like ANYOF_ONLY_HAS_BITMAP is.  This flag is
  467  * used by all ANYOF node types, and it could be used to avoid calling the
  468  * handler function, as the macro REGINCLASS in regexec.c does now for other
  469  * cases.
  470  *
  471  * Another possibility is based on the fact that ANYOF_MATCHES_POSIXL is
  472  * redundant with the node type ANYOFPOSIXL.  That flag could be removed, but
  473  * at the expense of extra code in regexec.c.  The flag has been retained
  474  * because it allows us to see if we need to call reginsert, or just use the
  475  * bitmap in one test.
  476  *
  477  * If this is done, an extension would be to make all ANYOFL nodes contain the
  478  * extra 32 bits that ANYOFPOSIXL ones do.  The posix flags only occupy 30
  479  * bits, so the ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD flags
  480  * and ANYOFL_FOLD could be moved to that extra space, but it would mean extra
  481  * instructions, as there are currently places in the code that assume those
  482  * two bits are zero.
  483  *
  484  * All told, 5 bits could be available for other uses if all of the above were
  485  * done.
  486  *
  487  * Some flags are not used in synthetic start class (SSC) nodes, so could be
  488  * shared should new flags be needed for SSCs, like SSC_MATCHES_EMPTY_STRING
  489  * now. */
  490 
  491 /* If this is set, the result of the match should be complemented.  regexec.c
  492  * is expecting this to be in the low bit.  Never in an SSC */
  493 #define ANYOF_INVERT                        0x01
  494 
  495 /* For the SSC node only, which cannot be inverted, so is shared with that bit.
  496  * This is used only during regex compilation. */
  497 #define SSC_MATCHES_EMPTY_STRING                ANYOF_INVERT
  498 
  499 /* Set if this is a regnode_charclass_posixl vs a regnode_charclass.  This
  500  * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
  501  * and the optimizer's synthetic start class.  Non-locale \d, etc are resolved
  502  * at compile-time.  Only set under /l; can be in SSC */
  503 #define ANYOF_MATCHES_POSIXL                    0x02
  504 
  505 /* The fold is calculated and stored in the bitmap where possible at compile
  506  * time.  However under locale, the actual folding varies depending on
  507  * what the locale is at the time of execution, so it has to be deferred until
  508  * then.  Only set under /l; never in an SSC  */
  509 #define ANYOFL_FOLD                             0x04
  510 
  511 /* Shared bit set only with ANYOFL and SSC nodes:
  512  *    If ANYOFL_FOLD is set, this flag indicates there are potential matches
  513  *      valid only if the locale is a UTF-8 one.
  514  *    If ANYOFL_FOLD is NOT set, this flag means to warn if the runtime locale
  515  *       isn't a UTF-8 one (and the generated node assumes a UTF-8 locale).
  516  *       None of INVERT, POSIXL,
  517  *       ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP
  518  *       can be set.  */
  519 #define ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD        0x08
  520 
  521 /* Convenience macros for teasing apart the meanings when reading the above bit
  522  * */
  523 #define ANYOFL_SOME_FOLDS_ONLY_IN_UTF8_LOCALE(flags)                        \
  524     ((flags & ( ANYOFL_FOLD /* Both bits are set */                         \
  525                |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD))   \
  526              == ( ANYOFL_FOLD                                               \
  527                  |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD))
  528 
  529 #define  ANYOFL_UTF8_LOCALE_REQD(flags)                                     \
  530     ((flags & ( ANYOFL_FOLD /* Only REQD bit is set */                      \
  531                |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD))   \
  532              == ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)
  533 
  534 /* Spare: Be sure to change ANYOF_FLAGS_ALL if this gets used  0x10 */
  535 
  536 /* If set, the node matches every code point NUM_ANYOF_CODE_POINTS and above.
  537  * Can be in an SSC */
  538 #define ANYOF_MATCHES_ALL_ABOVE_BITMAP          0x20
  539 
  540 /* Shared bit:
  541  *      Under /d it means the ANYOFD node matches more things if the target
  542  *          string is encoded in UTF-8; any such things will be non-ASCII,
  543  *          characters that are < 256, and can be accessed via the inversion
  544  *          list.
  545  *      When not under /d, it means the ANYOF node contains a user-defined
  546  *      property that wasn't yet defined at the time the regex was compiled,
  547  *      and so must be looked up at runtime, by creating an inversion list.
  548  * (These uses are mutually exclusive because a user-defined property is
  549  * specified by \p{}, and \p{} implies /u which deselects /d).  The long macro
  550  * name is to make sure that you are cautioned about its shared nature.  Only
  551  * the non-/d meaning can be in an SSC */
  552 #define ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP  0x40
  553 
  554 /* Shared bit:
  555  *      Under /d it means the ANYOFD node matches all non-ASCII Latin1
  556  *          characters when the target string is not in utf8.
  557  *      When not under /d, it means the ANYOF node should raise a warning if
  558  *          matching against an above-Unicode code point.
  559  * (These uses are mutually exclusive because the warning requires a \p{}, and
  560  * \p{} implies /u which deselects /d).  An SSC node only has this bit set if
  561  * what is meant is the warning.  The long macro name is to make sure that you
  562  * are cautioned about its shared nature */
  563 #define ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER 0x80
  564 
  565 #define ANYOF_FLAGS_ALL     (0xff & ~0x10)
  566 
  567 #define ANYOF_LOCALE_FLAGS (ANYOFL_FOLD | ANYOF_MATCHES_POSIXL)
  568 
  569 /* These are the flags that apply to both regular ANYOF nodes and synthetic
  570  * start class nodes during construction of the SSC.  During finalization of
  571  * the SSC, other of the flags may get added to it */
  572 #define ANYOF_COMMON_FLAGS      0
  573 
  574 /* Character classes for node->classflags of ANYOF */
  575 /* Should be synchronized with a table in regprop() */
  576 /* 2n should be the normal one, paired with its complement at 2n+1 */
  577 
  578 #define ANYOF_ALPHA    ((_CC_ALPHA) * 2)
  579 #define ANYOF_NALPHA   ((ANYOF_ALPHA) + 1)
  580 #define ANYOF_ALPHANUMERIC   ((_CC_ALPHANUMERIC) * 2)    /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
  581 #define ANYOF_NALPHANUMERIC  ((ANYOF_ALPHANUMERIC) + 1)
  582 #define ANYOF_ASCII    ((_CC_ASCII) * 2)
  583 #define ANYOF_NASCII   ((ANYOF_ASCII) + 1)
  584 #define ANYOF_BLANK    ((_CC_BLANK) * 2)     /* GNU extension: space and tab: non-vertical space */
  585 #define ANYOF_NBLANK   ((ANYOF_BLANK) + 1)
  586 #define ANYOF_CASED    ((_CC_CASED) * 2)    /* Pseudo class for [:lower:] or
  587                                                [:upper:] under /i */
  588 #define ANYOF_NCASED   ((ANYOF_CASED) + 1)
  589 #define ANYOF_CNTRL    ((_CC_CNTRL) * 2)
  590 #define ANYOF_NCNTRL   ((ANYOF_CNTRL) + 1)
  591 #define ANYOF_DIGIT    ((_CC_DIGIT) * 2)     /* \d */
  592 #define ANYOF_NDIGIT   ((ANYOF_DIGIT) + 1)
  593 #define ANYOF_GRAPH    ((_CC_GRAPH) * 2)
  594 #define ANYOF_NGRAPH   ((ANYOF_GRAPH) + 1)
  595 #define ANYOF_LOWER    ((_CC_LOWER) * 2)
  596 #define ANYOF_NLOWER   ((ANYOF_LOWER) + 1)
  597 #define ANYOF_PRINT    ((_CC_PRINT) * 2)
  598 #define ANYOF_NPRINT   ((ANYOF_PRINT) + 1)
  599 #define ANYOF_PUNCT    ((_CC_PUNCT) * 2)
  600 #define ANYOF_NPUNCT   ((ANYOF_PUNCT) + 1)
  601 #define ANYOF_SPACE    ((_CC_SPACE) * 2)     /* \s */
  602 #define ANYOF_NSPACE   ((ANYOF_SPACE) + 1)
  603 #define ANYOF_UPPER    ((_CC_UPPER) * 2)
  604 #define ANYOF_NUPPER   ((ANYOF_UPPER) + 1)
  605 #define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2)  /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
  606 #define ANYOF_NWORDCHAR   ((ANYOF_WORDCHAR) + 1)
  607 #define ANYOF_XDIGIT   ((_CC_XDIGIT) * 2)
  608 #define ANYOF_NXDIGIT  ((ANYOF_XDIGIT) + 1)
  609 
  610 /* pseudo classes below this, not stored in the class bitmap, but used as flags
  611    during compilation of char classes */
  612 
  613 #define ANYOF_VERTWS    ((_CC_VERTSPACE) * 2)
  614 #define ANYOF_NVERTWS   ((ANYOF_VERTWS)+1)
  615 
  616 /* It is best if this is the last one, as all above it are stored as bits in a
  617  * bitmap, and it isn't part of that bitmap */
  618 #if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
  619 #   error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
  620 #endif
  621 
  622 #define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
  623                                          *       '< ANYOF_MAX'
  624                                          * Hence doesn't include VERTWS, as that
  625                                          * is a pseudo class */
  626 #define ANYOF_MAX      ANYOF_POSIXL_MAX
  627 
  628 #if (ANYOF_POSIXL_MAX > 32)   /* Must fit in 32-bit word */
  629 #   error Problem with handy.h _CC_foo #defines
  630 #endif
  631 
  632 #define ANYOF_HORIZWS   ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
  633 #define ANYOF_NHORIZWS  ((ANYOF_POSIXL_MAX)+3)
  634 
  635 #define ANYOF_UNIPROP   ((ANYOF_POSIXL_MAX)+4)  /* Used to indicate a Unicode
  636                                                    property: \p{} or \P{} */
  637 
  638 /* Backward source code compatibility. */
  639 
  640 #define ANYOF_ALNUML     ANYOF_ALNUM
  641 #define ANYOF_NALNUML    ANYOF_NALNUM
  642 #define ANYOF_SPACEL     ANYOF_SPACE
  643 #define ANYOF_NSPACEL    ANYOF_NSPACE
  644 #define ANYOF_ALNUM ANYOF_WORDCHAR
  645 #define ANYOF_NALNUM ANYOF_NWORDCHAR
  646 
  647 /* Utility macros for the bitmap and classes of ANYOF */
  648 
  649 #define ANYOF_FLAGS(p)      ((p)->flags)
  650 
  651 #define ANYOF_BIT(c)        (1U << ((c) & 7))
  652 
  653 #define POSIXL_SET(field, c)    ((field) |= (1U << (c)))
  654 #define ANYOF_POSIXL_SET(p, c)  POSIXL_SET(((regnode_charclass_posixl*) (p))->classflags, (c))
  655 
  656 #define POSIXL_CLEAR(field, c) ((field) &= ~ (1U <<(c)))
  657 #define ANYOF_POSIXL_CLEAR(p, c) POSIXL_CLEAR(((regnode_charclass_posixl*) (p))->classflags, (c))
  658 
  659 #define POSIXL_TEST(field, c)   ((field) & (1U << (c)))
  660 #define ANYOF_POSIXL_TEST(p, c) POSIXL_TEST(((regnode_charclass_posixl*) (p))->classflags, (c))
  661 
  662 #define POSIXL_ZERO(field)  STMT_START { (field) = 0; } STMT_END
  663 #define ANYOF_POSIXL_ZERO(ret)  POSIXL_ZERO(((regnode_charclass_posixl*) (ret))->classflags)
  664 
  665 #define ANYOF_POSIXL_SET_TO_BITMAP(p, bits)                                 \
  666      STMT_START {                                                           \
  667                     ((regnode_charclass_posixl*) (p))->classflags = (bits); \
  668      } STMT_END
  669 
  670 /* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
  671 #define ANYOF_POSIXL_SETALL(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1; } STMT_END
  672 #define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
  673 
  674 #define ANYOF_POSIXL_TEST_ANY_SET(p)                               \
  675         ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL)                           \
  676      && (((regnode_charclass_posixl*)(p))->classflags))
  677 #define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
  678 
  679 /* Since an SSC always has this field, we don't have to test for that; nor do
  680  * we want to because the bit isn't set for SSC during its construction */
  681 #define ANYOF_POSIXL_SSC_TEST_ANY_SET(p)                               \
  682                             cBOOL(((regnode_ssc*)(p))->classflags)
  683 #define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */       \
  684         (((regnode_ssc*) (p))->classflags                              \
  685                         == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
  686 
  687 #define ANYOF_POSIXL_TEST_ALL_SET(p)                                   \
  688         ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL)                               \
  689          && ((regnode_charclass_posixl*) (p))->classflags              \
  690                         == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
  691 
  692 #define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
  693 #define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
  694 
  695 #define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
  696 
  697 #define ANYOF_BITMAP_ZERO(ret)  Zero(((regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
  698 #define ANYOF_BITMAP(p)     ((regnode_charclass*)(p))->bitmap
  699 #define ANYOF_BITMAP_BYTE(p, c) BITMAP_BYTE(ANYOF_BITMAP(p), c)
  700 #define ANYOF_BITMAP_SET(p, c)  (ANYOF_BITMAP_BYTE(p, c) |=  ANYOF_BIT(c))
  701 #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
  702 #define ANYOF_BITMAP_TEST(p, c) cBOOL(ANYOF_BITMAP_BYTE(p, c) &   ANYOF_BIT(c))
  703 
  704 #define ANYOF_BITMAP_SETALL(p)      \
  705     memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
  706 #define ANYOF_BITMAP_CLEARALL(p)    \
  707     Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
  708 
  709 /*
  710  * Utility definitions.
  711  */
  712 #ifndef CHARMASK
  713 #  define UCHARAT(p)    ((int)*(const U8*)(p))
  714 #else
  715 #  define UCHARAT(p)    ((int)*(p)&CHARMASK)
  716 #endif
  717 
  718 #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
  719 
  720 #define REG_ZERO_LEN_SEEN                   0x00000001
  721 #define REG_LOOKBEHIND_SEEN                 0x00000002
  722 #define REG_GPOS_SEEN                       0x00000004
  723 /* spare */
  724 #define REG_RECURSE_SEEN                    0x00000020
  725 #define REG_TOP_LEVEL_BRANCHES_SEEN         0x00000040
  726 #define REG_VERBARG_SEEN                    0x00000080
  727 #define REG_CUTGROUP_SEEN                   0x00000100
  728 #define REG_RUN_ON_COMMENT_SEEN             0x00000200
  729 #define REG_UNFOLDED_MULTI_SEEN             0x00000400
  730 /* spare */
  731 #define REG_UNBOUNDED_QUANTIFIER_SEEN       0x00001000
  732 
  733 
  734 START_EXTERN_C
  735 
  736 #ifdef PLUGGABLE_RE_EXTENSION
  737 #include "re_nodes.h"
  738 #else
  739 #include "regnodes.h"
  740 #endif
  741 
  742 #ifndef PLUGGABLE_RE_EXTENSION
  743 #ifndef DOINIT
  744 EXTCONST regexp_engine PL_core_reg_engine;
  745 #else /* DOINIT */
  746 EXTCONST regexp_engine PL_core_reg_engine = { 
  747         Perl_re_compile,
  748         Perl_regexec_flags,
  749         Perl_re_intuit_start,
  750         Perl_re_intuit_string, 
  751         Perl_regfree_internal,
  752         Perl_reg_numbered_buff_fetch,
  753         Perl_reg_numbered_buff_store,
  754         Perl_reg_numbered_buff_length,
  755         Perl_reg_named_buff,
  756         Perl_reg_named_buff_iter,
  757         Perl_reg_qr_package,
  758 #if defined(USE_ITHREADS)        
  759         Perl_regdupe_internal,
  760 #endif        
  761         Perl_re_op_compile
  762 };
  763 #endif /* DOINIT */
  764 #endif /* PLUGGABLE_RE_EXTENSION */
  765 
  766 
  767 END_EXTERN_C
  768 
  769 
  770 /* .what is a character array with one character for each member of .data
  771  * The character describes the function of the corresponding .data item:
  772  *   a - AV for paren_name_list under DEBUGGING
  773  *   f - start-class data for regstclass optimization
  774  *   l - start op for literal (?{EVAL}) item
  775  *   L - start op for literal (?{EVAL}) item, with separate CV (qr//)
  776  *   r - pointer to an embedded code-containing qr, e.g. /ab$qr/
  777  *   s - inversion list for Unicode-style character class, and the
  778  *       multicharacter strings resulting from casefolding the single-character
  779  *       entries in the character class
  780  *   t - trie struct
  781  *   u - trie struct's widecharmap (a HV, so can't share, must dup)
  782  *       also used for revcharmap and words under DEBUGGING
  783  *   T - aho-trie struct
  784  *   S - sv for named capture lookup
  785  * 20010712 mjd@plover.com
  786  * (Remember to update re_dup() and pregfree() if you add any items.)
  787  */
  788 struct reg_data {
  789     U32 count;
  790     U8 *what;
  791     void* data[1];
  792 };
  793 
  794 /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
  795    anchored* and float* via array indexes 0 and 1.  */
  796 #define anchored_substr substrs->data[0].substr
  797 #define anchored_utf8 substrs->data[0].utf8_substr
  798 #define anchored_offset substrs->data[0].min_offset
  799 #define anchored_end_shift substrs->data[0].end_shift
  800 
  801 #define float_substr substrs->data[1].substr
  802 #define float_utf8 substrs->data[1].utf8_substr
  803 #define float_min_offset substrs->data[1].min_offset
  804 #define float_max_offset substrs->data[1].max_offset
  805 #define float_end_shift substrs->data[1].end_shift
  806 
  807 #define check_substr substrs->data[2].substr
  808 #define check_utf8 substrs->data[2].utf8_substr
  809 #define check_offset_min substrs->data[2].min_offset
  810 #define check_offset_max substrs->data[2].max_offset
  811 #define check_end_shift substrs->data[2].end_shift
  812 
  813 #define RX_ANCHORED_SUBSTR(rx)  (ReANY(rx)->anchored_substr)
  814 #define RX_ANCHORED_UTF8(rx)    (ReANY(rx)->anchored_utf8)
  815 #define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
  816 #define RX_FLOAT_UTF8(rx)   (ReANY(rx)->float_utf8)
  817 
  818 /* trie related stuff */
  819 
  820 /* a transition record for the state machine. the
  821    check field determines which state "owns" the
  822    transition. the char the transition is for is
  823    determined by offset from the owning states base
  824    field.  the next field determines which state
  825    is to be transitioned to if any.
  826 */
  827 struct _reg_trie_trans {
  828   U32 next;
  829   U32 check;
  830 };
  831 
  832 /* a transition list element for the list based representation */
  833 struct _reg_trie_trans_list_elem {
  834     U16 forid;
  835     U32 newstate;
  836 };
  837 typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
  838 
  839 /* a state for compressed nodes. base is an offset
  840   into an array of reg_trie_trans array. If wordnum is
  841   nonzero the state is accepting. if base is zero then
  842   the state has no children (and will be accepting)
  843 */
  844 struct _reg_trie_state {
  845   U16 wordnum;
  846   union {
  847     U32                base;
  848     reg_trie_trans_le* list;
  849   } trans;
  850 };
  851 
  852 /* info per word; indexed by wordnum */
  853 typedef struct {
  854     U16  prev;  /* previous word in acceptance chain; eg in
  855          * zzz|abc|ab/ after matching the chars abc, the
  856          * accepted word is #2, and the previous accepted
  857          * word is #3 */
  858     U32 len;    /* how many chars long is this word? */
  859     U32 accept; /* accept state for this word */
  860 } reg_trie_wordinfo;
  861 
  862 
  863 typedef struct _reg_trie_state    reg_trie_state;
  864 typedef struct _reg_trie_trans    reg_trie_trans;
  865 
  866 
  867 /* anything in here that needs to be freed later
  868    should be dealt with in pregfree.
  869    refcount is first in both this and _reg_ac_data to allow a space
  870    optimisation in Perl_regdupe.  */
  871 struct _reg_trie_data {
  872     U32             refcount;        /* number of times this trie is referenced */
  873     U32             lasttrans;       /* last valid transition element */
  874     U16             *charmap;        /* byte to charid lookup array */
  875     reg_trie_state  *states;         /* state data */
  876     reg_trie_trans  *trans;          /* array of transition elements */
  877     char            *bitmap;         /* stclass bitmap */
  878     U16         *jump;           /* optional 1 indexed array of offsets before tail 
  879                                         for the node following a given word. */
  880     reg_trie_wordinfo *wordinfo;     /* array of info per word */
  881     U16             uniquecharcount; /* unique chars in trie (width of trans table) */
  882     U32             startstate;      /* initial state - used for common prefix optimisation */
  883     STRLEN          minlen;          /* minimum length of words in trie - build/opt only? */
  884     STRLEN          maxlen;          /* maximum length of words in trie - build/opt only? */
  885     U32             prefixlen;       /* #chars in common prefix */
  886     U32             statecount;      /* Build only - number of states in the states array 
  887                                         (including the unused zero state) */
  888     U32             wordcount;       /* Build only */
  889 #ifdef DEBUGGING
  890     STRLEN          charcount;       /* Build only */
  891 #endif
  892 };
  893 /* There is one (3 under DEBUGGING) pointers that logically belong in this
  894    structure, but are held outside as they need duplication on thread cloning,
  895    whereas the rest of the structure can be read only:
  896     HV              *widecharmap;    code points > 255 to charid
  897 #ifdef DEBUGGING
  898     AV              *words;          Array of words contained in trie, for dumping
  899     AV              *revcharmap;     Map of each charid back to its character representation
  900 #endif
  901 */
  902 
  903 #define TRIE_WORDS_OFFSET 2
  904 
  905 typedef struct _reg_trie_data reg_trie_data;
  906 
  907 /* refcount is first in both this and _reg_trie_data to allow a space
  908    optimisation in Perl_regdupe.  */
  909 struct _reg_ac_data {
  910     U32              refcount;
  911     U32              trie;
  912     U32              *fail;
  913     reg_trie_state   *states;
  914 };
  915 typedef struct _reg_ac_data reg_ac_data;
  916 
  917 /* ANY_BIT doesn't use the structure, so we can borrow it here.
  918    This is simpler than refactoring all of it as wed end up with
  919    three different sets... */
  920 
  921 #define TRIE_BITMAP(p)      (((reg_trie_data *)(p))->bitmap)
  922 #define TRIE_BITMAP_BYTE(p, c)  BITMAP_BYTE(TRIE_BITMAP(p), c)
  923 #define TRIE_BITMAP_SET(p, c)   (TRIE_BITMAP_BYTE(p, c) |=  ANYOF_BIT((U8)c))
  924 #define TRIE_BITMAP_CLEAR(p,c)  (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
  925 #define TRIE_BITMAP_TEST(p, c)  (TRIE_BITMAP_BYTE(p, c) &   ANYOF_BIT((U8)c))
  926 
  927 #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
  928 #define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
  929 
  930 
  931 #define BITMAP_BYTE(p, c)   (( (U8*) p) [ ( ( (UV) (c)) >> 3) ] )
  932 #define BITMAP_TEST(p, c)   (BITMAP_BYTE(p, c) &   ANYOF_BIT((U8)c))
  933 
  934 /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
  935 #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
  936 #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
  937 #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
  938 
  939 #ifdef DEBUGGING
  940 #define TRIE_CHARCOUNT(trie) ((trie)->charcount)
  941 #else
  942 #define TRIE_CHARCOUNT(trie) (trie_charcount)
  943 #endif
  944 
  945 #define RE_TRIE_MAXBUF_INIT 65536
  946 #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
  947 #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
  948 
  949 #define RE_COMPILE_RECURSION_INIT 1000
  950 #define RE_COMPILE_RECURSION_LIMIT "\022E_COMPILE_RECURSION_LIMIT"
  951 
  952 /*
  953 
  954 RE_DEBUG_FLAGS is used to control what debug output is emitted
  955 its divided into three groups of options, some of which interact.
  956 The three groups are: Compile, Execute, Extra. There is room for a
  957 further group, as currently only the low three bytes are used.
  958 
  959     Compile Options:
  960     
  961     PARSE
  962     PEEP
  963     TRIE
  964     PROGRAM
  965     OFFSETS
  966 
  967     Execute Options:
  968 
  969     INTUIT
  970     MATCH
  971     TRIE
  972 
  973     Extra Options
  974 
  975     TRIE
  976     OFFSETS
  977 
  978 If you modify any of these make sure you make corresponding changes to
  979 re.pm, especially to the documentation.
  980 
  981 */
  982 
  983 
  984 /* Compile */
  985 #define RE_DEBUG_COMPILE_MASK      0x0000FF
  986 #define RE_DEBUG_COMPILE_PARSE     0x000001
  987 #define RE_DEBUG_COMPILE_OPTIMISE  0x000002
  988 #define RE_DEBUG_COMPILE_TRIE      0x000004
  989 #define RE_DEBUG_COMPILE_DUMP      0x000008
  990 #define RE_DEBUG_COMPILE_FLAGS     0x000010
  991 #define RE_DEBUG_COMPILE_TEST      0x000020
  992 
  993 /* Execute */
  994 #define RE_DEBUG_EXECUTE_MASK      0x00FF00
  995 #define RE_DEBUG_EXECUTE_INTUIT    0x000100
  996 #define RE_DEBUG_EXECUTE_MATCH     0x000200
  997 #define RE_DEBUG_EXECUTE_TRIE      0x000400
  998 
  999 /* Extra */
 1000 #define RE_DEBUG_EXTRA_MASK        0xFF0000
 1001 #define RE_DEBUG_EXTRA_TRIE        0x010000
 1002 #define RE_DEBUG_EXTRA_OFFSETS     0x020000
 1003 #define RE_DEBUG_EXTRA_OFFDEBUG    0x040000
 1004 #define RE_DEBUG_EXTRA_STATE       0x080000
 1005 #define RE_DEBUG_EXTRA_OPTIMISE    0x100000
 1006 #define RE_DEBUG_EXTRA_BUFFERS     0x400000
 1007 #define RE_DEBUG_EXTRA_GPOS        0x800000
 1008 /* combined */
 1009 #define RE_DEBUG_EXTRA_STACK       0x280000
 1010 
 1011 #define RE_DEBUG_FLAG(x) (re_debug_flags & x)
 1012 /* Compile */
 1013 #define DEBUG_COMPILE_r(x) DEBUG_r( \
 1014     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_MASK)) x  )
 1015 #define DEBUG_PARSE_r(x) DEBUG_r( \
 1016     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_PARSE)) x  )
 1017 #define DEBUG_OPTIMISE_r(x) DEBUG_r( \
 1018     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE)) x  )
 1019 #define DEBUG_DUMP_r(x) DEBUG_r( \
 1020     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_DUMP)) x  )
 1021 #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
 1022     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_TRIE)) x )
 1023 #define DEBUG_FLAGS_r(x) DEBUG_r( \
 1024     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_FLAGS)) x )
 1025 #define DEBUG_TEST_r(x) DEBUG_r( \
 1026     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_COMPILE_TEST)) x )
 1027 /* Execute */
 1028 #define DEBUG_EXECUTE_r(x) DEBUG_r( \
 1029     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_MASK)) x  )
 1030 #define DEBUG_INTUIT_r(x) DEBUG_r( \
 1031     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT)) x  )
 1032 #define DEBUG_MATCH_r(x) DEBUG_r( \
 1033     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_MATCH)) x  )
 1034 #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
 1035     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXECUTE_TRIE)) x )
 1036 
 1037 /* Extra */
 1038 #define DEBUG_EXTRA_r(x) DEBUG_r( \
 1039     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_MASK)) x  )
 1040 #define DEBUG_OFFSETS_r(x) DEBUG_r( \
 1041     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS)) x  )
 1042 #define DEBUG_STATE_r(x) DEBUG_r( \
 1043     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_STATE)) x )
 1044 #define DEBUG_STACK_r(x) DEBUG_r( \
 1045     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_STACK)) x )
 1046 #define DEBUG_BUFFERS_r(x) DEBUG_r( \
 1047     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS)) x )
 1048 
 1049 #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
 1050     if (DEBUG_v_TEST || ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
 1051          (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)))) x )
 1052 #define MJD_OFFSET_DEBUG(x) DEBUG_r( \
 1053     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG)) \
 1054         Perl_warn_nocontext x )
 1055 #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
 1056     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_TRIE)) x )
 1057 #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
 1058     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_TRIE)) x )
 1059 
 1060 #define DEBUG_TRIE_r(x) DEBUG_r( \
 1061     if (DEBUG_v_TEST || (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
 1062         | RE_DEBUG_EXECUTE_TRIE ))) x )
 1063 #define DEBUG_GPOS_r(x) DEBUG_r( \
 1064     if (DEBUG_v_TEST || (re_debug_flags & RE_DEBUG_EXTRA_GPOS)) x )
 1065 
 1066 /* initialization */
 1067 /* get_sv() can return NULL during global destruction. */
 1068 #define GET_RE_DEBUG_FLAGS DEBUG_r({ \
 1069         SV * re_debug_flags_sv = NULL; \
 1070         re_debug_flags_sv = PL_curcop ? get_sv(RE_DEBUG_FLAGS, GV_ADD) : NULL; \
 1071         if (re_debug_flags_sv) { \
 1072             if (!SvIOK(re_debug_flags_sv)) \
 1073                 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
 1074             re_debug_flags=SvIV(re_debug_flags_sv); \
 1075         }\
 1076 })
 1077 
 1078 #ifdef DEBUGGING
 1079 
 1080 #define GET_RE_DEBUG_FLAGS_DECL volatile IV re_debug_flags = 0; \
 1081         PERL_UNUSED_VAR(re_debug_flags); GET_RE_DEBUG_FLAGS;
 1082 
 1083 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)   \
 1084     const char * const rpv =                                \
 1085         pv_pretty((dsv), (pv), (l), (m),                    \
 1086             PL_colors[(c1)],PL_colors[(c2)],                \
 1087             PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) );         \
 1088     const int rlen = SvCUR(dsv)
 1089 
 1090 /* This is currently unsed in the core */
 1091 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)                            \
 1092     const char * const rpv =                                        \
 1093         pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m),  \
 1094             PL_colors[(c1)],PL_colors[(c2)],                        \
 1095             PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
 1096 
 1097 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)                     \
 1098     const char * const rpv =                                        \
 1099         pv_pretty((dsv), (pv), (l), (m),                            \
 1100             PL_colors[0], PL_colors[1],                             \
 1101             ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
 1102               ((isuni) ? PERL_PV_ESCAPE_UNI : 0))                  \
 1103         )
 1104 
 1105 #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
 1106 #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
 1107     
 1108 #else /* if not DEBUGGING */
 1109 
 1110 #define GET_RE_DEBUG_FLAGS_DECL
 1111 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
 1112 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
 1113 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
 1114 #define RE_SV_DUMPLEN(ItEm)
 1115 #define RE_SV_TAIL(ItEm)
 1116 
 1117 #endif /* DEBUG RELATED DEFINES */
 1118 
 1119 #define FIRST_NON_ASCII_DECIMAL_DIGIT 0x660  /* ARABIC_INDIC_DIGIT_ZERO */
 1120 
 1121 typedef enum {
 1122     TRADITIONAL_BOUND = _CC_WORDCHAR,
 1123     GCB_BOUND,
 1124     LB_BOUND,
 1125     SB_BOUND,
 1126     WB_BOUND
 1127 } bound_type;
 1128 
 1129 #endif /* PERL_REGCOMP_H_ */
 1130 
 1131 /*
 1132  * ex: set ts=8 sts=4 sw=4 et:
 1133  */