dfa.h

早期freebsd实现

  _OR,				/* _OR is an operator of two arguments that
				   matches either of its arguments. */

  _LPAREN,			/* _LPAREN never appears in the parse tree,
				   it is only a lexeme. */

  _RPAREN,			/* _RPAREN never appears in the parse tree. */

  _SET				/* _SET and (and any value greater) is a
				   terminal symbol that matches any of a
				   class of characters. */
} _token;

#else /* ! __STDC__ */

typedef short _token;

#define _END -1
#define _EMPTY _NOTCHAR
#define _BACKREF (_EMPTY + 1)
#define _BEGLINE (_EMPTY + 2)
#define _ALLBEGLINE (_EMPTY + 3)
#define _ENDLINE (_EMPTY + 4)
#define _ALLENDLINE (_EMPTY + 5)
#define _BEGWORD (_EMPTY + 6)
#define _ENDWORD (_EMPTY + 7)
#define _LIMWORD (_EMPTY + 8)
#define _NOTLIMWORD (_EMPTY + 9)
#define _QMARK (_EMPTY + 10)
#define _STAR (_EMPTY + 11)
#define _PLUS (_EMPTY + 12)
#define _CAT (_EMPTY + 13)
#define _OR (_EMPTY + 14)
#define _LPAREN (_EMPTY + 15)
#define _RPAREN (_EMPTY + 16)
#define _SET (_EMPTY + 17)

#endif /* ! __STDC__ */

/* Sets are stored in an array in the compiled regexp; the index of the
   array corresponding to a given set token is given by _SET_INDEX(t). */
#define _SET_INDEX(t) ((t) - _SET)

/* Sometimes characters can only be matched depending on the surrounding
   context.  Such context decisions depend on what the previous character
   was, and the value of the current (lookahead) character.  Context
   dependent constraints are encoded as 8 bit integers.  Each bit that
   is set indicates that the constraint succeeds in the corresponding
   context.

   bit 7 - previous and current are newlines
   bit 6 - previous was newline, current isn't
   bit 5 - previous wasn't newline, current is
   bit 4 - neither previous nor current is a newline
   bit 3 - previous and current are word-constituents
   bit 2 - previous was word-constituent, current isn't
   bit 1 - previous wasn't word-constituent, current is
   bit 0 - neither previous nor current is word-constituent

   Word-constituent characters are those that satisfy isalnum().

   The macro _SUCCEEDS_IN_CONTEXT determines whether a a given constraint
   succeeds in a particular context.  Prevn is true if the previous character
   was a newline, currn is true if the lookahead character is a newline.
   Prevl and currl similarly depend upon whether the previous and current
   characters are word-constituent letters. */
#define _MATCHES_NEWLINE_CONTEXT(constraint, prevn, currn) \
  ((constraint) & (1 << (((prevn) ? 2 : 0) + ((currn) ? 1 : 0) + 4)))
#define _MATCHES_LETTER_CONTEXT(constraint, prevl, currl) \
  ((constraint) & (1 << (((prevl) ? 2 : 0) + ((currl) ? 1 : 0))))
#define _SUCCEEDS_IN_CONTEXT(constraint, prevn, currn, prevl, currl) \
  (_MATCHES_NEWLINE_CONTEXT(constraint, prevn, currn)		     \
   && _MATCHES_LETTER_CONTEXT(constraint, prevl, currl))

/* The following macros give information about what a constraint depends on. */
#define _PREV_NEWLINE_DEPENDENT(constraint) \
  (((constraint) & 0xc0) >> 2 != ((constraint) & 0x30))
#define _PREV_LETTER_DEPENDENT(constraint) \
  (((constraint) & 0x0c) >> 2 != ((constraint) & 0x03))

/* Tokens that match the empty string subject to some constraint actually
   work by applying that constraint to determine what may follow them,
   taking into account what has gone before.  The following values are
   the constraints corresponding to the special tokens previously defined. */
#define _NO_CONSTRAINT 0xff
#define _BEGLINE_CONSTRAINT 0xcf
#define _ENDLINE_CONSTRAINT 0xaf
#define _BEGWORD_CONSTRAINT 0xf2
#define _ENDWORD_CONSTRAINT 0xf4
#define _LIMWORD_CONSTRAINT 0xf6
#define _NOTLIMWORD_CONSTRAINT 0xf9

/* States of the recognizer correspond to sets of positions in the parse
   tree, together with the constraints under which they may be matched.
   So a position is encoded as an index into the parse tree together with
   a constraint. */
typedef struct
{
  unsigned index;		/* Index into the parse array. */
  unsigned constraint;		/* Constraint for matching this position. */
} _position;

/* Sets of positions are stored as arrays. */
typedef struct
{
  _position *elems;		/* Elements of this position set. */
  int nelem;			/* Number of elements in this set. */
} _position_set;

/* A state of the regexp consists of a set of positions, some flags,
   and the token value of the lowest-numbered position of the state that
   contains an _END token. */
typedef struct
{
  int hash;			/* Hash of the positions of this state. */
  _position_set elems;		/* Positions this state could match. */
  char newline;			/* True if previous state matched newline. */
  char letter;			/* True if previous state matched a letter. */
  char backref;			/* True if this state matches a \<digit>. */
  unsigned char constraint;	/* Constraint for this state to accept. */
  int first_end;		/* Token value of the first _END in elems. */
} _dfa_state;

/* If an r.e. is at most MUST_MAX characters long, we look for a string which
   must appear in it; whatever's found is dropped into the struct reg. */

#define MUST_MAX	50

/* A compiled regular expression. */
struct regexp
{
  /* Stuff built by the scanner. */
  _charset *charsets;		/* Array of character sets for _SET tokens. */
  int cindex;			/* Index for adding new charsets. */
  int calloc;			/* Number of charsets currently allocated. */

  /* Stuff built by the parser. */
  _token *tokens;		/* Postfix parse array. */
  int tindex;			/* Index for adding new tokens. */
  int talloc;			/* Number of tokens currently allocated. */
  int depth;			/* Depth required of an evaluation stack
				   used for depth-first traversal of the
				   parse tree. */
  int nleaves;			/* Number of leaves on the parse tree. */
  int nregexps;			/* Count of parallel regexps being built
				   with regparse(). */

  /* Stuff owned by the state builder. */
  _dfa_state *states;		/* States of the regexp. */
  int sindex;			/* Index for adding new states. */
  int salloc;			/* Number of states currently allocated. */

  /* Stuff built by the structure analyzer. */
  _position_set *follows;	/* Array of follow sets, indexed by position
				   index.  The follow of a position is the set
				   of positions containing characters that
				   could conceivably follow a character
				   matching the given position in a string
				   matching the regexp.  Allocated to the
				   maximum possible position index. */
  int searchflag;		/* True if we are supposed to build a searching
				   as opposed to an exact matcher.  A searching
				   matcher finds the first and shortest string
				   matching a regexp anywhere in the buffer,
				   whereas an exact matcher finds the longest
				   string matching, but anchored to the
				   beginning of the buffer. */

  /* Stuff owned by the executor. */
  int tralloc;			/* Number of transition tables that have
				   slots so far. */
  int trcount;			/* Number of transition tables that have
				   actually been built. */
  int **trans;			/* Transition tables for states that can
				   never accept.  If the transitions for a
				   state have not yet been computed, or the
				   state could possibly accept, its entry in
				   this table is NULL. */
  int **realtrans;		/* Trans always points to realtrans + 1; this
				   is so trans[-1] can contain NULL. */
  int **fails;			/* Transition tables after failing to accept
				   on a state that potentially could do so. */
  int *success;			/* Table of acceptance conditions used in
				   regexecute and computed in build_state. */
  int *newlines;		/* Transitions on newlines.  The entry for a
				   newline in any transition table is always
				   -1 so we can count lines without wasting
				   too many cycles.  The transition for a
				   newline is stored separately and handled
				   as a special case.  Newline is also used
				   as a sentinel at the end of the buffer. */
  char must[MUST_MAX];
  int mustn;
};

/* Some macros for user access to regexp internals. */

/* ACCEPTING returns true if s could possibly be an accepting state of r. */
#define ACCEPTING(s, r) ((r).states[s].constraint)

/* ACCEPTS_IN_CONTEXT returns true if the given state accepts in the
   specified context. */
#define ACCEPTS_IN_CONTEXT(prevn, currn, prevl, currl, state, reg) \
  _SUCCEEDS_IN_CONTEXT((reg).states[state].constraint,		   \
		       prevn, currn, prevl, currl)

/* FIRST_MATCHING_REGEXP returns the index number of the first of parallel
   regexps that a given state could accept.  Parallel regexps are numbered
   starting at 1. */
#define FIRST_MATCHING_REGEXP(state, reg) (-(reg).states[state].first_end)

/* Entry points. */

#ifdef __STDC__

/* Regsyntax() takes two arguments; the first sets the syntax bits described
   earlier in this file, and the second sets the case-folding flag. */
extern void regsyntax(long, int);

/* Compile the given string of the given length into the given struct regexp.
   Final argument is a flag specifying whether to build a searching or an
   exact matcher. */
extern void regcompile(const char *, size_t, struct regexp *, int);

/* Execute the given struct regexp on the buffer of characters.  The
   first char * points to the beginning, and the second points to the
   first character after the end of the buffer, which must be a writable
   place so a sentinel end-of-buffer marker can be stored there.  The
   second-to-last argument is a flag telling whether to allow newlines to
   be part of a string matching the regexp.  The next-to-last argument,
   if non-NULL, points to a place to increment every time we see a
   newline.  The final argument, if non-NULL, points to a flag that will
   be set if further examination by a backtracking matcher is needed in
   order to verify backreferencing; otherwise the flag will be cleared.
   Returns NULL if no match is found, or a pointer to the first
   character after the first & shortest matching string in the buffer. */
extern char *regexecute(struct regexp *, char *, char *, int, int *, int *);

/* Free the storage held by the components of a struct regexp. */
extern void reg_free(struct regexp *);

/* Entry points for people who know what they're doing. */

/* Initialize the components of a struct regexp. */
extern void reginit(struct regexp *);

/* Incrementally parse a string of given length into a struct regexp. */
extern void regparse(const char *, size_t, struct regexp *);

/* Analyze a parsed regexp; second argument tells whether to build a searching
   or an exact matcher. */
extern void reganalyze(struct regexp *, int);

/* Compute, for each possible character, the transitions out of a given
   state, storing them in an array of integers. */
extern void regstate(int, struct regexp *, int []);

/* Error handling. */

/* Regerror() is called by the regexp routines whenever an error occurs.  It
   takes a single argument, a NUL-terminated string describing the error.
   The default reg_error() prints the error message to stderr and exits.
   The user can provide a different reg_free() if so desired. */
extern void reg_error(const char *);

#else /* ! __STDC__ */
extern void regsyntax(), regcompile(), reg_free(), reginit(), regparse();
extern void reganalyze(), regstate(), reg_error();
extern char *regexecute();
#endif