1 /*- 2 * Copyright (c) 1992, 1993, 1994 Henry Spencer. 3 * Copyright (c) 1992, 1993, 1994 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * Henry Spencer. 8 * 9 * %sccs.include.redist.c% 10 * 11 * @(#)regex2.h 8.4 (Berkeley) 03/20/94 12 */ 13 14 /* 15 * First, the stuff that ends up in the outside-world include file 16 = typedef off_t regoff_t; 17 = typedef struct { 18 = int re_magic; 19 = size_t re_nsub; // number of parenthesized subexpressions 20 = const char *re_endp; // end pointer for REG_PEND 21 = struct re_guts *re_g; // none of your business :-) 22 = } regex_t; 23 = typedef struct { 24 = regoff_t rm_so; // start of match 25 = regoff_t rm_eo; // end of match 26 = } regmatch_t; 27 */ 28 /* 29 * internals of regex_t 30 */ 31 #define MAGIC1 ((('r'^0200)<<8) | 'e') 32 33 /* 34 * The internal representation is a *strip*, a sequence of 35 * operators ending with an endmarker. (Some terminology etc. is a 36 * historical relic of earlier versions which used multiple strips.) 37 * Certain oddities in the representation are there to permit running 38 * the machinery backwards; in particular, any deviation from sequential 39 * flow must be marked at both its source and its destination. Some 40 * fine points: 41 * 42 * - OPLUS_ and O_PLUS are *inside* the loop they create. 43 * - OQUEST_ and O_QUEST are *outside* the bypass they create. 44 * - OCH_ and O_CH are *outside* the multi-way branch they create, while 45 * OOR1 and OOR2 are respectively the end and the beginning of one of 46 * the branches. Note that there is an implicit OOR2 following OCH_ 47 * and an implicit OOR1 preceding O_CH. 48 * 49 * In state representations, an operator's bit is on to signify a state 50 * immediately *preceding* "execution" of that operator. 51 */ 52 typedef unsigned long sop; /* strip operator */ 53 typedef long sopno; 54 #define OPRMASK 0xf8000000 55 #define OPDMASK 0x07ffffff 56 #define OPSHIFT ((unsigned)27) 57 #define OP(n) ((n)&OPRMASK) 58 #define OPND(n) ((n)&OPDMASK) 59 #define SOP(op, opnd) ((op)|(opnd)) 60 /* operators meaning operand */ 61 /* (back, fwd are offsets) */ 62 #define OEND (1<<OPSHIFT) /* endmarker - */ 63 #define OCHAR (2<<OPSHIFT) /* character unsigned char */ 64 #define OBOL (3<<OPSHIFT) /* left anchor - */ 65 #define OEOL (4<<OPSHIFT) /* right anchor - */ 66 #define OANY (5<<OPSHIFT) /* . - */ 67 #define OANYOF (6<<OPSHIFT) /* [...] set number */ 68 #define OBACK_ (7<<OPSHIFT) /* begin \d paren number */ 69 #define O_BACK (8<<OPSHIFT) /* end \d paren number */ 70 #define OPLUS_ (9<<OPSHIFT) /* + prefix fwd to suffix */ 71 #define O_PLUS (10<<OPSHIFT) /* + suffix back to prefix */ 72 #define OQUEST_ (11<<OPSHIFT) /* ? prefix fwd to suffix */ 73 #define O_QUEST (12<<OPSHIFT) /* ? suffix back to prefix */ 74 #define OLPAREN (13<<OPSHIFT) /* ( fwd to ) */ 75 #define ORPAREN (14<<OPSHIFT) /* ) back to ( */ 76 #define OCH_ (15<<OPSHIFT) /* begin choice fwd to OOR2 */ 77 #define OOR1 (16<<OPSHIFT) /* | pt. 1 back to OOR1 or OCH_ */ 78 #define OOR2 (17<<OPSHIFT) /* | pt. 2 fwd to OOR2 or O_CH */ 79 #define O_CH (18<<OPSHIFT) /* end choice back to OOR1 */ 80 #define OBOW (19<<OPSHIFT) /* begin word - */ 81 #define OEOW (20<<OPSHIFT) /* end word - */ 82 83 /* 84 * Structure for [] character-set representation. Character sets are 85 * done as bit vectors, grouped 8 to a byte vector for compactness. 86 * The individual set therefore has both a pointer to the byte vector 87 * and a mask to pick out the relevant bit of each byte. A hash code 88 * simplifies testing whether two sets could be identical. 89 * 90 * This will get trickier for multicharacter collating elements. As 91 * preliminary hooks for dealing with such things, we also carry along 92 * a string of multi-character elements, and decide the size of the 93 * vectors at run time. 94 */ 95 typedef struct { 96 uch *ptr; /* -> uch [csetsize] */ 97 uch mask; /* bit within array */ 98 uch hash; /* hash code */ 99 size_t smultis; 100 char *multis; /* -> char[smulti] ab\0cd\0ef\0\0 */ 101 } cset; 102 /* note that CHadd and CHsub are unsafe, and CHIN doesn't yield 0/1 */ 103 #define CHadd(cs, c) ((cs)->ptr[(uch)(c)] |= (cs)->mask, (cs)->hash += (c)) 104 #define CHsub(cs, c) ((cs)->ptr[(uch)(c)] &= ~(cs)->mask, (cs)->hash -= (c)) 105 #define CHIN(cs, c) ((cs)->ptr[(uch)(c)] & (cs)->mask) 106 #define MCadd(p, cs, cp) mcadd(p, cs, cp) /* regcomp() internal fns */ 107 #define MCsub(p, cs, cp) mcsub(p, cs, cp) 108 #define MCin(p, cs, cp) mcin(p, cs, cp) 109 110 /* stuff for character categories */ 111 typedef unsigned char cat_t; 112 113 /* 114 * main compiled-expression structure 115 */ 116 struct re_guts { 117 int magic; 118 # define MAGIC2 ((('R'^0200)<<8)|'E') 119 sop *strip; /* malloced area for strip */ 120 int csetsize; /* number of bits in a cset vector */ 121 int ncsets; /* number of csets in use */ 122 cset *sets; /* -> cset [ncsets] */ 123 uch *setbits; /* -> uch[csetsize][ncsets/CHAR_BIT] */ 124 int cflags; /* copy of regcomp() cflags argument */ 125 sopno nstates; /* = number of sops */ 126 sopno firststate; /* the initial OEND (normally 0) */ 127 sopno laststate; /* the final OEND */ 128 int iflags; /* internal flags */ 129 # define USEBOL 01 /* used ^ */ 130 # define USEEOL 02 /* used $ */ 131 # define BAD 04 /* something wrong */ 132 int nbol; /* number of ^ used */ 133 int neol; /* number of $ used */ 134 int ncategories; /* how many character categories */ 135 cat_t *categories; /* ->catspace[-CHAR_MIN] */ 136 char *must; /* match must contain this string */ 137 int mlen; /* length of must */ 138 size_t nsub; /* copy of re_nsub */ 139 int backrefs; /* does it use back references? */ 140 sopno nplus; /* how deep does it nest +s? */ 141 /* catspace must be last */ 142 cat_t catspace[1]; /* actually [NC] */ 143 }; 144 145 /* misc utilities */ 146 #define OUT (CHAR_MAX+1) /* a non-character value */ 147 #define ISWORD(c) (isalnum(c) || (c) == '_') 148