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