5 * First, the stuff that ends up in the outside-world include file
6 = typedef off_t regoff_t;
9 = size_t re_nsub; // number of parenthesized subexpressions
10 = const char *re_endp; // end pointer for REG_PEND
11 = struct re_guts *re_g; // none of your business :-)
14 = regoff_t rm_so; // start of match
15 = regoff_t rm_eo; // end of match
19 * internals of regex_t
21 #define MAGIC1 ((('r'^0200)<<8) | 'e')
24 * The internal representation is a *strip*, a sequence of
25 * operators ending with an endmarker. (Some terminology etc. is a
26 * historical relic of earlier versions which used multiple strips.)
27 * Certain oddities in the representation are there to permit running
28 * the machinery backwards; in particular, any deviation from sequential
29 * flow must be marked at both its source and its destination. Some
32 * - OPLUS_ and O_PLUS are *inside* the loop they create.
33 * - OQUEST_ and O_QUEST are *outside* the bypass they create.
34 * - OCH_ and O_CH are *outside* the multi-way branch they create, while
35 * OOR1 and OOR2 are respectively the end and the beginning of one of
36 * the branches. Note that there is an implicit OOR2 following OCH_
37 * and an implicit OOR1 preceding O_CH.
39 * In state representations, an operator's bit is on to signify a state
40 * immediately *preceding* "execution" of that operator.
42 typedef unsigned long sop; /* strip operator */
43 typedef unsigned long sopno;
44 #define OPRMASK 0xf8000000
45 #define OPDMASK 0x07ffffff
46 #define OPSHIFT ((unsigned)27)
47 #define OP(n) ((n)&OPRMASK)
48 #define OPND(n) ((n)&OPDMASK)
49 #define SOP(op, opnd) ((op)|(opnd))
53 #define UNSIGNEDLONG1 1ul
55 #define UNSIGNEDLONG1 ((unsigned long)1)
59 #define MAKE_UNSIGNED_LONG(num) (UNSIGNEDLONG1*num)
60 #define SHIFTED_UL(num) (MAKE_UNSIGNED_LONG(num) << OPSHIFT)
62 /* operators meaning operand */
63 /* (back, fwd are offsets) */
64 #define OEND (SHIFTED_UL(1)) /* endmarker - */
65 #define OCHAR (SHIFTED_UL(2)) /* character unsigned char */
66 #define OBOL (SHIFTED_UL(3)) /* left anchor - */
67 #define OEOL (SHIFTED_UL(4)) /* right anchor - */
68 #define OANY (SHIFTED_UL(5)) /* . - */
69 #define OANYOF (SHIFTED_UL(6)) /* [...] set number */
70 #define OBACK_ (SHIFTED_UL(7)) /* begin \d paren number */
71 #define O_BACK (SHIFTED_UL(8)) /* end \d paren number */
72 #define OPLUS_ (SHIFTED_UL(9)) /* + prefix fwd to suffix */
73 #define O_PLUS (SHIFTED_UL(10)) /* + suffix back to prefix */
74 #define OQUEST_ (SHIFTED_UL(11)) /* ? prefix fwd to suffix */
75 #define O_QUEST (SHIFTED_UL(12)) /* ? suffix back to prefix */
76 #define OLPAREN (SHIFTED_UL(13)) /* ( fwd to ) */
77 #define ORPAREN (SHIFTED_UL(14)) /* ) back to ( */
78 #define OCH_ (SHIFTED_UL(15)) /* begin choice fwd to OOR2 */
79 #define OOR1 (SHIFTED_UL(16)) /* | pt. 1 back to OOR1 or OCH_ */
80 #define OOR2 (SHIFTED_UL(17)) /* | pt. 2 fwd to OOR2 or O_CH */
81 #define O_CH (SHIFTED_UL(18)) /* end choice back to OOR1 */
82 #define OBOW (SHIFTED_UL(19)) /* begin word - */
83 #define OEOW (SHIFTED_UL(20)) /* end word - */
86 * Structure for [] character-set representation. Character sets are
87 * done as bit vectors, grouped 8 to a byte vector for compactness.
88 * The individual set therefore has both a pointer to the byte vector
89 * and a mask to pick out the relevant bit of each byte. A hash code
90 * simplifies testing whether two sets could be identical.
92 * This will get trickier for multicharacter collating elements. As
93 * preliminary hooks for dealing with such things, we also carry along
94 * a string of multi-character elements, and decide the size of the
95 * vectors at run time.
98 uch *ptr; /* -> uch [csetsize] */
99 uch mask; /* bit within array */
100 uch hash; /* hash code */
102 char *multis; /* -> char[smulti] ab\0cd\0ef\0\0 */
104 /* note that CHadd and CHsub are unsafe, and CHIN doesn't yield 0/1 */
105 #define CHadd(cs, c) do { \
106 (cs)->ptr[(uch)(c)] = (uch)((cs)->ptr[(uch)(c)] | (cs)->mask); \
107 (cs)->hash = (uch)((cs)->hash + (c)); \
110 #define CHsub(cs, c) do { \
111 (cs)->ptr[(uch)(c)] = (uch)((cs)->ptr[(uch)(c)] & ~(cs)->mask); \
112 (cs)->hash = (uch)((cs)->hash - (c)); \
115 #define CHIN(cs, c) ((cs)->ptr[(uch)(c)] & (cs)->mask)
116 #define MCadd(p, cs, cp) mcadd(p, cs, cp) /* regcomp() internal fns */
117 #define MCsub(p, cs, cp) mcsub(p, cs, cp)
118 #define MCin(p, cs, cp) mcin(p, cs, cp)
120 /* stuff for character categories */
121 typedef unsigned char cat_t;
124 * main compiled-expression structure
128 # define MAGIC2 ((('R'^0200)<<8)|'E')
129 sop *strip; /* malloced area for strip */
130 int csetsize; /* number of bits in a cset vector */
131 int ncsets; /* number of csets in use */
132 cset *sets; /* -> cset [ncsets] */
133 uch *setbits; /* -> uch[csetsize][ncsets/CHAR_BIT] */
134 int cflags; /* copy of regcomp() cflags argument */
135 sopno nstates; /* = number of sops */
136 sopno firststate; /* the initial OEND (normally 0) */
137 sopno laststate; /* the final OEND */
138 int iflags; /* internal flags */
139 # define USEBOL 01 /* used ^ */
140 # define USEEOL 02 /* used $ */
141 # define BAD 04 /* something wrong */
142 int nbol; /* number of ^ used */
143 int neol; /* number of $ used */
144 int ncategories; /* how many character categories */
145 cat_t *categories; /* ->catspace[-CHAR_MIN] */
146 char *must; /* match must contain this string */
147 sopno mlen; /* length of must */
148 size_t nsub; /* copy of re_nsub */
149 int backrefs; /* does it use back references? */
150 sopno nplus; /* how deep does it nest +s? */
151 /* catspace must be last */
152 cat_t catspace[1]; /* actually [NC] */
156 #define OUT (CHAR_MAX+1) /* a non-character value */
157 #define ISWORD(c) (isalnum(c) || (c) == '_')
159 #endif /* !REGEX2_H */