1 /* vi: set sw=4 ts=4: */
3 * awk implementation for busybox
5 * Copyright (C) 2002 by Dmitry Zakharov <dmit@crp.bank.gov.ua>
7 * Licensed under the GPL v2 or later, see the file LICENSE in this tarball.
14 /* This is a NOEXEC applet. Be very careful! */
21 #define VF_NUMBER 0x0001 /* 1 = primary type is number */
22 #define VF_ARRAY 0x0002 /* 1 = it's an array */
24 #define VF_CACHED 0x0100 /* 1 = num/str value has cached str/num eq */
25 #define VF_USER 0x0200 /* 1 = user input (may be numeric string) */
26 #define VF_SPECIAL 0x0400 /* 1 = requires extra handling when changed */
27 #define VF_WALK 0x0800 /* 1 = variable has alloc'd x.walker list */
28 #define VF_FSTR 0x1000 /* 1 = var::string points to fstring buffer */
29 #define VF_CHILD 0x2000 /* 1 = function arg; x.parent points to source */
30 #define VF_DIRTY 0x4000 /* 1 = variable was set explicitly */
32 /* these flags are static, don't change them when value is changed */
33 #define VF_DONTTOUCH (VF_ARRAY | VF_SPECIAL | VF_WALK | VF_CHILD | VF_DIRTY)
36 typedef struct var_s {
37 unsigned type; /* flags */
41 int aidx; /* func arg idx (for compilation stage) */
42 struct xhash_s *array; /* array ptr */
43 struct var_s *parent; /* for func args, ptr to actual parameter */
44 char **walker; /* list of array elements (for..in) */
48 /* Node chain (pattern-action chain, BEGIN, END, function bodies) */
49 typedef struct chain_s {
52 const char *programname;
56 typedef struct func_s {
62 typedef struct rstream_s {
71 typedef struct hash_item_s {
73 struct var_s v; /* variable/array hash */
74 struct rstream_s rs; /* redirect streams hash */
75 struct func_s f; /* functions hash */
77 struct hash_item_s *next; /* next in chain */
78 char name[1]; /* really it's longer */
81 typedef struct xhash_s {
82 unsigned nel; /* num of elements */
83 unsigned csize; /* current hash size */
84 unsigned nprime; /* next hash size in PRIMES[] */
85 unsigned glen; /* summary length of item names */
86 struct hash_item_s **items;
90 typedef struct node_s {
111 /* Block of temporary variables */
112 typedef struct nvblock_s {
115 struct nvblock_s *prev;
116 struct nvblock_s *next;
120 typedef struct tsplitter_s {
125 /* simple token classes */
126 /* Order and hex values are very important!!! See next_token() */
127 #define TC_SEQSTART 1 /* ( */
128 #define TC_SEQTERM (1 << 1) /* ) */
129 #define TC_REGEXP (1 << 2) /* /.../ */
130 #define TC_OUTRDR (1 << 3) /* | > >> */
131 #define TC_UOPPOST (1 << 4) /* unary postfix operator */
132 #define TC_UOPPRE1 (1 << 5) /* unary prefix operator */
133 #define TC_BINOPX (1 << 6) /* two-opnd operator */
134 #define TC_IN (1 << 7)
135 #define TC_COMMA (1 << 8)
136 #define TC_PIPE (1 << 9) /* input redirection pipe */
137 #define TC_UOPPRE2 (1 << 10) /* unary prefix operator */
138 #define TC_ARRTERM (1 << 11) /* ] */
139 #define TC_GRPSTART (1 << 12) /* { */
140 #define TC_GRPTERM (1 << 13) /* } */
141 #define TC_SEMICOL (1 << 14)
142 #define TC_NEWLINE (1 << 15)
143 #define TC_STATX (1 << 16) /* ctl statement (for, next...) */
144 #define TC_WHILE (1 << 17)
145 #define TC_ELSE (1 << 18)
146 #define TC_BUILTIN (1 << 19)
147 #define TC_GETLINE (1 << 20)
148 #define TC_FUNCDECL (1 << 21) /* `function' `func' */
149 #define TC_BEGIN (1 << 22)
150 #define TC_END (1 << 23)
151 #define TC_EOF (1 << 24)
152 #define TC_VARIABLE (1 << 25)
153 #define TC_ARRAY (1 << 26)
154 #define TC_FUNCTION (1 << 27)
155 #define TC_STRING (1 << 28)
156 #define TC_NUMBER (1 << 29)
158 #define TC_UOPPRE (TC_UOPPRE1 | TC_UOPPRE2)
160 /* combined token classes */
161 #define TC_BINOP (TC_BINOPX | TC_COMMA | TC_PIPE | TC_IN)
162 #define TC_UNARYOP (TC_UOPPRE | TC_UOPPOST)
163 #define TC_OPERAND (TC_VARIABLE | TC_ARRAY | TC_FUNCTION \
164 | TC_BUILTIN | TC_GETLINE | TC_SEQSTART | TC_STRING | TC_NUMBER)
166 #define TC_STATEMNT (TC_STATX | TC_WHILE)
167 #define TC_OPTERM (TC_SEMICOL | TC_NEWLINE)
169 /* word tokens, cannot mean something else if not expected */
170 #define TC_WORD (TC_IN | TC_STATEMNT | TC_ELSE | TC_BUILTIN \
171 | TC_GETLINE | TC_FUNCDECL | TC_BEGIN | TC_END)
173 /* discard newlines after these */
174 #define TC_NOTERM (TC_COMMA | TC_GRPSTART | TC_GRPTERM \
175 | TC_BINOP | TC_OPTERM)
177 /* what can expression begin with */
178 #define TC_OPSEQ (TC_OPERAND | TC_UOPPRE | TC_REGEXP)
179 /* what can group begin with */
180 #define TC_GRPSEQ (TC_OPSEQ | TC_OPTERM | TC_STATEMNT | TC_GRPSTART)
182 /* if previous token class is CONCAT1 and next is CONCAT2, concatenation */
183 /* operator is inserted between them */
184 #define TC_CONCAT1 (TC_VARIABLE | TC_ARRTERM | TC_SEQTERM \
185 | TC_STRING | TC_NUMBER | TC_UOPPOST)
186 #define TC_CONCAT2 (TC_OPERAND | TC_UOPPRE)
188 #define OF_RES1 0x010000
189 #define OF_RES2 0x020000
190 #define OF_STR1 0x040000
191 #define OF_STR2 0x080000
192 #define OF_NUM1 0x100000
193 #define OF_CHECKED 0x200000
195 /* combined operator flags */
198 #define xS (OF_RES2 | OF_STR2)
200 #define VV (OF_RES1 | OF_RES2)
201 #define Nx (OF_RES1 | OF_NUM1)
202 #define NV (OF_RES1 | OF_NUM1 | OF_RES2)
203 #define Sx (OF_RES1 | OF_STR1)
204 #define SV (OF_RES1 | OF_STR1 | OF_RES2)
205 #define SS (OF_RES1 | OF_STR1 | OF_RES2 | OF_STR2)
207 #define OPCLSMASK 0xFF00
208 #define OPNMASK 0x007F
210 /* operator priority is a highest byte (even: r->l, odd: l->r grouping)
211 * For builtins it has different meaning: n n s3 s2 s1 v3 v2 v1,
212 * n - min. number of args, vN - resolve Nth arg to var, sN - resolve to string
214 #define P(x) (x << 24)
215 #define PRIMASK 0x7F000000
216 #define PRIMASK2 0x7E000000
218 /* Operation classes */
220 #define SHIFT_TIL_THIS 0x0600
221 #define RECUR_FROM_THIS 0x1000
224 OC_DELETE = 0x0100, OC_EXEC = 0x0200, OC_NEWSOURCE = 0x0300,
225 OC_PRINT = 0x0400, OC_PRINTF = 0x0500, OC_WALKINIT = 0x0600,
227 OC_BR = 0x0700, OC_BREAK = 0x0800, OC_CONTINUE = 0x0900,
228 OC_EXIT = 0x0a00, OC_NEXT = 0x0b00, OC_NEXTFILE = 0x0c00,
229 OC_TEST = 0x0d00, OC_WALKNEXT = 0x0e00,
231 OC_BINARY = 0x1000, OC_BUILTIN = 0x1100, OC_COLON = 0x1200,
232 OC_COMMA = 0x1300, OC_COMPARE = 0x1400, OC_CONCAT = 0x1500,
233 OC_FBLTIN = 0x1600, OC_FIELD = 0x1700, OC_FNARG = 0x1800,
234 OC_FUNC = 0x1900, OC_GETLINE = 0x1a00, OC_IN = 0x1b00,
235 OC_LAND = 0x1c00, OC_LOR = 0x1d00, OC_MATCH = 0x1e00,
236 OC_MOVE = 0x1f00, OC_PGETLINE = 0x2000, OC_REGEXP = 0x2100,
237 OC_REPLACE = 0x2200, OC_RETURN = 0x2300, OC_SPRINTF = 0x2400,
238 OC_TERNARY = 0x2500, OC_UNARY = 0x2600, OC_VAR = 0x2700,
241 ST_IF = 0x3000, ST_DO = 0x3100, ST_FOR = 0x3200,
245 /* simple builtins */
247 F_in, F_rn, F_co, F_ex, F_lg, F_si, F_sq, F_sr,
248 F_ti, F_le, F_sy, F_ff, F_cl
253 B_a2, B_ix, B_ma, B_sp, B_ss, B_ti, B_mt, B_lo, B_up,
255 B_an, B_co, B_ls, B_or, B_rs, B_xo,
258 /* tokens and their corresponding info values */
260 #define NTC "\377" /* switch to next token class (tc<<1) */
263 #define OC_B OC_BUILTIN
265 static const char tokenlist[] ALIGN1 =
268 "\1/" NTC /* REGEXP */
269 "\2>>" "\1>" "\1|" NTC /* OUTRDR */
270 "\2++" "\2--" NTC /* UOPPOST */
271 "\2++" "\2--" "\1$" NTC /* UOPPRE1 */
272 "\2==" "\1=" "\2+=" "\2-=" /* BINOPX */
273 "\2*=" "\2/=" "\2%=" "\2^="
274 "\1+" "\1-" "\3**=" "\2**"
275 "\1/" "\1%" "\1^" "\1*"
276 "\2!=" "\2>=" "\2<=" "\1>"
277 "\1<" "\2!~" "\1~" "\2&&"
278 "\2||" "\1?" "\1:" NTC
282 "\1+" "\1-" "\1!" NTC /* UOPPRE2 */
288 "\2if" "\2do" "\3for" "\5break" /* STATX */
289 "\10continue" "\6delete" "\5print"
290 "\6printf" "\4next" "\10nextfile"
291 "\6return" "\4exit" NTC
295 "\3and" "\5compl" "\6lshift" "\2or"
297 "\5close" "\6system" "\6fflush" "\5atan2" /* BUILTIN */
298 "\3cos" "\3exp" "\3int" "\3log"
299 "\4rand" "\3sin" "\4sqrt" "\5srand"
300 "\6gensub" "\4gsub" "\5index" "\6length"
301 "\5match" "\5split" "\7sprintf" "\3sub"
302 "\6substr" "\7systime" "\10strftime" "\6mktime"
303 "\7tolower" "\7toupper" NTC
305 "\4func" "\10function" NTC
310 static const uint32_t tokeninfo[] = {
314 xS|'a', xS|'w', xS|'|',
315 OC_UNARY|xV|P(9)|'p', OC_UNARY|xV|P(9)|'m',
316 OC_UNARY|xV|P(9)|'P', OC_UNARY|xV|P(9)|'M',
318 OC_COMPARE|VV|P(39)|5, OC_MOVE|VV|P(74),
319 OC_REPLACE|NV|P(74)|'+', OC_REPLACE|NV|P(74)|'-',
320 OC_REPLACE|NV|P(74)|'*', OC_REPLACE|NV|P(74)|'/',
321 OC_REPLACE|NV|P(74)|'%', OC_REPLACE|NV|P(74)|'&',
322 OC_BINARY|NV|P(29)|'+', OC_BINARY|NV|P(29)|'-',
323 OC_REPLACE|NV|P(74)|'&', OC_BINARY|NV|P(15)|'&',
324 OC_BINARY|NV|P(25)|'/', OC_BINARY|NV|P(25)|'%',
325 OC_BINARY|NV|P(15)|'&', OC_BINARY|NV|P(25)|'*',
326 OC_COMPARE|VV|P(39)|4, OC_COMPARE|VV|P(39)|3,
327 OC_COMPARE|VV|P(39)|0, OC_COMPARE|VV|P(39)|1,
328 OC_COMPARE|VV|P(39)|2, OC_MATCH|Sx|P(45)|'!',
329 OC_MATCH|Sx|P(45)|'~', OC_LAND|Vx|P(55),
330 OC_LOR|Vx|P(59), OC_TERNARY|Vx|P(64)|'?',
331 OC_COLON|xx|P(67)|':',
334 OC_PGETLINE|SV|P(37),
335 OC_UNARY|xV|P(19)|'+', OC_UNARY|xV|P(19)|'-',
336 OC_UNARY|xV|P(19)|'!',
342 ST_IF, ST_DO, ST_FOR, OC_BREAK,
343 OC_CONTINUE, OC_DELETE|Vx, OC_PRINT,
344 OC_PRINTF, OC_NEXT, OC_NEXTFILE,
345 OC_RETURN|Vx, OC_EXIT|Nx,
349 OC_B|B_an|P(0x83), OC_B|B_co|P(0x41), OC_B|B_ls|P(0x83), OC_B|B_or|P(0x83),
350 OC_B|B_rs|P(0x83), OC_B|B_xo|P(0x83),
351 OC_FBLTIN|Sx|F_cl, OC_FBLTIN|Sx|F_sy, OC_FBLTIN|Sx|F_ff, OC_B|B_a2|P(0x83),
352 OC_FBLTIN|Nx|F_co, OC_FBLTIN|Nx|F_ex, OC_FBLTIN|Nx|F_in, OC_FBLTIN|Nx|F_lg,
353 OC_FBLTIN|F_rn, OC_FBLTIN|Nx|F_si, OC_FBLTIN|Nx|F_sq, OC_FBLTIN|Nx|F_sr,
354 OC_B|B_ge|P(0xd6), OC_B|B_gs|P(0xb6), OC_B|B_ix|P(0x9b), OC_FBLTIN|Sx|F_le,
355 OC_B|B_ma|P(0x89), OC_B|B_sp|P(0x8b), OC_SPRINTF, OC_B|B_su|P(0xb6),
356 OC_B|B_ss|P(0x8f), OC_FBLTIN|F_ti, OC_B|B_ti|P(0x0b), OC_B|B_mt|P(0x0b),
357 OC_B|B_lo|P(0x49), OC_B|B_up|P(0x49),
364 /* internal variable names and their initial values */
365 /* asterisk marks SPECIAL vars; $ is just no-named Field0 */
367 CONVFMT, OFMT, FS, OFS,
368 ORS, RS, RT, FILENAME,
369 SUBSEP, F0, ARGIND, ARGC,
370 ARGV, ERRNO, FNR, NR,
371 NF, IGNORECASE, ENVIRON, NUM_INTERNAL_VARS
374 static const char vNames[] ALIGN1 =
375 "CONVFMT\0" "OFMT\0" "FS\0*" "OFS\0"
376 "ORS\0" "RS\0*" "RT\0" "FILENAME\0"
377 "SUBSEP\0" "$\0*" "ARGIND\0" "ARGC\0"
378 "ARGV\0" "ERRNO\0" "FNR\0" "NR\0"
379 "NF\0*" "IGNORECASE\0*" "ENVIRON\0" "\0";
381 static const char vValues[] ALIGN1 =
382 "%.6g\0" "%.6g\0" " \0" " \0"
383 "\n\0" "\n\0" "\0" "\0"
384 "\034\0" "\0" "\377";
386 /* hash size may grow to these values */
387 #define FIRST_PRIME 61
388 static const uint16_t PRIMES[] ALIGN2 = { 251, 1021, 4093, 16381, 65521 };
391 /* Globals. Split in two parts so that first one is addressed
392 * with (mostly short) negative offsets.
393 * NB: it's unsafe to put members of type "double"
394 * into globals2 (gcc may fail to align them).
398 chain beginseq, mainseq, endseq;
400 node *break_ptr, *continue_ptr;
402 xhash *vhash, *ahash, *fdhash, *fnhash;
403 const char *g_progname;
406 int maxfields; /* used in fsrealloc() only */
415 smallint is_f0_split;
418 uint32_t t_info; /* often used */
424 var *intvar[NUM_INTERNAL_VARS]; /* often used */
426 /* former statics from various functions */
427 char *split_f0__fstrings;
429 uint32_t next_token__save_tclass;
430 uint32_t next_token__save_info;
431 uint32_t next_token__ltclass;
432 smallint next_token__concat_inserted;
434 smallint next_input_file__files_happen;
435 rstream next_input_file__rsm;
437 var *evaluate__fnargs;
438 unsigned evaluate__seed;
439 regex_t evaluate__sreg;
443 tsplitter exec_builtin__tspl;
445 /* biggest and least used members go last */
446 tsplitter fsplitter, rsplitter;
448 #define G1 (ptr_to_globals[-1])
449 #define G (*(struct globals2 *)ptr_to_globals)
450 /* For debug. nm --size-sort awk.o | grep -vi ' [tr] ' */
451 /*char G1size[sizeof(G1)]; - 0x74 */
452 /*char Gsize[sizeof(G)]; - 0x1c4 */
453 /* Trying to keep most of members accessible with short offsets: */
454 /*char Gofs_seed[offsetof(struct globals2, evaluate__seed)]; - 0x90 */
455 #define t_double (G1.t_double )
456 #define beginseq (G1.beginseq )
457 #define mainseq (G1.mainseq )
458 #define endseq (G1.endseq )
459 #define seq (G1.seq )
460 #define break_ptr (G1.break_ptr )
461 #define continue_ptr (G1.continue_ptr)
463 #define vhash (G1.vhash )
464 #define ahash (G1.ahash )
465 #define fdhash (G1.fdhash )
466 #define fnhash (G1.fnhash )
467 #define g_progname (G1.g_progname )
468 #define g_lineno (G1.g_lineno )
469 #define nfields (G1.nfields )
470 #define maxfields (G1.maxfields )
471 #define Fields (G1.Fields )
472 #define g_cb (G1.g_cb )
473 #define g_pos (G1.g_pos )
474 #define g_buf (G1.g_buf )
475 #define icase (G1.icase )
476 #define exiting (G1.exiting )
477 #define nextrec (G1.nextrec )
478 #define nextfile (G1.nextfile )
479 #define is_f0_split (G1.is_f0_split )
480 #define t_info (G.t_info )
481 #define t_tclass (G.t_tclass )
482 #define t_string (G.t_string )
483 #define t_lineno (G.t_lineno )
484 #define t_rollback (G.t_rollback )
485 #define intvar (G.intvar )
486 #define fsplitter (G.fsplitter )
487 #define rsplitter (G.rsplitter )
488 #define INIT_G() do { \
489 SET_PTR_TO_GLOBALS((char*)xzalloc(sizeof(G1)+sizeof(G)) + sizeof(G1)); \
490 G.next_token__ltclass = TC_OPTERM; \
491 G.evaluate__seed = 1; \
495 /* function prototypes */
496 static void handle_special(var *);
497 static node *parse_expr(uint32_t);
498 static void chain_group(void);
499 static var *evaluate(node *, var *);
500 static rstream *next_input_file(void);
501 static int fmt_num(char *, int, const char *, double, int);
502 static int awk_exit(int) NORETURN;
504 /* ---- error handling ---- */
506 static const char EMSG_INTERNAL_ERROR[] ALIGN1 = "Internal error";
507 static const char EMSG_UNEXP_EOS[] ALIGN1 = "Unexpected end of string";
508 static const char EMSG_UNEXP_TOKEN[] ALIGN1 = "Unexpected token";
509 static const char EMSG_DIV_BY_ZERO[] ALIGN1 = "Division by zero";
510 static const char EMSG_INV_FMT[] ALIGN1 = "Invalid format specifier";
511 static const char EMSG_TOO_FEW_ARGS[] ALIGN1 = "Too few arguments for builtin";
512 static const char EMSG_NOT_ARRAY[] ALIGN1 = "Not an array";
513 static const char EMSG_POSSIBLE_ERROR[] ALIGN1 = "Possible syntax error";
514 static const char EMSG_UNDEF_FUNC[] ALIGN1 = "Call to undefined function";
515 #if !ENABLE_FEATURE_AWK_LIBM
516 static const char EMSG_NO_MATH[] ALIGN1 = "Math support is not compiled in";
519 static void zero_out_var(var *vp)
521 memset(vp, 0, sizeof(*vp));
524 static void syntax_error(const char *message) NORETURN;
525 static void syntax_error(const char *message)
527 bb_error_msg_and_die("%s:%i: %s", g_progname, g_lineno, message);
530 /* ---- hash stuff ---- */
532 static unsigned hashidx(const char *name)
537 idx = *name++ + (idx << 6) - idx;
541 /* create new hash */
542 static xhash *hash_init(void)
546 newhash = xzalloc(sizeof(*newhash));
547 newhash->csize = FIRST_PRIME;
548 newhash->items = xzalloc(FIRST_PRIME * sizeof(newhash->items[0]));
553 /* find item in hash, return ptr to data, NULL if not found */
554 static void *hash_search(xhash *hash, const char *name)
558 hi = hash->items[hashidx(name) % hash->csize];
560 if (strcmp(hi->name, name) == 0)
567 /* grow hash if it becomes too big */
568 static void hash_rebuild(xhash *hash)
570 unsigned newsize, i, idx;
571 hash_item **newitems, *hi, *thi;
573 if (hash->nprime == ARRAY_SIZE(PRIMES))
576 newsize = PRIMES[hash->nprime++];
577 newitems = xzalloc(newsize * sizeof(newitems[0]));
579 for (i = 0; i < hash->csize; i++) {
584 idx = hashidx(thi->name) % newsize;
585 thi->next = newitems[idx];
591 hash->csize = newsize;
592 hash->items = newitems;
595 /* find item in hash, add it if necessary. Return ptr to data */
596 static void *hash_find(xhash *hash, const char *name)
602 hi = hash_search(hash, name);
604 if (++hash->nel / hash->csize > 10)
607 l = strlen(name) + 1;
608 hi = xzalloc(sizeof(*hi) + l);
609 strcpy(hi->name, name);
611 idx = hashidx(name) % hash->csize;
612 hi->next = hash->items[idx];
613 hash->items[idx] = hi;
619 #define findvar(hash, name) ((var*) hash_find((hash), (name)))
620 #define newvar(name) ((var*) hash_find(vhash, (name)))
621 #define newfile(name) ((rstream*)hash_find(fdhash, (name)))
622 #define newfunc(name) ((func*) hash_find(fnhash, (name)))
624 static void hash_remove(xhash *hash, const char *name)
626 hash_item *hi, **phi;
628 phi = &(hash->items[hashidx(name) % hash->csize]);
631 if (strcmp(hi->name, name) == 0) {
632 hash->glen -= (strlen(name) + 1);
642 /* ------ some useful functions ------ */
644 static void skip_spaces(char **s)
649 if (*p == '\\' && p[1] == '\n') {
652 } else if (*p != ' ' && *p != '\t') {
660 static char *nextword(char **s)
668 static char nextchar(char **s)
675 c = bb_process_escape_sequence((const char**)s);
676 if (c == '\\' && *s == pps)
681 static ALWAYS_INLINE int isalnum_(int c)
683 return (isalnum(c) || c == '_');
686 static double my_strtod(char **pp)
690 && ((((*pp)[1] | 0x20) == 'x') || isdigit((*pp)[1]))
692 return strtoull(*pp, pp, 0);
695 return strtod(*pp, pp);
698 /* -------- working with variables (set/get/copy/etc) -------- */
700 static xhash *iamarray(var *v)
704 while (a->type & VF_CHILD)
707 if (!(a->type & VF_ARRAY)) {
709 a->x.array = hash_init();
714 static void clear_array(xhash *array)
719 for (i = 0; i < array->csize; i++) {
720 hi = array->items[i];
724 free(thi->data.v.string);
727 array->items[i] = NULL;
729 array->glen = array->nel = 0;
732 /* clear a variable */
733 static var *clrvar(var *v)
735 if (!(v->type & VF_FSTR))
738 v->type &= VF_DONTTOUCH;
744 /* assign string value to variable */
745 static var *setvar_p(var *v, char *value)
753 /* same as setvar_p but make a copy of string */
754 static var *setvar_s(var *v, const char *value)
756 return setvar_p(v, (value && *value) ? xstrdup(value) : NULL);
759 /* same as setvar_s but sets USER flag */
760 static var *setvar_u(var *v, const char *value)
762 v = setvar_s(v, value);
767 /* set array element to user string */
768 static void setari_u(var *a, int idx, const char *s)
772 v = findvar(iamarray(a), itoa(idx));
776 /* assign numeric value to variable */
777 static var *setvar_i(var *v, double value)
780 v->type |= VF_NUMBER;
786 static const char *getvar_s(var *v)
788 /* if v is numeric and has no cached string, convert it to string */
789 if ((v->type & (VF_NUMBER | VF_CACHED)) == VF_NUMBER) {
790 fmt_num(g_buf, MAXVARFMT, getvar_s(intvar[CONVFMT]), v->number, TRUE);
791 v->string = xstrdup(g_buf);
792 v->type |= VF_CACHED;
794 return (v->string == NULL) ? "" : v->string;
797 static double getvar_i(var *v)
801 if ((v->type & (VF_NUMBER | VF_CACHED)) == 0) {
805 v->number = my_strtod(&s);
806 if (v->type & VF_USER) {
814 v->type |= VF_CACHED;
819 /* Used for operands of bitwise ops */
820 static unsigned long getvar_i_int(var *v)
822 double d = getvar_i(v);
824 /* Casting doubles to longs is undefined for values outside
825 * of target type range. Try to widen it as much as possible */
827 return (unsigned long)d;
828 /* Why? Think about d == -4294967295.0 (assuming 32bit longs) */
829 return - (long) (unsigned long) (-d);
832 static var *copyvar(var *dest, const var *src)
836 dest->type |= (src->type & ~(VF_DONTTOUCH | VF_FSTR));
837 dest->number = src->number;
839 dest->string = xstrdup(src->string);
841 handle_special(dest);
845 static var *incvar(var *v)
847 return setvar_i(v, getvar_i(v) + 1.0);
850 /* return true if v is number or numeric string */
851 static int is_numeric(var *v)
854 return ((v->type ^ VF_DIRTY) & (VF_NUMBER | VF_USER | VF_DIRTY));
857 /* return 1 when value of v corresponds to true, 0 otherwise */
858 static int istrue(var *v)
861 return (v->number != 0);
862 return (v->string && v->string[0]);
865 /* temporary variables allocator. Last allocated should be first freed */
866 static var *nvalloc(int n)
874 if ((g_cb->pos - g_cb->nv) + n <= g_cb->size)
880 size = (n <= MINNVBLOCK) ? MINNVBLOCK : n;
881 g_cb = xzalloc(sizeof(nvblock) + size * sizeof(var));
883 g_cb->pos = g_cb->nv;
885 /*g_cb->next = NULL; - xzalloc did it */
893 while (v < g_cb->pos) {
902 static void nvfree(var *v)
906 if (v < g_cb->nv || v >= g_cb->pos)
907 syntax_error(EMSG_INTERNAL_ERROR);
909 for (p = v; p < g_cb->pos; p++) {
910 if ((p->type & (VF_ARRAY | VF_CHILD)) == VF_ARRAY) {
911 clear_array(iamarray(p));
912 free(p->x.array->items);
915 if (p->type & VF_WALK)
922 while (g_cb->prev && g_cb->pos == g_cb->nv) {
927 /* ------- awk program text parsing ------- */
929 /* Parse next token pointed by global pos, place results into global ttt.
930 * If token isn't expected, give away. Return token class
932 static uint32_t next_token(uint32_t expected)
934 #define concat_inserted (G.next_token__concat_inserted)
935 #define save_tclass (G.next_token__save_tclass)
936 #define save_info (G.next_token__save_info)
937 /* Initialized to TC_OPTERM: */
938 #define ltclass (G.next_token__ltclass)
949 } else if (concat_inserted) {
950 concat_inserted = FALSE;
951 t_tclass = save_tclass;
960 while (*p != '\n' && *p != '\0')
969 } else if (*p == '\"') {
973 if (*p == '\0' || *p == '\n')
974 syntax_error(EMSG_UNEXP_EOS);
975 *(s++) = nextchar(&p);
981 } else if ((expected & TC_REGEXP) && *p == '/') {
985 if (*p == '\0' || *p == '\n')
986 syntax_error(EMSG_UNEXP_EOS);
990 *(s-1) = bb_process_escape_sequence((const char **)&p);
1001 } else if (*p == '.' || isdigit(*p)) {
1003 t_double = my_strtod(&p);
1005 syntax_error(EMSG_UNEXP_TOKEN);
1009 /* search for something known */
1019 /* if token class is expected, token
1020 * matches and it's not a longer word,
1021 * then this is what we are looking for
1023 if ((tc & (expected | TC_WORD | TC_NEWLINE))
1024 && *tl == *p && strncmp(p, tl, l) == 0
1025 && !((tc & TC_WORD) && isalnum_(p[l]))
1036 /* it's a name (var/array/function),
1037 * otherwise it's something wrong
1040 syntax_error(EMSG_UNEXP_TOKEN);
1043 while (isalnum_(*(++p))) {
1048 /* also consume whitespace between functionname and bracket */
1049 if (!(expected & TC_VARIABLE))
1063 /* skipping newlines in some cases */
1064 if ((ltclass & TC_NOTERM) && (tc & TC_NEWLINE))
1067 /* insert concatenation operator when needed */
1068 if ((ltclass & TC_CONCAT1) && (tc & TC_CONCAT2) && (expected & TC_BINOP)) {
1069 concat_inserted = TRUE;
1073 t_info = OC_CONCAT | SS | P(35);
1080 /* Are we ready for this? */
1081 if (!(ltclass & expected))
1082 syntax_error((ltclass & (TC_NEWLINE | TC_EOF)) ?
1083 EMSG_UNEXP_EOS : EMSG_UNEXP_TOKEN);
1086 #undef concat_inserted
1092 static void rollback_token(void)
1097 static node *new_node(uint32_t info)
1101 n = xzalloc(sizeof(node));
1103 n->lineno = g_lineno;
1107 static node *mk_re_node(const char *s, node *n, regex_t *re)
1109 n->info = OC_REGEXP;
1112 xregcomp(re, s, REG_EXTENDED);
1113 xregcomp(re + 1, s, REG_EXTENDED | REG_ICASE);
1118 static node *condition(void)
1120 next_token(TC_SEQSTART);
1121 return parse_expr(TC_SEQTERM);
1124 /* parse expression terminated by given argument, return ptr
1125 * to built subtree. Terminator is eaten by parse_expr */
1126 static node *parse_expr(uint32_t iexp)
1135 sn.r.n = glptr = NULL;
1136 xtc = TC_OPERAND | TC_UOPPRE | TC_REGEXP | iexp;
1138 while (!((tc = next_token(xtc)) & iexp)) {
1139 if (glptr && (t_info == (OC_COMPARE | VV | P(39) | 2))) {
1140 /* input redirection (<) attached to glptr node */
1141 cn = glptr->l.n = new_node(OC_CONCAT | SS | P(37));
1143 xtc = TC_OPERAND | TC_UOPPRE;
1146 } else if (tc & (TC_BINOP | TC_UOPPOST)) {
1147 /* for binary and postfix-unary operators, jump back over
1148 * previous operators with higher priority */
1150 while (((t_info & PRIMASK) > (vn->a.n->info & PRIMASK2))
1151 || ((t_info == vn->info) && ((t_info & OPCLSMASK) == OC_COLON))
1155 if ((t_info & OPCLSMASK) == OC_TERNARY)
1157 cn = vn->a.n->r.n = new_node(t_info);
1159 if (tc & TC_BINOP) {
1161 xtc = TC_OPERAND | TC_UOPPRE | TC_REGEXP;
1162 if ((t_info & OPCLSMASK) == OC_PGETLINE) {
1164 next_token(TC_GETLINE);
1165 /* give maximum priority to this pipe */
1166 cn->info &= ~PRIMASK;
1167 xtc = TC_OPERAND | TC_UOPPRE | TC_BINOP | iexp;
1171 xtc = TC_OPERAND | TC_UOPPRE | TC_BINOP | iexp;
1176 /* for operands and prefix-unary operators, attach them
1179 cn = vn->r.n = new_node(t_info);
1181 xtc = TC_OPERAND | TC_UOPPRE | TC_REGEXP;
1182 if (tc & (TC_OPERAND | TC_REGEXP)) {
1183 xtc = TC_UOPPRE | TC_UOPPOST | TC_BINOP | TC_OPERAND | iexp;
1184 /* one should be very careful with switch on tclass -
1185 * only simple tclasses should be used! */
1190 v = hash_search(ahash, t_string);
1192 cn->info = OC_FNARG;
1193 cn->l.i = v->x.aidx;
1195 cn->l.v = newvar(t_string);
1197 if (tc & TC_ARRAY) {
1199 cn->r.n = parse_expr(TC_ARRTERM);
1206 v = cn->l.v = xzalloc(sizeof(var));
1208 setvar_i(v, t_double);
1210 setvar_s(v, t_string);
1214 mk_re_node(t_string, cn, xzalloc(sizeof(regex_t)*2));
1219 cn->r.f = newfunc(t_string);
1220 cn->l.n = condition();
1224 cn = vn->r.n = parse_expr(TC_SEQTERM);
1230 xtc = TC_OPERAND | TC_UOPPRE | TC_BINOP | iexp;
1234 cn->l.n = condition();
1243 /* add node to chain. Return ptr to alloc'd node */
1244 static node *chain_node(uint32_t info)
1249 seq->first = seq->last = new_node(0);
1251 if (seq->programname != g_progname) {
1252 seq->programname = g_progname;
1253 n = chain_node(OC_NEWSOURCE);
1254 n->l.s = xstrdup(g_progname);
1259 seq->last = n->a.n = new_node(OC_DONE);
1264 static void chain_expr(uint32_t info)
1268 n = chain_node(info);
1269 n->l.n = parse_expr(TC_OPTERM | TC_GRPTERM);
1270 if (t_tclass & TC_GRPTERM)
1274 static node *chain_loop(node *nn)
1276 node *n, *n2, *save_brk, *save_cont;
1278 save_brk = break_ptr;
1279 save_cont = continue_ptr;
1281 n = chain_node(OC_BR | Vx);
1282 continue_ptr = new_node(OC_EXEC);
1283 break_ptr = new_node(OC_EXEC);
1285 n2 = chain_node(OC_EXEC | Vx);
1288 continue_ptr->a.n = n2;
1289 break_ptr->a.n = n->r.n = seq->last;
1291 continue_ptr = save_cont;
1292 break_ptr = save_brk;
1297 /* parse group and attach it to chain */
1298 static void chain_group(void)
1304 c = next_token(TC_GRPSEQ);
1305 } while (c & TC_NEWLINE);
1307 if (c & TC_GRPSTART) {
1308 while (next_token(TC_GRPSEQ | TC_GRPTERM) != TC_GRPTERM) {
1309 if (t_tclass & TC_NEWLINE) continue;
1313 } else if (c & (TC_OPSEQ | TC_OPTERM)) {
1315 chain_expr(OC_EXEC | Vx);
1316 } else { /* TC_STATEMNT */
1317 switch (t_info & OPCLSMASK) {
1319 n = chain_node(OC_BR | Vx);
1320 n->l.n = condition();
1322 n2 = chain_node(OC_EXEC);
1324 if (next_token(TC_GRPSEQ | TC_GRPTERM | TC_ELSE) == TC_ELSE) {
1326 n2->a.n = seq->last;
1334 n = chain_loop(NULL);
1339 n2 = chain_node(OC_EXEC);
1340 n = chain_loop(NULL);
1342 next_token(TC_WHILE);
1343 n->l.n = condition();
1347 next_token(TC_SEQSTART);
1348 n2 = parse_expr(TC_SEMICOL | TC_SEQTERM);
1349 if (t_tclass & TC_SEQTERM) { /* for-in */
1350 if ((n2->info & OPCLSMASK) != OC_IN)
1351 syntax_error(EMSG_UNEXP_TOKEN);
1352 n = chain_node(OC_WALKINIT | VV);
1355 n = chain_loop(NULL);
1356 n->info = OC_WALKNEXT | Vx;
1358 } else { /* for (;;) */
1359 n = chain_node(OC_EXEC | Vx);
1361 n2 = parse_expr(TC_SEMICOL);
1362 n3 = parse_expr(TC_SEQTERM);
1372 n = chain_node(t_info);
1373 n->l.n = parse_expr(TC_OPTERM | TC_OUTRDR | TC_GRPTERM);
1374 if (t_tclass & TC_OUTRDR) {
1376 n->r.n = parse_expr(TC_OPTERM | TC_GRPTERM);
1378 if (t_tclass & TC_GRPTERM)
1383 n = chain_node(OC_EXEC);
1388 n = chain_node(OC_EXEC);
1389 n->a.n = continue_ptr;
1392 /* delete, next, nextfile, return, exit */
1399 static void parse_program(char *p)
1408 while ((tclass = next_token(TC_EOF | TC_OPSEQ | TC_GRPSTART |
1409 TC_OPTERM | TC_BEGIN | TC_END | TC_FUNCDECL)) != TC_EOF) {
1411 if (tclass & TC_OPTERM)
1415 if (tclass & TC_BEGIN) {
1419 } else if (tclass & TC_END) {
1423 } else if (tclass & TC_FUNCDECL) {
1424 next_token(TC_FUNCTION);
1426 f = newfunc(t_string);
1427 f->body.first = NULL;
1429 while (next_token(TC_VARIABLE | TC_SEQTERM) & TC_VARIABLE) {
1430 v = findvar(ahash, t_string);
1431 v->x.aidx = (f->nargs)++;
1433 if (next_token(TC_COMMA | TC_SEQTERM) & TC_SEQTERM)
1440 } else if (tclass & TC_OPSEQ) {
1442 cn = chain_node(OC_TEST);
1443 cn->l.n = parse_expr(TC_OPTERM | TC_EOF | TC_GRPSTART);
1444 if (t_tclass & TC_GRPSTART) {
1448 chain_node(OC_PRINT);
1450 cn->r.n = mainseq.last;
1452 } else /* if (tclass & TC_GRPSTART) */ {
1460 /* -------- program execution part -------- */
1462 static node *mk_splitter(const char *s, tsplitter *spl)
1470 if ((n->info & OPCLSMASK) == OC_REGEXP) {
1472 regfree(ire); // TODO: nuke ire, use re+1?
1474 if (strlen(s) > 1) {
1475 mk_re_node(s, n, re);
1477 n->info = (uint32_t) *s;
1483 /* use node as a regular expression. Supplied with node ptr and regex_t
1484 * storage space. Return ptr to regex (if result points to preg, it should
1485 * be later regfree'd manually
1487 static regex_t *as_regex(node *op, regex_t *preg)
1493 if ((op->info & OPCLSMASK) == OC_REGEXP) {
1494 return icase ? op->r.ire : op->l.re;
1497 s = getvar_s(evaluate(op, v));
1499 cflags = icase ? REG_EXTENDED | REG_ICASE : REG_EXTENDED;
1500 /* Testcase where REG_EXTENDED fails (unpaired '{'):
1501 * echo Hi | awk 'gsub("@(samp|code|file)\{","");'
1502 * gawk 3.1.5 eats this. We revert to ~REG_EXTENDED
1503 * (maybe gsub is not supposed to use REG_EXTENDED?).
1505 if (regcomp(preg, s, cflags)) {
1506 cflags &= ~REG_EXTENDED;
1507 xregcomp(preg, s, cflags);
1513 /* gradually increasing buffer */
1514 static void qrealloc(char **b, int n, int *size)
1516 if (!*b || n >= *size) {
1517 *size = n + (n>>1) + 80;
1518 *b = xrealloc(*b, *size);
1522 /* resize field storage space */
1523 static void fsrealloc(int size)
1527 if (size >= maxfields) {
1529 maxfields = size + 16;
1530 Fields = xrealloc(Fields, maxfields * sizeof(var));
1531 for (; i < maxfields; i++) {
1532 Fields[i].type = VF_SPECIAL;
1533 Fields[i].string = NULL;
1537 if (size < nfields) {
1538 for (i = size; i < nfields; i++) {
1545 static int awk_split(const char *s, node *spl, char **slist)
1550 regmatch_t pmatch[2]; // TODO: why [2]? [1] is enough...
1552 /* in worst case, each char would be a separate field */
1553 *slist = s1 = xzalloc(strlen(s) * 2 + 3);
1556 c[0] = c[1] = (char)spl->info;
1558 if (*getvar_s(intvar[RS]) == '\0')
1561 if ((spl->info & OPCLSMASK) == OC_REGEXP) { /* regex split */
1563 return n; /* "": zero fields */
1564 n++; /* at least one field will be there */
1566 l = strcspn(s, c+2); /* len till next NUL or \n */
1567 if (regexec(icase ? spl->r.ire : spl->l.re, s, 1, pmatch, 0) == 0
1568 && pmatch[0].rm_so <= l
1570 l = pmatch[0].rm_so;
1571 if (pmatch[0].rm_eo == 0) {
1575 n++; /* we saw yet another delimiter */
1577 pmatch[0].rm_eo = l;
1582 /* make sure we remove *all* of the separator chars */
1585 } while (++l < pmatch[0].rm_eo);
1587 s += pmatch[0].rm_eo;
1591 if (c[0] == '\0') { /* null split */
1599 if (c[0] != ' ') { /* single-character split */
1601 c[0] = toupper(c[0]);
1602 c[1] = tolower(c[1]);
1605 while ((s1 = strpbrk(s1, c))) {
1613 s = skip_whitespace(s);
1616 while (*s && !isspace(*s))
1623 static void split_f0(void)
1625 /* static char *fstrings; */
1626 #define fstrings (G.split_f0__fstrings)
1637 n = awk_split(getvar_s(intvar[F0]), &fsplitter.n, &fstrings);
1640 for (i = 0; i < n; i++) {
1641 Fields[i].string = nextword(&s);
1642 Fields[i].type |= (VF_FSTR | VF_USER | VF_DIRTY);
1645 /* set NF manually to avoid side effects */
1647 intvar[NF]->type = VF_NUMBER | VF_SPECIAL;
1648 intvar[NF]->number = nfields;
1652 /* perform additional actions when some internal variables changed */
1653 static void handle_special(var *v)
1657 const char *sep, *s;
1658 int sl, l, len, i, bsize;
1660 if (!(v->type & VF_SPECIAL))
1663 if (v == intvar[NF]) {
1664 n = (int)getvar_i(v);
1667 /* recalculate $0 */
1668 sep = getvar_s(intvar[OFS]);
1672 for (i = 0; i < n; i++) {
1673 s = getvar_s(&Fields[i]);
1676 memcpy(b+len, sep, sl);
1679 qrealloc(&b, len+l+sl, &bsize);
1680 memcpy(b+len, s, l);
1685 setvar_p(intvar[F0], b);
1688 } else if (v == intvar[F0]) {
1689 is_f0_split = FALSE;
1691 } else if (v == intvar[FS]) {
1692 mk_splitter(getvar_s(v), &fsplitter);
1694 } else if (v == intvar[RS]) {
1695 mk_splitter(getvar_s(v), &rsplitter);
1697 } else if (v == intvar[IGNORECASE]) {
1701 n = getvar_i(intvar[NF]);
1702 setvar_i(intvar[NF], n > v-Fields ? n : v-Fields+1);
1703 /* right here v is invalid. Just to note... */
1707 /* step through func/builtin/etc arguments */
1708 static node *nextarg(node **pn)
1713 if (n && (n->info & OPCLSMASK) == OC_COMMA) {
1722 static void hashwalk_init(var *v, xhash *array)
1728 if (v->type & VF_WALK)
1732 w = v->x.walker = xzalloc(2 + 2*sizeof(char *) + array->glen);
1733 w[0] = w[1] = (char *)(w + 2);
1734 for (i = 0; i < array->csize; i++) {
1735 hi = array->items[i];
1737 strcpy(*w, hi->name);
1744 static int hashwalk_next(var *v)
1752 setvar_s(v, nextword(w+1));
1756 /* evaluate node, return 1 when result is true, 0 otherwise */
1757 static int ptest(node *pattern)
1759 /* ptest__v is "static": to save stack space? */
1760 return istrue(evaluate(pattern, &G.ptest__v));
1763 /* read next record from stream rsm into a variable v */
1764 static int awk_getline(rstream *rsm, var *v)
1767 regmatch_t pmatch[2];
1768 int a, p, pp=0, size;
1769 int fd, so, eo, r, rp;
1772 /* we're using our own buffer since we need access to accumulating
1775 fd = fileno(rsm->F);
1780 c = (char) rsplitter.n.info;
1783 if (!m) qrealloc(&m, 256, &size);
1789 if ((rsplitter.n.info & OPCLSMASK) == OC_REGEXP) {
1790 if (regexec(icase ? rsplitter.n.r.ire : rsplitter.n.l.re,
1791 b, 1, pmatch, 0) == 0) {
1792 so = pmatch[0].rm_so;
1793 eo = pmatch[0].rm_eo;
1797 } else if (c != '\0') {
1798 s = strchr(b+pp, c);
1799 if (!s) s = memchr(b+pp, '\0', p - pp);
1806 while (b[rp] == '\n')
1808 s = strstr(b+rp, "\n\n");
1811 while (b[eo] == '\n') eo++;
1819 memmove(m, (const void *)(m+a), p+1);
1824 qrealloc(&m, a+p+128, &size);
1827 p += safe_read(fd, b+p, size-p-1);
1831 setvar_i(intvar[ERRNO], errno);
1840 c = b[so]; b[so] = '\0';
1844 c = b[eo]; b[eo] = '\0';
1845 setvar_s(intvar[RT], b+so);
1857 static int fmt_num(char *b, int size, const char *format, double n, int int_as_int)
1861 const char *s = format;
1863 if (int_as_int && n == (int)n) {
1864 r = snprintf(b, size, "%d", (int)n);
1866 do { c = *s; } while (c && *++s);
1867 if (strchr("diouxX", c)) {
1868 r = snprintf(b, size, format, (int)n);
1869 } else if (strchr("eEfgG", c)) {
1870 r = snprintf(b, size, format, n);
1872 syntax_error(EMSG_INV_FMT);
1878 /* formatted output into an allocated buffer, return ptr to buffer */
1879 static char *awk_printf(node *n)
1884 int i, j, incr, bsize;
1889 fmt = f = xstrdup(getvar_s(evaluate(nextarg(&n), v)));
1894 while (*f && (*f != '%' || *(++f) == '%'))
1896 while (*f && !isalpha(*f)) {
1898 syntax_error("%*x formats are not supported");
1902 incr = (f - s) + MAXVARFMT;
1903 qrealloc(&b, incr + i, &bsize);
1908 arg = evaluate(nextarg(&n), v);
1911 if (c == 'c' || !c) {
1912 i += sprintf(b+i, s, is_numeric(arg) ?
1913 (char)getvar_i(arg) : *getvar_s(arg));
1914 } else if (c == 's') {
1916 qrealloc(&b, incr+i+strlen(s1), &bsize);
1917 i += sprintf(b+i, s, s1);
1919 i += fmt_num(b+i, incr, s, getvar_i(arg), FALSE);
1923 /* if there was an error while sprintf, return value is negative */
1927 b = xrealloc(b, i + 1);
1934 /* common substitution routine
1935 * replace (nm) substring of (src) that match (n) with (repl), store
1936 * result into (dest), return number of substitutions. If nm=0, replace
1937 * all matches. If src or dst is NULL, use $0. If ex=TRUE, enable
1938 * subexpression matching (\1-\9)
1940 static int awk_sub(node *rn, const char *repl, int nm, var *src, var *dest, int ex)
1945 int c, i, j, di, rl, so, eo, nbs, n, dssize;
1946 regmatch_t pmatch[10];
1949 re = as_regex(rn, &sreg);
1950 if (!src) src = intvar[F0];
1951 if (!dest) dest = intvar[F0];
1956 while (regexec(re, sp, 10, pmatch, sp==getvar_s(src) ? 0 : REG_NOTBOL) == 0) {
1957 so = pmatch[0].rm_so;
1958 eo = pmatch[0].rm_eo;
1960 qrealloc(&ds, di + eo + rl, &dssize);
1961 memcpy(ds + di, sp, eo);
1967 for (s = repl; *s; s++) {
1973 if (c == '&' || (ex && c >= '0' && c <= '9')) {
1974 di -= ((nbs + 3) >> 1);
1983 n = pmatch[j].rm_eo - pmatch[j].rm_so;
1984 qrealloc(&ds, di + rl + n, &dssize);
1985 memcpy(ds + di, sp + pmatch[j].rm_so, n);
2003 qrealloc(&ds, di + strlen(sp), &dssize);
2004 strcpy(ds + di, sp);
2011 static NOINLINE int do_mktime(const char *ds)
2016 /*memset(&then, 0, sizeof(then)); - not needed */
2017 then.tm_isdst = -1; /* default is unknown */
2019 /* manpage of mktime says these fields are ints,
2020 * so we can sscanf stuff directly into them */
2021 count = sscanf(ds, "%u %u %u %u %u %u %d",
2022 &then.tm_year, &then.tm_mon, &then.tm_mday,
2023 &then.tm_hour, &then.tm_min, &then.tm_sec,
2027 || (unsigned)then.tm_mon < 1
2028 || (unsigned)then.tm_year < 1900
2034 then.tm_year -= 1900;
2036 return mktime(&then);
2039 static NOINLINE var *exec_builtin(node *op, var *res)
2041 #define tspl (G.exec_builtin__tspl)
2047 regmatch_t pmatch[2];
2057 isr = info = op->info;
2060 av[2] = av[3] = NULL;
2061 for (i = 0; i < 4 && op; i++) {
2062 an[i] = nextarg(&op);
2063 if (isr & 0x09000000) av[i] = evaluate(an[i], &tv[i]);
2064 if (isr & 0x08000000) as[i] = getvar_s(av[i]);
2069 if ((uint32_t)nargs < (info >> 30))
2070 syntax_error(EMSG_TOO_FEW_ARGS);
2076 #if ENABLE_FEATURE_AWK_LIBM
2077 setvar_i(res, atan2(getvar_i(av[0]), getvar_i(av[1])));
2079 syntax_error(EMSG_NO_MATH);
2085 spl = (an[2]->info & OPCLSMASK) == OC_REGEXP ?
2086 an[2] : mk_splitter(getvar_s(evaluate(an[2], &tv[2])), &tspl);
2091 n = awk_split(as[0], spl, &s);
2093 clear_array(iamarray(av[1]));
2094 for (i = 1; i <= n; i++)
2095 setari_u(av[1], i, nextword(&s1));
2102 i = getvar_i(av[1]) - 1;
2105 n = (nargs > 2) ? getvar_i(av[2]) : l-i;
2107 s = xstrndup(as[0]+i, n);
2111 /* Bitwise ops must assume that operands are unsigned. GNU Awk 3.1.5:
2112 * awk '{ print or(-1,1) }' gives "4.29497e+09", not "-2.xxxe+09" */
2114 setvar_i(res, getvar_i_int(av[0]) & getvar_i_int(av[1]));
2118 setvar_i(res, ~getvar_i_int(av[0]));
2122 setvar_i(res, getvar_i_int(av[0]) << getvar_i_int(av[1]));
2126 setvar_i(res, getvar_i_int(av[0]) | getvar_i_int(av[1]));
2130 setvar_i(res, getvar_i_int(av[0]) >> getvar_i_int(av[1]));
2134 setvar_i(res, getvar_i_int(av[0]) ^ getvar_i_int(av[1]));
2139 s1 = s = xstrdup(as[0]);
2141 //*s1 = (info == B_up) ? toupper(*s1) : tolower(*s1);
2142 if ((unsigned char)((*s1 | 0x20) - 'a') <= ('z' - 'a'))
2143 *s1 = (info == B_up) ? (*s1 & 0xdf) : (*s1 | 0x20);
2152 l = strlen(as[0]) - ll;
2153 if (ll > 0 && l >= 0) {
2155 s = strstr(as[0], as[1]);
2156 if (s) n = (s - as[0]) + 1;
2158 /* this piece of code is terribly slow and
2159 * really should be rewritten
2161 for (i=0; i<=l; i++) {
2162 if (strncasecmp(as[0]+i, as[1], ll) == 0) {
2174 tt = getvar_i(av[1]);
2177 //s = (nargs > 0) ? as[0] : "%a %b %d %H:%M:%S %Z %Y";
2178 i = strftime(g_buf, MAXVARFMT,
2179 ((nargs > 0) ? as[0] : "%a %b %d %H:%M:%S %Z %Y"),
2182 setvar_s(res, g_buf);
2186 setvar_i(res, do_mktime(as[0]));
2190 re = as_regex(an[1], &sreg);
2191 n = regexec(re, as[0], 1, pmatch, 0);
2196 pmatch[0].rm_so = 0;
2197 pmatch[0].rm_eo = -1;
2199 setvar_i(newvar("RSTART"), pmatch[0].rm_so);
2200 setvar_i(newvar("RLENGTH"), pmatch[0].rm_eo - pmatch[0].rm_so);
2201 setvar_i(res, pmatch[0].rm_so);
2202 if (re == &sreg) regfree(re);
2206 awk_sub(an[0], as[1], getvar_i(av[2]), av[3], res, TRUE);
2210 setvar_i(res, awk_sub(an[0], as[1], 0, av[2], av[2], FALSE));
2214 setvar_i(res, awk_sub(an[0], as[1], 1, av[2], av[2], FALSE));
2224 * Evaluate node - the heart of the program. Supplied with subtree
2225 * and place where to store result. returns ptr to result.
2227 #define XC(n) ((n) >> 8)
2229 static var *evaluate(node *op, var *res)
2231 /* This procedure is recursive so we should count every byte */
2232 #define fnargs (G.evaluate__fnargs)
2233 /* seed is initialized to 1 */
2234 #define seed (G.evaluate__seed)
2235 #define sreg (G.evaluate__sreg)
2257 return setvar_s(res, NULL);
2263 opn = (opinfo & OPNMASK);
2264 g_lineno = op->lineno;
2266 /* execute inevitable things */
2268 if (opinfo & OF_RES1) X.v = L.v = evaluate(op1, v1);
2269 if (opinfo & OF_RES2) R.v = evaluate(op->r.n, v1+1);
2270 if (opinfo & OF_STR1) L.s = getvar_s(L.v);
2271 if (opinfo & OF_STR2) R.s = getvar_s(R.v);
2272 if (opinfo & OF_NUM1) L.d = getvar_i(L.v);
2274 switch (XC(opinfo & OPCLSMASK)) {
2276 /* -- iterative node type -- */
2280 if ((op1->info & OPCLSMASK) == OC_COMMA) {
2281 /* it's range pattern */
2282 if ((opinfo & OF_CHECKED) || ptest(op1->l.n)) {
2283 op->info |= OF_CHECKED;
2284 if (ptest(op1->r.n))
2285 op->info &= ~OF_CHECKED;
2292 op = (ptest(op1)) ? op->a.n : op->r.n;
2296 /* just evaluate an expression, also used as unconditional jump */
2300 /* branch, used in if-else and various loops */
2302 op = istrue(L.v) ? op->a.n : op->r.n;
2305 /* initialize for-in loop */
2306 case XC( OC_WALKINIT ):
2307 hashwalk_init(L.v, iamarray(R.v));
2310 /* get next array item */
2311 case XC( OC_WALKNEXT ):
2312 op = hashwalk_next(L.v) ? op->a.n : op->r.n;
2315 case XC( OC_PRINT ):
2316 case XC( OC_PRINTF ):
2319 X.rsm = newfile(R.s);
2322 X.rsm->F = popen(R.s, "w");
2323 if (X.rsm->F == NULL)
2324 bb_perror_msg_and_die("popen");
2327 X.rsm->F = xfopen(R.s, opn=='w' ? "w" : "a");
2333 if ((opinfo & OPCLSMASK) == OC_PRINT) {
2335 fputs(getvar_s(intvar[F0]), X.F);
2338 L.v = evaluate(nextarg(&op1), v1);
2339 if (L.v->type & VF_NUMBER) {
2340 fmt_num(g_buf, MAXVARFMT, getvar_s(intvar[OFMT]),
2341 getvar_i(L.v), TRUE);
2344 fputs(getvar_s(L.v), X.F);
2347 if (op1) fputs(getvar_s(intvar[OFS]), X.F);
2350 fputs(getvar_s(intvar[ORS]), X.F);
2352 } else { /* OC_PRINTF */
2353 L.s = awk_printf(op1);
2360 case XC( OC_DELETE ):
2361 X.info = op1->info & OPCLSMASK;
2362 if (X.info == OC_VAR) {
2364 } else if (X.info == OC_FNARG) {
2365 R.v = &fnargs[op1->l.i];
2367 syntax_error(EMSG_NOT_ARRAY);
2372 L.s = getvar_s(evaluate(op1->r.n, v1));
2373 hash_remove(iamarray(R.v), L.s);
2375 clear_array(iamarray(R.v));
2379 case XC( OC_NEWSOURCE ):
2380 g_progname = op->l.s;
2383 case XC( OC_RETURN ):
2387 case XC( OC_NEXTFILE ):
2398 /* -- recursive node type -- */
2402 if (L.v == intvar[NF])
2406 case XC( OC_FNARG ):
2407 L.v = &fnargs[op->l.i];
2409 res = op->r.n ? findvar(iamarray(L.v), R.s) : L.v;
2413 setvar_i(res, hash_search(iamarray(R.v), L.s) ? 1 : 0);
2416 case XC( OC_REGEXP ):
2418 L.s = getvar_s(intvar[F0]);
2421 case XC( OC_MATCH ):
2424 X.re = as_regex(op1, &sreg);
2425 R.i = regexec(X.re, L.s, 0, NULL, 0);
2426 if (X.re == &sreg) regfree(X.re);
2427 setvar_i(res, (R.i == 0) ^ (opn == '!'));
2431 /* if source is a temporary string, jusk relink it to dest */
2432 //Disabled: if R.v is numeric but happens to have cached R.v->string,
2433 //then L.v ends up being a string, which is wrong
2434 // if (R.v == v1+1 && R.v->string) {
2435 // res = setvar_p(L.v, R.v->string);
2436 // R.v->string = NULL;
2438 res = copyvar(L.v, R.v);
2442 case XC( OC_TERNARY ):
2443 if ((op->r.n->info & OPCLSMASK) != OC_COLON)
2444 syntax_error(EMSG_POSSIBLE_ERROR);
2445 res = evaluate(istrue(L.v) ? op->r.n->l.n : op->r.n->r.n, res);
2449 if (!op->r.f->body.first)
2450 syntax_error(EMSG_UNDEF_FUNC);
2452 X.v = R.v = nvalloc(op->r.f->nargs + 1);
2454 L.v = evaluate(nextarg(&op1), v1);
2456 R.v->type |= VF_CHILD;
2457 R.v->x.parent = L.v;
2458 if (++R.v - X.v >= op->r.f->nargs)
2466 res = evaluate(op->r.f->body.first, res);
2473 case XC( OC_GETLINE ):
2474 case XC( OC_PGETLINE ):
2476 X.rsm = newfile(L.s);
2478 if ((opinfo & OPCLSMASK) == OC_PGETLINE) {
2479 X.rsm->F = popen(L.s, "r");
2480 X.rsm->is_pipe = TRUE;
2482 X.rsm->F = fopen_for_read(L.s); /* not xfopen! */
2486 if (!iF) iF = next_input_file();
2491 setvar_i(intvar[ERRNO], errno);
2499 L.i = awk_getline(X.rsm, R.v);
2502 incvar(intvar[FNR]);
2509 /* simple builtins */
2510 case XC( OC_FBLTIN ):
2518 R.d = (double)rand() / (double)RAND_MAX;
2520 #if ENABLE_FEATURE_AWK_LIBM
2546 syntax_error(EMSG_NO_MATH);
2551 seed = op1 ? (unsigned)L.d : (unsigned)time(NULL);
2561 L.s = getvar_s(intvar[F0]);
2567 R.d = (ENABLE_FEATURE_ALLOW_EXEC && L.s && *L.s)
2568 ? (system(L.s) >> 8) : 0;
2576 X.rsm = newfile(L.s);
2585 X.rsm = (rstream *)hash_search(fdhash, L.s);
2587 R.i = X.rsm->is_pipe ? pclose(X.rsm->F) : fclose(X.rsm->F);
2588 free(X.rsm->buffer);
2589 hash_remove(fdhash, L.s);
2592 setvar_i(intvar[ERRNO], errno);
2599 case XC( OC_BUILTIN ):
2600 res = exec_builtin(op, res);
2603 case XC( OC_SPRINTF ):
2604 setvar_p(res, awk_printf(op1));
2607 case XC( OC_UNARY ):
2609 L.d = R.d = getvar_i(R.v);
2635 case XC( OC_FIELD ):
2636 R.i = (int)getvar_i(R.v);
2643 res = &Fields[R.i - 1];
2647 /* concatenation (" ") and index joining (",") */
2648 case XC( OC_CONCAT ):
2649 case XC( OC_COMMA ):
2650 opn = strlen(L.s) + strlen(R.s) + 2;
2653 if ((opinfo & OPCLSMASK) == OC_COMMA) {
2654 L.s = getvar_s(intvar[SUBSEP]);
2655 X.s = xrealloc(X.s, opn + strlen(L.s));
2663 setvar_i(res, istrue(L.v) ? ptest(op->r.n) : 0);
2667 setvar_i(res, istrue(L.v) ? 1 : ptest(op->r.n));
2670 case XC( OC_BINARY ):
2671 case XC( OC_REPLACE ):
2672 R.d = getvar_i(R.v);
2685 syntax_error(EMSG_DIV_BY_ZERO);
2689 #if ENABLE_FEATURE_AWK_LIBM
2690 L.d = pow(L.d, R.d);
2692 syntax_error(EMSG_NO_MATH);
2697 syntax_error(EMSG_DIV_BY_ZERO);
2698 L.d -= (int)(L.d / R.d) * R.d;
2701 res = setvar_i(((opinfo & OPCLSMASK) == OC_BINARY) ? res : X.v, L.d);
2704 case XC( OC_COMPARE ):
2705 if (is_numeric(L.v) && is_numeric(R.v)) {
2706 L.d = getvar_i(L.v) - getvar_i(R.v);
2708 L.s = getvar_s(L.v);
2709 R.s = getvar_s(R.v);
2710 L.d = icase ? strcasecmp(L.s, R.s) : strcmp(L.s, R.s);
2712 switch (opn & 0xfe) {
2723 setvar_i(res, (opn & 1 ? R.i : !R.i) ? 1 : 0);
2727 syntax_error(EMSG_POSSIBLE_ERROR);
2729 if ((opinfo & OPCLSMASK) <= SHIFT_TIL_THIS)
2731 if ((opinfo & OPCLSMASK) >= RECUR_FROM_THIS)
2744 /* -------- main & co. -------- */
2746 static int awk_exit(int r)
2757 evaluate(endseq.first, &tv);
2760 /* waiting for children */
2761 for (i = 0; i < fdhash->csize; i++) {
2762 hi = fdhash->items[i];
2764 if (hi->data.rs.F && hi->data.rs.is_pipe)
2765 pclose(hi->data.rs.F);
2773 /* if expr looks like "var=value", perform assignment and return 1,
2774 * otherwise return 0 */
2775 static int is_assignment(const char *expr)
2777 char *exprc, *s, *s0, *s1;
2779 exprc = xstrdup(expr);
2780 if (!isalnum_(*exprc) || (s = strchr(exprc, '=')) == NULL) {
2788 *(s1++) = nextchar(&s);
2791 setvar_u(newvar(exprc), s0);
2796 /* switch to next input file */
2797 static rstream *next_input_file(void)
2799 #define rsm (G.next_input_file__rsm)
2800 #define files_happen (G.next_input_file__files_happen)
2803 const char *fname, *ind;
2808 rsm.pos = rsm.adv = 0;
2811 if (getvar_i(intvar[ARGIND])+1 >= getvar_i(intvar[ARGC])) {
2817 ind = getvar_s(incvar(intvar[ARGIND]));
2818 fname = getvar_s(findvar(iamarray(intvar[ARGV]), ind));
2819 if (fname && *fname && !is_assignment(fname))
2820 F = xfopen_stdin(fname);
2824 files_happen = TRUE;
2825 setvar_s(intvar[FILENAME], fname);
2832 int awk_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
2833 int awk_main(int argc, char **argv)
2836 char *opt_F, *opt_W;
2837 llist_t *list_v = NULL;
2838 llist_t *list_f = NULL;
2843 char *vnames = (char *)vNames; /* cheat */
2844 char *vvalues = (char *)vValues;
2848 /* Undo busybox.c, or else strtod may eat ','! This breaks parsing:
2849 * $1,$2 == '$1,' '$2', NOT '$1' ',' '$2' */
2850 if (ENABLE_LOCALE_SUPPORT)
2851 setlocale(LC_NUMERIC, "C");
2855 /* allocate global buffer */
2856 g_buf = xmalloc(MAXVARFMT + 1);
2858 vhash = hash_init();
2859 ahash = hash_init();
2860 fdhash = hash_init();
2861 fnhash = hash_init();
2863 /* initialize variables */
2864 for (i = 0; *vnames; i++) {
2865 intvar[i] = v = newvar(nextword(&vnames));
2866 if (*vvalues != '\377')
2867 setvar_s(v, nextword(&vvalues));
2871 if (*vnames == '*') {
2872 v->type |= VF_SPECIAL;
2877 handle_special(intvar[FS]);
2878 handle_special(intvar[RS]);
2880 newfile("/dev/stdin")->F = stdin;
2881 newfile("/dev/stdout")->F = stdout;
2882 newfile("/dev/stderr")->F = stderr;
2884 /* Huh, people report that sometimes environ is NULL. Oh well. */
2885 if (environ) for (envp = environ; *envp; envp++) {
2886 /* environ is writable, thus we don't strdup it needlessly */
2888 char *s1 = strchr(s, '=');
2891 /* Both findvar and setvar_u take const char*
2892 * as 2nd arg -> environment is not trashed */
2893 setvar_u(findvar(iamarray(intvar[ENVIRON]), s), s1 + 1);
2897 opt_complementary = "v::f::"; /* -v and -f can occur multiple times */
2898 opt = getopt32(argv, "F:v:f:W:", &opt_F, &list_v, &list_f, &opt_W);
2902 setvar_s(intvar[FS], opt_F); // -F
2903 while (list_v) { /* -v */
2904 if (!is_assignment(llist_pop(&list_v)))
2907 if (list_f) { /* -f */
2912 g_progname = llist_pop(&list_f);
2913 from_file = xfopen_stdin(g_progname);
2914 /* one byte is reserved for some trick in next_token */
2915 for (i = j = 1; j > 0; i += j) {
2916 s = xrealloc(s, i + 4096);
2917 j = fread(s + i, 1, 4094, from_file);
2921 parse_program(s + 1);
2925 } else { // no -f: take program from 1st parameter
2928 g_progname = "cmd. line";
2929 parse_program(*argv++);
2931 if (opt & 0x8) // -W
2932 bb_error_msg("warning: unrecognized option '-W %s' ignored", opt_W);
2934 /* fill in ARGV array */
2935 setvar_i(intvar[ARGC], argc);
2936 setari_u(intvar[ARGV], 0, "awk");
2939 setari_u(intvar[ARGV], ++i, *argv++);
2941 evaluate(beginseq.first, &tv);
2942 if (!mainseq.first && !endseq.first)
2943 awk_exit(EXIT_SUCCESS);
2945 /* input file could already be opened in BEGIN block */
2947 iF = next_input_file();
2949 /* passing through input files */
2952 setvar_i(intvar[FNR], 0);
2954 while ((i = awk_getline(iF, intvar[F0])) > 0) {
2957 incvar(intvar[FNR]);
2958 evaluate(mainseq.first, &tv);
2965 syntax_error(strerror(errno));
2967 iF = next_input_file();
2970 awk_exit(EXIT_SUCCESS);