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 GPLv2 or later, see file LICENSE in this source tree.
10 //config: bool "awk (22 kb)"
13 //config: Awk is used as a pattern scanning and processing language.
15 //config:config FEATURE_AWK_LIBM
16 //config: bool "Enable math functions (requires libm)"
18 //config: depends on AWK
20 //config: Enable math functions of the Awk programming language.
21 //config: NOTE: This requires libm to be present for linking.
23 //config:config FEATURE_AWK_GNU_EXTENSIONS
24 //config: bool "Enable a few GNU extensions"
26 //config: depends on AWK
28 //config: Enable a few features from gawk:
29 //config: * command line option -e AWK_PROGRAM
30 //config: * simultaneous use of -f and -e on the command line.
31 //config: This enables the use of awk library files.
32 //config: Example: awk -f mylib.awk -e '{print myfunction($1);}' ...
34 //applet:IF_AWK(APPLET_NOEXEC(awk, awk, BB_DIR_USR_BIN, BB_SUID_DROP, awk))
36 //kbuild:lib-$(CONFIG_AWK) += awk.o
38 //usage:#define awk_trivial_usage
39 //usage: "[OPTIONS] [AWK_PROGRAM] [FILE]..."
40 //usage:#define awk_full_usage "\n\n"
41 //usage: " -v VAR=VAL Set variable"
42 //usage: "\n -F SEP Use SEP as field separator"
43 //usage: "\n -f FILE Read program from FILE"
44 //usage: IF_FEATURE_AWK_GNU_EXTENSIONS(
45 //usage: "\n -e AWK_PROGRAM"
52 /* This is a NOEXEC applet. Be very careful! */
55 /* If you comment out one of these below, it will be #defined later
56 * to perform debug printfs to stderr: */
57 #define debug_printf_walker(...) do {} while (0)
58 #define debug_printf_eval(...) do {} while (0)
59 #define debug_printf_parse(...) do {} while (0)
61 #ifndef debug_printf_walker
62 # define debug_printf_walker(...) (fprintf(stderr, __VA_ARGS__))
64 #ifndef debug_printf_eval
65 # define debug_printf_eval(...) (fprintf(stderr, __VA_ARGS__))
67 #ifndef debug_printf_parse
68 # define debug_printf_parse(...) (fprintf(stderr, __VA_ARGS__))
72 /* "+": stop on first non-option:
73 * $ awk 'BEGIN { for(i=1; i<ARGC; ++i) { print i ": " ARGV[i] }}' -argz
76 #define OPTSTR_AWK "+" \
78 IF_FEATURE_AWK_GNU_EXTENSIONS("e:*") \
81 OPTBIT_F, /* define field separator */
82 OPTBIT_v, /* define variable */
83 OPTBIT_f, /* pull in awk program from file */
84 IF_FEATURE_AWK_GNU_EXTENSIONS(OPTBIT_e,) /* -e AWK_PROGRAM */
85 OPTBIT_W, /* -W ignored */
86 OPT_F = 1 << OPTBIT_F,
87 OPT_v = 1 << OPTBIT_v,
88 OPT_f = 1 << OPTBIT_f,
89 OPT_e = IF_FEATURE_AWK_GNU_EXTENSIONS((1 << OPTBIT_e)) + 0,
97 #define VF_NUMBER 0x0001 /* 1 = primary type is number */
98 #define VF_ARRAY 0x0002 /* 1 = it's an array */
100 #define VF_CACHED 0x0100 /* 1 = num/str value has cached str/num eq */
101 #define VF_USER 0x0200 /* 1 = user input (may be numeric string) */
102 #define VF_SPECIAL 0x0400 /* 1 = requires extra handling when changed */
103 #define VF_WALK 0x0800 /* 1 = variable has alloc'd x.walker list */
104 #define VF_FSTR 0x1000 /* 1 = var::string points to fstring buffer */
105 #define VF_CHILD 0x2000 /* 1 = function arg; x.parent points to source */
106 #define VF_DIRTY 0x4000 /* 1 = variable was set explicitly */
108 /* these flags are static, don't change them when value is changed */
109 #define VF_DONTTOUCH (VF_ARRAY | VF_SPECIAL | VF_WALK | VF_CHILD | VF_DIRTY)
111 typedef struct walker_list {
114 struct walker_list *prev;
119 typedef struct var_s {
120 unsigned type; /* flags */
124 int aidx; /* func arg idx (for compilation stage) */
125 struct xhash_s *array; /* array ptr */
126 struct var_s *parent; /* for func args, ptr to actual parameter */
127 walker_list *walker; /* list of array elements (for..in) */
131 /* Node chain (pattern-action chain, BEGIN, END, function bodies) */
132 typedef struct chain_s {
133 struct node_s *first;
135 const char *programname;
139 typedef struct func_s {
145 typedef struct rstream_s {
154 typedef struct hash_item_s {
156 struct var_s v; /* variable/array hash */
157 struct rstream_s rs; /* redirect streams hash */
158 struct func_s f; /* functions hash */
160 struct hash_item_s *next; /* next in chain */
161 char name[1]; /* really it's longer */
164 typedef struct xhash_s {
165 unsigned nel; /* num of elements */
166 unsigned csize; /* current hash size */
167 unsigned nprime; /* next hash size in PRIMES[] */
168 unsigned glen; /* summary length of item names */
169 struct hash_item_s **items;
173 typedef struct node_s {
193 /* Block of temporary variables */
194 typedef struct nvblock_s {
197 struct nvblock_s *prev;
198 struct nvblock_s *next;
202 typedef struct tsplitter_s {
207 /* simple token classes */
208 /* Order and hex values are very important!!! See next_token() */
209 #define TC_SEQSTART (1 << 0) /* ( */
210 #define TC_SEQTERM (1 << 1) /* ) */
211 #define TC_REGEXP (1 << 2) /* /.../ */
212 #define TC_OUTRDR (1 << 3) /* | > >> */
213 #define TC_UOPPOST (1 << 4) /* unary postfix operator */
214 #define TC_UOPPRE1 (1 << 5) /* unary prefix operator */
215 #define TC_BINOPX (1 << 6) /* two-opnd operator */
216 #define TC_IN (1 << 7)
217 #define TC_COMMA (1 << 8)
218 #define TC_PIPE (1 << 9) /* input redirection pipe */
219 #define TC_UOPPRE2 (1 << 10) /* unary prefix operator */
220 #define TC_ARRTERM (1 << 11) /* ] */
221 #define TC_GRPSTART (1 << 12) /* { */
222 #define TC_GRPTERM (1 << 13) /* } */
223 #define TC_SEMICOL (1 << 14)
224 #define TC_NEWLINE (1 << 15)
225 #define TC_STATX (1 << 16) /* ctl statement (for, next...) */
226 #define TC_WHILE (1 << 17)
227 #define TC_ELSE (1 << 18)
228 #define TC_BUILTIN (1 << 19)
229 /* This costs ~50 bytes of code.
230 * A separate class to support deprecated "length" form. If we don't need that
231 * (i.e. if we demand that only "length()" with () is valid), then TC_LENGTH
232 * can be merged with TC_BUILTIN:
234 #define TC_LENGTH (1 << 20)
235 #define TC_GETLINE (1 << 21)
236 #define TC_FUNCDECL (1 << 22) /* 'function' 'func' */
237 #define TC_BEGIN (1 << 23)
238 #define TC_END (1 << 24)
239 #define TC_EOF (1 << 25)
240 #define TC_VARIABLE (1 << 26)
241 #define TC_ARRAY (1 << 27)
242 #define TC_FUNCTION (1 << 28)
243 #define TC_STRING (1 << 29)
244 #define TC_NUMBER (1 << 30)
246 #define TC_UOPPRE (TC_UOPPRE1 | TC_UOPPRE2)
248 /* combined token classes */
249 #define TC_BINOP (TC_BINOPX | TC_COMMA | TC_PIPE | TC_IN)
250 //#define TC_UNARYOP (TC_UOPPRE | TC_UOPPOST)
251 #define TC_OPERAND (TC_VARIABLE | TC_ARRAY | TC_FUNCTION \
252 | TC_BUILTIN | TC_LENGTH | TC_GETLINE \
253 | TC_SEQSTART | TC_STRING | TC_NUMBER)
255 #define TC_STATEMNT (TC_STATX | TC_WHILE)
256 #define TC_OPTERM (TC_SEMICOL | TC_NEWLINE)
258 /* word tokens, cannot mean something else if not expected */
259 #define TC_WORD (TC_IN | TC_STATEMNT | TC_ELSE \
260 | TC_BUILTIN | TC_LENGTH | TC_GETLINE \
261 | TC_FUNCDECL | TC_BEGIN | TC_END)
263 /* discard newlines after these */
264 #define TC_NOTERM (TC_COMMA | TC_GRPSTART | TC_GRPTERM \
265 | TC_BINOP | TC_OPTERM)
267 /* what can expression begin with */
268 #define TC_OPSEQ (TC_OPERAND | TC_UOPPRE | TC_REGEXP)
269 /* what can group begin with */
270 #define TC_GRPSEQ (TC_OPSEQ | TC_OPTERM | TC_STATEMNT | TC_GRPSTART)
272 /* if previous token class is CONCAT1 and next is CONCAT2, concatenation */
273 /* operator is inserted between them */
274 #define TC_CONCAT1 (TC_VARIABLE | TC_ARRTERM | TC_SEQTERM \
275 | TC_STRING | TC_NUMBER | TC_UOPPOST)
276 #define TC_CONCAT2 (TC_OPERAND | TC_UOPPRE)
278 #define OF_RES1 0x010000
279 #define OF_RES2 0x020000
280 #define OF_STR1 0x040000
281 #define OF_STR2 0x080000
282 #define OF_NUM1 0x100000
283 #define OF_CHECKED 0x200000
285 /* combined operator flags */
288 #define xS (OF_RES2 | OF_STR2)
290 #define VV (OF_RES1 | OF_RES2)
291 #define Nx (OF_RES1 | OF_NUM1)
292 #define NV (OF_RES1 | OF_NUM1 | OF_RES2)
293 #define Sx (OF_RES1 | OF_STR1)
294 #define SV (OF_RES1 | OF_STR1 | OF_RES2)
295 #define SS (OF_RES1 | OF_STR1 | OF_RES2 | OF_STR2)
297 #define OPCLSMASK 0xFF00
298 #define OPNMASK 0x007F
300 /* operator priority is a highest byte (even: r->l, odd: l->r grouping)
301 * For builtins it has different meaning: n n s3 s2 s1 v3 v2 v1,
302 * n - min. number of args, vN - resolve Nth arg to var, sN - resolve to string
307 #define P(x) (x << 24)
308 #define PRIMASK 0x7F000000
309 #define PRIMASK2 0x7E000000
311 /* Operation classes */
313 #define SHIFT_TIL_THIS 0x0600
314 #define RECUR_FROM_THIS 0x1000
317 OC_DELETE = 0x0100, OC_EXEC = 0x0200, OC_NEWSOURCE = 0x0300,
318 OC_PRINT = 0x0400, OC_PRINTF = 0x0500, OC_WALKINIT = 0x0600,
320 OC_BR = 0x0700, OC_BREAK = 0x0800, OC_CONTINUE = 0x0900,
321 OC_EXIT = 0x0a00, OC_NEXT = 0x0b00, OC_NEXTFILE = 0x0c00,
322 OC_TEST = 0x0d00, OC_WALKNEXT = 0x0e00,
324 OC_BINARY = 0x1000, OC_BUILTIN = 0x1100, OC_COLON = 0x1200,
325 OC_COMMA = 0x1300, OC_COMPARE = 0x1400, OC_CONCAT = 0x1500,
326 OC_FBLTIN = 0x1600, OC_FIELD = 0x1700, OC_FNARG = 0x1800,
327 OC_FUNC = 0x1900, OC_GETLINE = 0x1a00, OC_IN = 0x1b00,
328 OC_LAND = 0x1c00, OC_LOR = 0x1d00, OC_MATCH = 0x1e00,
329 OC_MOVE = 0x1f00, OC_PGETLINE = 0x2000, OC_REGEXP = 0x2100,
330 OC_REPLACE = 0x2200, OC_RETURN = 0x2300, OC_SPRINTF = 0x2400,
331 OC_TERNARY = 0x2500, OC_UNARY = 0x2600, OC_VAR = 0x2700,
334 ST_IF = 0x3000, ST_DO = 0x3100, ST_FOR = 0x3200,
338 /* simple builtins */
340 F_in, F_rn, F_co, F_ex, F_lg, F_si, F_sq, F_sr,
341 F_ti, F_le, F_sy, F_ff, F_cl
346 B_a2, B_ix, B_ma, B_sp, B_ss, B_ti, B_mt, B_lo, B_up,
348 B_an, B_co, B_ls, B_or, B_rs, B_xo,
351 /* tokens and their corresponding info values */
353 #define NTC "\377" /* switch to next token class (tc<<1) */
356 static const char tokenlist[] ALIGN1 =
357 "\1(" NTC /* TC_SEQSTART */
358 "\1)" NTC /* TC_SEQTERM */
359 "\1/" NTC /* TC_REGEXP */
360 "\2>>" "\1>" "\1|" NTC /* TC_OUTRDR */
361 "\2++" "\2--" NTC /* TC_UOPPOST */
362 "\2++" "\2--" "\1$" NTC /* TC_UOPPRE1 */
363 "\2==" "\1=" "\2+=" "\2-=" /* TC_BINOPX */
364 "\2*=" "\2/=" "\2%=" "\2^="
365 "\1+" "\1-" "\3**=" "\2**"
366 "\1/" "\1%" "\1^" "\1*"
367 "\2!=" "\2>=" "\2<=" "\1>"
368 "\1<" "\2!~" "\1~" "\2&&"
369 "\2||" "\1?" "\1:" NTC
370 "\2in" NTC /* TC_IN */
371 "\1," NTC /* TC_COMMA */
372 "\1|" NTC /* TC_PIPE */
373 "\1+" "\1-" "\1!" NTC /* TC_UOPPRE2 */
374 "\1]" NTC /* TC_ARRTERM */
375 "\1{" NTC /* TC_GRPSTART */
376 "\1}" NTC /* TC_GRPTERM */
377 "\1;" NTC /* TC_SEMICOL */
378 "\1\n" NTC /* TC_NEWLINE */
379 "\2if" "\2do" "\3for" "\5break" /* TC_STATX */
380 "\10continue" "\6delete" "\5print"
381 "\6printf" "\4next" "\10nextfile"
382 "\6return" "\4exit" NTC
383 "\5while" NTC /* TC_WHILE */
384 "\4else" NTC /* TC_ELSE */
385 "\3and" "\5compl" "\6lshift" "\2or" /* TC_BUILTIN */
387 "\5close" "\6system" "\6fflush" "\5atan2"
388 "\3cos" "\3exp" "\3int" "\3log"
389 "\4rand" "\3sin" "\4sqrt" "\5srand"
390 "\6gensub" "\4gsub" "\5index" /* "\6length" was here */
391 "\5match" "\5split" "\7sprintf" "\3sub"
392 "\6substr" "\7systime" "\10strftime" "\6mktime"
393 "\7tolower" "\7toupper" NTC
394 "\6length" NTC /* TC_LENGTH */
395 "\7getline" NTC /* TC_GETLINE */
396 "\4func" "\10function" NTC /* TC_FUNCDECL */
397 "\5BEGIN" NTC /* TC_BEGIN */
399 /* compiler adds trailing "\0" */
402 #define OC_B OC_BUILTIN
404 static const uint32_t tokeninfo[] = {
408 xS|'a', xS|'w', xS|'|',
409 OC_UNARY|xV|P(9)|'p', OC_UNARY|xV|P(9)|'m',
410 OC_UNARY|xV|P(9)|'P', OC_UNARY|xV|P(9)|'M', OC_FIELD|xV|P(5),
411 OC_COMPARE|VV|P(39)|5, OC_MOVE|VV|P(74), OC_REPLACE|NV|P(74)|'+', OC_REPLACE|NV|P(74)|'-',
412 OC_REPLACE|NV|P(74)|'*', OC_REPLACE|NV|P(74)|'/', OC_REPLACE|NV|P(74)|'%', OC_REPLACE|NV|P(74)|'&',
413 OC_BINARY|NV|P(29)|'+', OC_BINARY|NV|P(29)|'-', OC_REPLACE|NV|P(74)|'&', OC_BINARY|NV|P(15)|'&',
414 OC_BINARY|NV|P(25)|'/', OC_BINARY|NV|P(25)|'%', OC_BINARY|NV|P(15)|'&', OC_BINARY|NV|P(25)|'*',
415 OC_COMPARE|VV|P(39)|4, OC_COMPARE|VV|P(39)|3, OC_COMPARE|VV|P(39)|0, OC_COMPARE|VV|P(39)|1,
416 OC_COMPARE|VV|P(39)|2, OC_MATCH|Sx|P(45)|'!', OC_MATCH|Sx|P(45)|'~', OC_LAND|Vx|P(55),
417 OC_LOR|Vx|P(59), OC_TERNARY|Vx|P(64)|'?', OC_COLON|xx|P(67)|':',
418 OC_IN|SV|P(49), /* TC_IN */
420 OC_PGETLINE|SV|P(37),
421 OC_UNARY|xV|P(19)|'+', OC_UNARY|xV|P(19)|'-', OC_UNARY|xV|P(19)|'!',
427 ST_IF, ST_DO, ST_FOR, OC_BREAK,
428 OC_CONTINUE, OC_DELETE|Vx, OC_PRINT,
429 OC_PRINTF, OC_NEXT, OC_NEXTFILE,
430 OC_RETURN|Vx, OC_EXIT|Nx,
433 OC_B|B_an|P(0x83), OC_B|B_co|P(0x41), OC_B|B_ls|P(0x83), OC_B|B_or|P(0x83),
434 OC_B|B_rs|P(0x83), OC_B|B_xo|P(0x83),
435 OC_FBLTIN|Sx|F_cl, OC_FBLTIN|Sx|F_sy, OC_FBLTIN|Sx|F_ff, OC_B|B_a2|P(0x83),
436 OC_FBLTIN|Nx|F_co, OC_FBLTIN|Nx|F_ex, OC_FBLTIN|Nx|F_in, OC_FBLTIN|Nx|F_lg,
437 OC_FBLTIN|F_rn, OC_FBLTIN|Nx|F_si, OC_FBLTIN|Nx|F_sq, OC_FBLTIN|Nx|F_sr,
438 OC_B|B_ge|P(0xd6), OC_B|B_gs|P(0xb6), OC_B|B_ix|P(0x9b), /* OC_FBLTIN|Sx|F_le, was here */
439 OC_B|B_ma|P(0x89), OC_B|B_sp|P(0x8b), OC_SPRINTF, OC_B|B_su|P(0xb6),
440 OC_B|B_ss|P(0x8f), OC_FBLTIN|F_ti, OC_B|B_ti|P(0x0b), OC_B|B_mt|P(0x0b),
441 OC_B|B_lo|P(0x49), OC_B|B_up|P(0x49),
442 OC_FBLTIN|Sx|F_le, /* TC_LENGTH */
449 /* internal variable names and their initial values */
450 /* asterisk marks SPECIAL vars; $ is just no-named Field0 */
452 CONVFMT, OFMT, FS, OFS,
453 ORS, RS, RT, FILENAME,
454 SUBSEP, F0, ARGIND, ARGC,
455 ARGV, ERRNO, FNR, NR,
456 NF, IGNORECASE, ENVIRON, NUM_INTERNAL_VARS
459 static const char vNames[] ALIGN1 =
460 "CONVFMT\0" "OFMT\0" "FS\0*" "OFS\0"
461 "ORS\0" "RS\0*" "RT\0" "FILENAME\0"
462 "SUBSEP\0" "$\0*" "ARGIND\0" "ARGC\0"
463 "ARGV\0" "ERRNO\0" "FNR\0" "NR\0"
464 "NF\0*" "IGNORECASE\0*" "ENVIRON\0" "\0";
466 static const char vValues[] ALIGN1 =
467 "%.6g\0" "%.6g\0" " \0" " \0"
468 "\n\0" "\n\0" "\0" "\0"
469 "\034\0" "\0" "\377";
471 /* hash size may grow to these values */
472 #define FIRST_PRIME 61
473 static const uint16_t PRIMES[] ALIGN2 = { 251, 1021, 4093, 16381, 65521 };
476 /* Globals. Split in two parts so that first one is addressed
477 * with (mostly short) negative offsets.
478 * NB: it's unsafe to put members of type "double"
479 * into globals2 (gcc may fail to align them).
483 chain beginseq, mainseq, endseq;
485 node *break_ptr, *continue_ptr;
487 xhash *vhash, *ahash, *fdhash, *fnhash;
488 const char *g_progname;
491 int maxfields; /* used in fsrealloc() only */
500 smallint is_f0_split;
504 uint32_t t_info; /* often used */
509 var *intvar[NUM_INTERNAL_VARS]; /* often used */
511 /* former statics from various functions */
512 char *split_f0__fstrings;
514 uint32_t next_token__save_tclass;
515 uint32_t next_token__save_info;
516 uint32_t next_token__ltclass;
517 smallint next_token__concat_inserted;
519 smallint next_input_file__files_happen;
520 rstream next_input_file__rsm;
522 var *evaluate__fnargs;
523 unsigned evaluate__seed;
524 regex_t evaluate__sreg;
528 tsplitter exec_builtin__tspl;
530 /* biggest and least used members go last */
531 tsplitter fsplitter, rsplitter;
533 #define G1 (ptr_to_globals[-1])
534 #define G (*(struct globals2 *)ptr_to_globals)
535 /* For debug. nm --size-sort awk.o | grep -vi ' [tr] ' */
536 /*char G1size[sizeof(G1)]; - 0x74 */
537 /*char Gsize[sizeof(G)]; - 0x1c4 */
538 /* Trying to keep most of members accessible with short offsets: */
539 /*char Gofs_seed[offsetof(struct globals2, evaluate__seed)]; - 0x90 */
540 #define t_double (G1.t_double )
541 #define beginseq (G1.beginseq )
542 #define mainseq (G1.mainseq )
543 #define endseq (G1.endseq )
544 #define seq (G1.seq )
545 #define break_ptr (G1.break_ptr )
546 #define continue_ptr (G1.continue_ptr)
548 #define vhash (G1.vhash )
549 #define ahash (G1.ahash )
550 #define fdhash (G1.fdhash )
551 #define fnhash (G1.fnhash )
552 #define g_progname (G1.g_progname )
553 #define g_lineno (G1.g_lineno )
554 #define nfields (G1.nfields )
555 #define maxfields (G1.maxfields )
556 #define Fields (G1.Fields )
557 #define g_cb (G1.g_cb )
558 #define g_pos (G1.g_pos )
559 #define g_buf (G1.g_buf )
560 #define icase (G1.icase )
561 #define exiting (G1.exiting )
562 #define nextrec (G1.nextrec )
563 #define nextfile (G1.nextfile )
564 #define is_f0_split (G1.is_f0_split )
565 #define t_rollback (G1.t_rollback )
566 #define t_info (G.t_info )
567 #define t_tclass (G.t_tclass )
568 #define t_string (G.t_string )
569 #define t_lineno (G.t_lineno )
570 #define intvar (G.intvar )
571 #define fsplitter (G.fsplitter )
572 #define rsplitter (G.rsplitter )
573 #define INIT_G() do { \
574 SET_PTR_TO_GLOBALS((char*)xzalloc(sizeof(G1)+sizeof(G)) + sizeof(G1)); \
575 G.next_token__ltclass = TC_OPTERM; \
576 G.evaluate__seed = 1; \
580 /* function prototypes */
581 static void handle_special(var *);
582 static node *parse_expr(uint32_t);
583 static void chain_group(void);
584 static var *evaluate(node *, var *);
585 static rstream *next_input_file(void);
586 static int fmt_num(char *, int, const char *, double, int);
587 static int awk_exit(int) NORETURN;
589 /* ---- error handling ---- */
591 static const char EMSG_INTERNAL_ERROR[] ALIGN1 = "Internal error";
592 static const char EMSG_UNEXP_EOS[] ALIGN1 = "Unexpected end of string";
593 static const char EMSG_UNEXP_TOKEN[] ALIGN1 = "Unexpected token";
594 static const char EMSG_DIV_BY_ZERO[] ALIGN1 = "Division by zero";
595 static const char EMSG_INV_FMT[] ALIGN1 = "Invalid format specifier";
596 static const char EMSG_TOO_FEW_ARGS[] ALIGN1 = "Too few arguments for builtin";
597 static const char EMSG_NOT_ARRAY[] ALIGN1 = "Not an array";
598 static const char EMSG_POSSIBLE_ERROR[] ALIGN1 = "Possible syntax error";
599 static const char EMSG_UNDEF_FUNC[] ALIGN1 = "Call to undefined function";
600 static const char EMSG_NO_MATH[] ALIGN1 = "Math support is not compiled in";
601 static const char EMSG_NEGATIVE_FIELD[] ALIGN1 = "Access to negative field";
603 static void zero_out_var(var *vp)
605 memset(vp, 0, sizeof(*vp));
608 static void syntax_error(const char *message) NORETURN;
609 static void syntax_error(const char *message)
611 bb_error_msg_and_die("%s:%i: %s", g_progname, g_lineno, message);
614 /* ---- hash stuff ---- */
616 static unsigned hashidx(const char *name)
621 idx = *name++ + (idx << 6) - idx;
625 /* create new hash */
626 static xhash *hash_init(void)
630 newhash = xzalloc(sizeof(*newhash));
631 newhash->csize = FIRST_PRIME;
632 newhash->items = xzalloc(FIRST_PRIME * sizeof(newhash->items[0]));
637 /* find item in hash, return ptr to data, NULL if not found */
638 static void *hash_search(xhash *hash, const char *name)
642 hi = hash->items[hashidx(name) % hash->csize];
644 if (strcmp(hi->name, name) == 0)
651 /* grow hash if it becomes too big */
652 static void hash_rebuild(xhash *hash)
654 unsigned newsize, i, idx;
655 hash_item **newitems, *hi, *thi;
657 if (hash->nprime == ARRAY_SIZE(PRIMES))
660 newsize = PRIMES[hash->nprime++];
661 newitems = xzalloc(newsize * sizeof(newitems[0]));
663 for (i = 0; i < hash->csize; i++) {
668 idx = hashidx(thi->name) % newsize;
669 thi->next = newitems[idx];
675 hash->csize = newsize;
676 hash->items = newitems;
679 /* find item in hash, add it if necessary. Return ptr to data */
680 static void *hash_find(xhash *hash, const char *name)
686 hi = hash_search(hash, name);
688 if (++hash->nel / hash->csize > 10)
691 l = strlen(name) + 1;
692 hi = xzalloc(sizeof(*hi) + l);
693 strcpy(hi->name, name);
695 idx = hashidx(name) % hash->csize;
696 hi->next = hash->items[idx];
697 hash->items[idx] = hi;
703 #define findvar(hash, name) ((var*) hash_find((hash), (name)))
704 #define newvar(name) ((var*) hash_find(vhash, (name)))
705 #define newfile(name) ((rstream*)hash_find(fdhash, (name)))
706 #define newfunc(name) ((func*) hash_find(fnhash, (name)))
708 static void hash_remove(xhash *hash, const char *name)
710 hash_item *hi, **phi;
712 phi = &hash->items[hashidx(name) % hash->csize];
715 if (strcmp(hi->name, name) == 0) {
716 hash->glen -= (strlen(name) + 1);
726 /* ------ some useful functions ------ */
728 static char *skip_spaces(char *p)
731 if (*p == '\\' && p[1] == '\n') {
734 } else if (*p != ' ' && *p != '\t') {
742 /* returns old *s, advances *s past word and terminating NUL */
743 static char *nextword(char **s)
746 while (*(*s)++ != '\0')
751 static char nextchar(char **s)
758 c = bb_process_escape_sequence((const char**)s);
759 /* Example awk statement:
761 * we must treat \" as "
763 if (c == '\\' && *s == pps) { /* unrecognized \z? */
764 c = *(*s); /* yes, fetch z */
766 (*s)++; /* advance unless z = NUL */
771 /* TODO: merge with strcpy_and_process_escape_sequences()?
773 static void unescape_string_in_place(char *s1)
776 while ((*s1 = nextchar(&s)) != '\0')
780 static ALWAYS_INLINE int isalnum_(int c)
782 return (isalnum(c) || c == '_');
785 static double my_strtod(char **pp)
788 if (ENABLE_DESKTOP && cp[0] == '0') {
789 /* Might be hex or octal integer: 0x123abc or 07777 */
790 char c = (cp[1] | 0x20);
791 if (c == 'x' || isdigit(cp[1])) {
792 unsigned long long ull = strtoull(cp, pp, 0);
796 if (!isdigit(c) && c != '.')
798 /* else: it may be a floating number. Examples:
799 * 009.123 (*pp points to '9')
800 * 000.123 (*pp points to '.')
801 * fall through to strtod.
805 return strtod(cp, pp);
808 /* -------- working with variables (set/get/copy/etc) -------- */
810 static xhash *iamarray(var *v)
814 while (a->type & VF_CHILD)
817 if (!(a->type & VF_ARRAY)) {
819 a->x.array = hash_init();
824 static void clear_array(xhash *array)
829 for (i = 0; i < array->csize; i++) {
830 hi = array->items[i];
834 free(thi->data.v.string);
837 array->items[i] = NULL;
839 array->glen = array->nel = 0;
842 /* clear a variable */
843 static var *clrvar(var *v)
845 if (!(v->type & VF_FSTR))
848 v->type &= VF_DONTTOUCH;
854 /* assign string value to variable */
855 static var *setvar_p(var *v, char *value)
863 /* same as setvar_p but make a copy of string */
864 static var *setvar_s(var *v, const char *value)
866 return setvar_p(v, (value && *value) ? xstrdup(value) : NULL);
869 /* same as setvar_s but sets USER flag */
870 static var *setvar_u(var *v, const char *value)
872 v = setvar_s(v, value);
877 /* set array element to user string */
878 static void setari_u(var *a, int idx, const char *s)
882 v = findvar(iamarray(a), itoa(idx));
886 /* assign numeric value to variable */
887 static var *setvar_i(var *v, double value)
890 v->type |= VF_NUMBER;
896 static const char *getvar_s(var *v)
898 /* if v is numeric and has no cached string, convert it to string */
899 if ((v->type & (VF_NUMBER | VF_CACHED)) == VF_NUMBER) {
900 fmt_num(g_buf, MAXVARFMT, getvar_s(intvar[CONVFMT]), v->number, TRUE);
901 v->string = xstrdup(g_buf);
902 v->type |= VF_CACHED;
904 return (v->string == NULL) ? "" : v->string;
907 static double getvar_i(var *v)
911 if ((v->type & (VF_NUMBER | VF_CACHED)) == 0) {
915 debug_printf_eval("getvar_i: '%s'->", s);
916 v->number = my_strtod(&s);
917 debug_printf_eval("%f (s:'%s')\n", v->number, s);
918 if (v->type & VF_USER) {
924 debug_printf_eval("getvar_i: '%s'->zero\n", s);
927 v->type |= VF_CACHED;
929 debug_printf_eval("getvar_i: %f\n", v->number);
933 /* Used for operands of bitwise ops */
934 static unsigned long getvar_i_int(var *v)
936 double d = getvar_i(v);
938 /* Casting doubles to longs is undefined for values outside
939 * of target type range. Try to widen it as much as possible */
941 return (unsigned long)d;
942 /* Why? Think about d == -4294967295.0 (assuming 32bit longs) */
943 return - (long) (unsigned long) (-d);
946 static var *copyvar(var *dest, const var *src)
950 dest->type |= (src->type & ~(VF_DONTTOUCH | VF_FSTR));
951 debug_printf_eval("copyvar: number:%f string:'%s'\n", src->number, src->string);
952 dest->number = src->number;
954 dest->string = xstrdup(src->string);
956 handle_special(dest);
960 static var *incvar(var *v)
962 return setvar_i(v, getvar_i(v) + 1.0);
965 /* return true if v is number or numeric string */
966 static int is_numeric(var *v)
969 return ((v->type ^ VF_DIRTY) & (VF_NUMBER | VF_USER | VF_DIRTY));
972 /* return 1 when value of v corresponds to true, 0 otherwise */
973 static int istrue(var *v)
976 return (v->number != 0);
977 return (v->string && v->string[0]);
980 /* temporary variables allocator. Last allocated should be first freed */
981 static var *nvalloc(int n)
989 if ((g_cb->pos - g_cb->nv) + n <= g_cb->size)
995 size = (n <= MINNVBLOCK) ? MINNVBLOCK : n;
996 g_cb = xzalloc(sizeof(nvblock) + size * sizeof(var));
998 g_cb->pos = g_cb->nv;
1000 /*g_cb->next = NULL; - xzalloc did it */
1008 while (v < g_cb->pos) {
1017 static void nvfree(var *v)
1021 if (v < g_cb->nv || v >= g_cb->pos)
1022 syntax_error(EMSG_INTERNAL_ERROR);
1024 for (p = v; p < g_cb->pos; p++) {
1025 if ((p->type & (VF_ARRAY | VF_CHILD)) == VF_ARRAY) {
1026 clear_array(iamarray(p));
1027 free(p->x.array->items);
1030 if (p->type & VF_WALK) {
1032 walker_list *w = p->x.walker;
1033 debug_printf_walker("nvfree: freeing walker @%p\n", &p->x.walker);
1037 debug_printf_walker(" free(%p)\n", w);
1046 while (g_cb->prev && g_cb->pos == g_cb->nv) {
1051 /* ------- awk program text parsing ------- */
1053 /* Parse next token pointed by global pos, place results into global ttt.
1054 * If token isn't expected, give away. Return token class
1056 static uint32_t next_token(uint32_t expected)
1058 #define concat_inserted (G.next_token__concat_inserted)
1059 #define save_tclass (G.next_token__save_tclass)
1060 #define save_info (G.next_token__save_info)
1061 /* Initialized to TC_OPTERM: */
1062 #define ltclass (G.next_token__ltclass)
1071 } else if (concat_inserted) {
1072 concat_inserted = FALSE;
1073 t_tclass = save_tclass;
1079 g_lineno = t_lineno;
1081 while (*p != '\n' && *p != '\0')
1089 debug_printf_parse("%s: token found: TC_EOF\n", __func__);
1090 } else if (*p == '\"') {
1093 while (*p != '\"') {
1095 if (*p == '\0' || *p == '\n')
1096 syntax_error(EMSG_UNEXP_EOS);
1098 *s++ = nextchar(&pp);
1104 debug_printf_parse("%s: token found:'%s' TC_STRING\n", __func__, t_string);
1105 } else if ((expected & TC_REGEXP) && *p == '/') {
1109 if (*p == '\0' || *p == '\n')
1110 syntax_error(EMSG_UNEXP_EOS);
1114 s[-1] = bb_process_escape_sequence((const char **)&pp);
1126 debug_printf_parse("%s: token found:'%s' TC_REGEXP\n", __func__, t_string);
1128 } else if (*p == '.' || isdigit(*p)) {
1131 t_double = my_strtod(&pp);
1134 syntax_error(EMSG_UNEXP_TOKEN);
1136 debug_printf_parse("%s: token found:%f TC_NUMBER\n", __func__, t_double);
1138 /* search for something known */
1143 int l = (unsigned char) *tl++;
1144 if (l == (unsigned char) NTCC) {
1148 /* if token class is expected,
1150 * and it's not a longer word,
1152 if ((tc & (expected | TC_WORD | TC_NEWLINE))
1153 && strncmp(p, tl, l) == 0
1154 && !((tc & TC_WORD) && isalnum_(p[l]))
1156 /* then this is what we are looking for */
1158 debug_printf_parse("%s: token found:'%.*s' t_info:%x\n", __func__, l, p, t_info);
1165 /* not a known token */
1167 /* is it a name? (var/array/function) */
1169 syntax_error(EMSG_UNEXP_TOKEN); /* no */
1172 while (isalnum_(*++p)) {
1177 /* also consume whitespace between functionname and bracket */
1178 if (!(expected & TC_VARIABLE) || (expected & TC_ARRAY))
1182 debug_printf_parse("%s: token found:'%s' TC_FUNCTION\n", __func__, t_string);
1187 debug_printf_parse("%s: token found:'%s' TC_ARRAY\n", __func__, t_string);
1189 debug_printf_parse("%s: token found:'%s' TC_VARIABLE\n", __func__, t_string);
1195 /* skipping newlines in some cases */
1196 if ((ltclass & TC_NOTERM) && (tc & TC_NEWLINE))
1199 /* insert concatenation operator when needed */
1200 if ((ltclass & TC_CONCAT1) && (tc & TC_CONCAT2) && (expected & TC_BINOP)) {
1201 concat_inserted = TRUE;
1205 t_info = OC_CONCAT | SS | P(35);
1212 /* Are we ready for this? */
1213 if (!(ltclass & expected)) {
1214 syntax_error((ltclass & (TC_NEWLINE | TC_EOF)) ?
1215 EMSG_UNEXP_EOS : EMSG_UNEXP_TOKEN);
1219 #undef concat_inserted
1225 static void rollback_token(void)
1230 static node *new_node(uint32_t info)
1234 n = xzalloc(sizeof(node));
1236 n->lineno = g_lineno;
1240 static void mk_re_node(const char *s, node *n, regex_t *re)
1242 n->info = OC_REGEXP;
1245 xregcomp(re, s, REG_EXTENDED);
1246 xregcomp(re + 1, s, REG_EXTENDED | REG_ICASE);
1249 static node *condition(void)
1251 next_token(TC_SEQSTART);
1252 return parse_expr(TC_SEQTERM);
1255 /* parse expression terminated by given argument, return ptr
1256 * to built subtree. Terminator is eaten by parse_expr */
1257 static node *parse_expr(uint32_t iexp)
1265 debug_printf_parse("%s(%x)\n", __func__, iexp);
1268 sn.r.n = glptr = NULL;
1269 xtc = TC_OPERAND | TC_UOPPRE | TC_REGEXP | iexp;
1271 while (!((tc = next_token(xtc)) & iexp)) {
1273 if (glptr && (t_info == (OC_COMPARE | VV | P(39) | 2))) {
1274 /* input redirection (<) attached to glptr node */
1275 debug_printf_parse("%s: input redir\n", __func__);
1276 cn = glptr->l.n = new_node(OC_CONCAT | SS | P(37));
1278 xtc = TC_OPERAND | TC_UOPPRE;
1281 } else if (tc & (TC_BINOP | TC_UOPPOST)) {
1282 debug_printf_parse("%s: TC_BINOP | TC_UOPPOST\n", __func__);
1283 /* for binary and postfix-unary operators, jump back over
1284 * previous operators with higher priority */
1286 while (((t_info & PRIMASK) > (vn->a.n->info & PRIMASK2))
1287 || ((t_info == vn->info) && ((t_info & OPCLSMASK) == OC_COLON))
1291 if ((t_info & OPCLSMASK) == OC_TERNARY)
1293 cn = vn->a.n->r.n = new_node(t_info);
1295 if (tc & TC_BINOP) {
1297 xtc = TC_OPERAND | TC_UOPPRE | TC_REGEXP;
1298 if ((t_info & OPCLSMASK) == OC_PGETLINE) {
1300 next_token(TC_GETLINE);
1301 /* give maximum priority to this pipe */
1302 cn->info &= ~PRIMASK;
1303 xtc = TC_OPERAND | TC_UOPPRE | TC_BINOP | iexp;
1307 xtc = TC_OPERAND | TC_UOPPRE | TC_BINOP | iexp;
1312 debug_printf_parse("%s: other\n", __func__);
1313 /* for operands and prefix-unary operators, attach them
1316 cn = vn->r.n = new_node(t_info);
1318 xtc = TC_OPERAND | TC_UOPPRE | TC_REGEXP;
1319 if (tc & (TC_OPERAND | TC_REGEXP)) {
1320 debug_printf_parse("%s: TC_OPERAND | TC_REGEXP\n", __func__);
1321 xtc = TC_UOPPRE | TC_UOPPOST | TC_BINOP | TC_OPERAND | iexp;
1322 /* one should be very careful with switch on tclass -
1323 * only simple tclasses should be used! */
1327 debug_printf_parse("%s: TC_VARIABLE | TC_ARRAY\n", __func__);
1329 v = hash_search(ahash, t_string);
1331 cn->info = OC_FNARG;
1332 cn->l.aidx = v->x.aidx;
1334 cn->l.v = newvar(t_string);
1336 if (tc & TC_ARRAY) {
1338 cn->r.n = parse_expr(TC_ARRTERM);
1344 debug_printf_parse("%s: TC_NUMBER | TC_STRING\n", __func__);
1346 v = cn->l.v = xzalloc(sizeof(var));
1348 setvar_i(v, t_double);
1350 setvar_s(v, t_string);
1354 debug_printf_parse("%s: TC_REGEXP\n", __func__);
1355 mk_re_node(t_string, cn, xzalloc(sizeof(regex_t)*2));
1359 debug_printf_parse("%s: TC_FUNCTION\n", __func__);
1361 cn->r.f = newfunc(t_string);
1362 cn->l.n = condition();
1366 debug_printf_parse("%s: TC_SEQSTART\n", __func__);
1367 cn = vn->r.n = parse_expr(TC_SEQTERM);
1369 syntax_error("Empty sequence");
1374 debug_printf_parse("%s: TC_GETLINE\n", __func__);
1376 xtc = TC_OPERAND | TC_UOPPRE | TC_BINOP | iexp;
1380 debug_printf_parse("%s: TC_BUILTIN\n", __func__);
1381 cn->l.n = condition();
1385 debug_printf_parse("%s: TC_LENGTH\n", __func__);
1386 next_token(TC_SEQSTART | TC_OPTERM | TC_GRPTERM);
1388 if (t_tclass & TC_SEQSTART) {
1389 /* It was a "(" token. Handle just like TC_BUILTIN */
1390 cn->l.n = condition();
1398 debug_printf_parse("%s() returns %p\n", __func__, sn.r.n);
1402 /* add node to chain. Return ptr to alloc'd node */
1403 static node *chain_node(uint32_t info)
1408 seq->first = seq->last = new_node(0);
1410 if (seq->programname != g_progname) {
1411 seq->programname = g_progname;
1412 n = chain_node(OC_NEWSOURCE);
1413 n->l.new_progname = xstrdup(g_progname);
1418 seq->last = n->a.n = new_node(OC_DONE);
1423 static void chain_expr(uint32_t info)
1427 n = chain_node(info);
1428 n->l.n = parse_expr(TC_OPTERM | TC_GRPTERM);
1429 if (t_tclass & TC_GRPTERM)
1433 static node *chain_loop(node *nn)
1435 node *n, *n2, *save_brk, *save_cont;
1437 save_brk = break_ptr;
1438 save_cont = continue_ptr;
1440 n = chain_node(OC_BR | Vx);
1441 continue_ptr = new_node(OC_EXEC);
1442 break_ptr = new_node(OC_EXEC);
1444 n2 = chain_node(OC_EXEC | Vx);
1447 continue_ptr->a.n = n2;
1448 break_ptr->a.n = n->r.n = seq->last;
1450 continue_ptr = save_cont;
1451 break_ptr = save_brk;
1456 /* parse group and attach it to chain */
1457 static void chain_group(void)
1463 c = next_token(TC_GRPSEQ);
1464 } while (c & TC_NEWLINE);
1466 if (c & TC_GRPSTART) {
1467 debug_printf_parse("%s: TC_GRPSTART\n", __func__);
1468 while (next_token(TC_GRPSEQ | TC_GRPTERM) != TC_GRPTERM) {
1469 debug_printf_parse("%s: !TC_GRPTERM\n", __func__);
1470 if (t_tclass & TC_NEWLINE)
1475 debug_printf_parse("%s: TC_GRPTERM\n", __func__);
1476 } else if (c & (TC_OPSEQ | TC_OPTERM)) {
1477 debug_printf_parse("%s: TC_OPSEQ | TC_OPTERM\n", __func__);
1479 chain_expr(OC_EXEC | Vx);
1482 debug_printf_parse("%s: TC_STATEMNT(?)\n", __func__);
1483 switch (t_info & OPCLSMASK) {
1485 debug_printf_parse("%s: ST_IF\n", __func__);
1486 n = chain_node(OC_BR | Vx);
1487 n->l.n = condition();
1489 n2 = chain_node(OC_EXEC);
1491 if (next_token(TC_GRPSEQ | TC_GRPTERM | TC_ELSE) == TC_ELSE) {
1493 n2->a.n = seq->last;
1500 debug_printf_parse("%s: ST_WHILE\n", __func__);
1502 n = chain_loop(NULL);
1507 debug_printf_parse("%s: ST_DO\n", __func__);
1508 n2 = chain_node(OC_EXEC);
1509 n = chain_loop(NULL);
1511 next_token(TC_WHILE);
1512 n->l.n = condition();
1516 debug_printf_parse("%s: ST_FOR\n", __func__);
1517 next_token(TC_SEQSTART);
1518 n2 = parse_expr(TC_SEMICOL | TC_SEQTERM);
1519 if (t_tclass & TC_SEQTERM) { /* for-in */
1520 if (!n2 || (n2->info & OPCLSMASK) != OC_IN)
1521 syntax_error(EMSG_UNEXP_TOKEN);
1522 n = chain_node(OC_WALKINIT | VV);
1525 n = chain_loop(NULL);
1526 n->info = OC_WALKNEXT | Vx;
1528 } else { /* for (;;) */
1529 n = chain_node(OC_EXEC | Vx);
1531 n2 = parse_expr(TC_SEMICOL);
1532 n3 = parse_expr(TC_SEQTERM);
1542 debug_printf_parse("%s: OC_PRINT[F]\n", __func__);
1543 n = chain_node(t_info);
1544 n->l.n = parse_expr(TC_OPTERM | TC_OUTRDR | TC_GRPTERM);
1545 if (t_tclass & TC_OUTRDR) {
1547 n->r.n = parse_expr(TC_OPTERM | TC_GRPTERM);
1549 if (t_tclass & TC_GRPTERM)
1554 debug_printf_parse("%s: OC_BREAK\n", __func__);
1555 n = chain_node(OC_EXEC);
1561 debug_printf_parse("%s: OC_CONTINUE\n", __func__);
1562 n = chain_node(OC_EXEC);
1563 n->a.n = continue_ptr;
1567 /* delete, next, nextfile, return, exit */
1569 debug_printf_parse("%s: default\n", __func__);
1575 static void parse_program(char *p)
1584 while ((tclass = next_token(TC_EOF | TC_OPSEQ | TC_GRPSTART |
1585 TC_OPTERM | TC_BEGIN | TC_END | TC_FUNCDECL)) != TC_EOF) {
1587 if (tclass & TC_OPTERM) {
1588 debug_printf_parse("%s: TC_OPTERM\n", __func__);
1593 if (tclass & TC_BEGIN) {
1594 debug_printf_parse("%s: TC_BEGIN\n", __func__);
1597 } else if (tclass & TC_END) {
1598 debug_printf_parse("%s: TC_END\n", __func__);
1601 } else if (tclass & TC_FUNCDECL) {
1602 debug_printf_parse("%s: TC_FUNCDECL\n", __func__);
1603 next_token(TC_FUNCTION);
1605 f = newfunc(t_string);
1606 f->body.first = NULL;
1608 while (next_token(TC_VARIABLE | TC_SEQTERM) & TC_VARIABLE) {
1609 v = findvar(ahash, t_string);
1610 v->x.aidx = f->nargs++;
1612 if (next_token(TC_COMMA | TC_SEQTERM) & TC_SEQTERM)
1618 } else if (tclass & TC_OPSEQ) {
1619 debug_printf_parse("%s: TC_OPSEQ\n", __func__);
1621 cn = chain_node(OC_TEST);
1622 cn->l.n = parse_expr(TC_OPTERM | TC_EOF | TC_GRPSTART);
1623 if (t_tclass & TC_GRPSTART) {
1624 debug_printf_parse("%s: TC_GRPSTART\n", __func__);
1628 debug_printf_parse("%s: !TC_GRPSTART\n", __func__);
1629 chain_node(OC_PRINT);
1631 cn->r.n = mainseq.last;
1632 } else /* if (tclass & TC_GRPSTART) */ {
1633 debug_printf_parse("%s: TC_GRPSTART(?)\n", __func__);
1638 debug_printf_parse("%s: TC_EOF\n", __func__);
1642 /* -------- program execution part -------- */
1644 static node *mk_splitter(const char *s, tsplitter *spl)
1652 if ((n->info & OPCLSMASK) == OC_REGEXP) {
1654 regfree(ire); // TODO: nuke ire, use re+1?
1656 if (s[0] && s[1]) { /* strlen(s) > 1 */
1657 mk_re_node(s, n, re);
1659 n->info = (uint32_t) s[0];
1665 /* use node as a regular expression. Supplied with node ptr and regex_t
1666 * storage space. Return ptr to regex (if result points to preg, it should
1667 * be later regfree'd manually
1669 static regex_t *as_regex(node *op, regex_t *preg)
1675 if ((op->info & OPCLSMASK) == OC_REGEXP) {
1676 return icase ? op->r.ire : op->l.re;
1679 s = getvar_s(evaluate(op, v));
1681 cflags = icase ? REG_EXTENDED | REG_ICASE : REG_EXTENDED;
1682 /* Testcase where REG_EXTENDED fails (unpaired '{'):
1683 * echo Hi | awk 'gsub("@(samp|code|file)\{","");'
1684 * gawk 3.1.5 eats this. We revert to ~REG_EXTENDED
1685 * (maybe gsub is not supposed to use REG_EXTENDED?).
1687 if (regcomp(preg, s, cflags)) {
1688 cflags &= ~REG_EXTENDED;
1689 xregcomp(preg, s, cflags);
1695 /* gradually increasing buffer.
1696 * note that we reallocate even if n == old_size,
1697 * and thus there is at least one extra allocated byte.
1699 static char* qrealloc(char *b, int n, int *size)
1701 if (!b || n >= *size) {
1702 *size = n + (n>>1) + 80;
1703 b = xrealloc(b, *size);
1708 /* resize field storage space */
1709 static void fsrealloc(int size)
1713 if (size >= maxfields) {
1715 maxfields = size + 16;
1716 Fields = xrealloc(Fields, maxfields * sizeof(Fields[0]));
1717 for (; i < maxfields; i++) {
1718 Fields[i].type = VF_SPECIAL;
1719 Fields[i].string = NULL;
1722 /* if size < nfields, clear extra field variables */
1723 for (i = size; i < nfields; i++) {
1729 static int awk_split(const char *s, node *spl, char **slist)
1734 regmatch_t pmatch[2]; // TODO: why [2]? [1] is enough...
1736 /* in worst case, each char would be a separate field */
1737 *slist = s1 = xzalloc(strlen(s) * 2 + 3);
1740 c[0] = c[1] = (char)spl->info;
1742 if (*getvar_s(intvar[RS]) == '\0')
1746 if ((spl->info & OPCLSMASK) == OC_REGEXP) { /* regex split */
1748 return n; /* "": zero fields */
1749 n++; /* at least one field will be there */
1751 l = strcspn(s, c+2); /* len till next NUL or \n */
1752 if (regexec(icase ? spl->r.ire : spl->l.re, s, 1, pmatch, 0) == 0
1753 && pmatch[0].rm_so <= l
1755 l = pmatch[0].rm_so;
1756 if (pmatch[0].rm_eo == 0) {
1760 n++; /* we saw yet another delimiter */
1762 pmatch[0].rm_eo = l;
1767 /* make sure we remove *all* of the separator chars */
1770 } while (++l < pmatch[0].rm_eo);
1772 s += pmatch[0].rm_eo;
1776 if (c[0] == '\0') { /* null split */
1784 if (c[0] != ' ') { /* single-character split */
1786 c[0] = toupper(c[0]);
1787 c[1] = tolower(c[1]);
1791 while ((s1 = strpbrk(s1, c)) != NULL) {
1799 s = skip_whitespace(s);
1803 while (*s && !isspace(*s))
1810 static void split_f0(void)
1812 /* static char *fstrings; */
1813 #define fstrings (G.split_f0__fstrings)
1824 n = awk_split(getvar_s(intvar[F0]), &fsplitter.n, &fstrings);
1827 for (i = 0; i < n; i++) {
1828 Fields[i].string = nextword(&s);
1829 Fields[i].type |= (VF_FSTR | VF_USER | VF_DIRTY);
1832 /* set NF manually to avoid side effects */
1834 intvar[NF]->type = VF_NUMBER | VF_SPECIAL;
1835 intvar[NF]->number = nfields;
1839 /* perform additional actions when some internal variables changed */
1840 static void handle_special(var *v)
1844 const char *sep, *s;
1845 int sl, l, len, i, bsize;
1847 if (!(v->type & VF_SPECIAL))
1850 if (v == intvar[NF]) {
1851 n = (int)getvar_i(v);
1854 /* recalculate $0 */
1855 sep = getvar_s(intvar[OFS]);
1859 for (i = 0; i < n; i++) {
1860 s = getvar_s(&Fields[i]);
1863 memcpy(b+len, sep, sl);
1866 b = qrealloc(b, len+l+sl, &bsize);
1867 memcpy(b+len, s, l);
1872 setvar_p(intvar[F0], b);
1875 } else if (v == intvar[F0]) {
1876 is_f0_split = FALSE;
1878 } else if (v == intvar[FS]) {
1880 * The POSIX-2008 standard says that changing FS should have no effect on the
1881 * current input line, but only on the next one. The language is:
1883 * > Before the first reference to a field in the record is evaluated, the record
1884 * > shall be split into fields, according to the rules in Regular Expressions,
1885 * > using the value of FS that was current at the time the record was read.
1887 * So, split up current line before assignment to FS:
1891 mk_splitter(getvar_s(v), &fsplitter);
1892 } else if (v == intvar[RS]) {
1893 mk_splitter(getvar_s(v), &rsplitter);
1894 } else if (v == intvar[IGNORECASE]) {
1897 n = getvar_i(intvar[NF]);
1898 setvar_i(intvar[NF], n > v-Fields ? n : v-Fields+1);
1899 /* right here v is invalid. Just to note... */
1903 /* step through func/builtin/etc arguments */
1904 static node *nextarg(node **pn)
1909 if (n && (n->info & OPCLSMASK) == OC_COMMA) {
1918 static void hashwalk_init(var *v, xhash *array)
1923 walker_list *prev_walker;
1925 if (v->type & VF_WALK) {
1926 prev_walker = v->x.walker;
1931 debug_printf_walker("hashwalk_init: prev_walker:%p\n", prev_walker);
1933 w = v->x.walker = xzalloc(sizeof(*w) + array->glen + 1); /* why + 1? */
1934 debug_printf_walker(" walker@%p=%p\n", &v->x.walker, w);
1935 w->cur = w->end = w->wbuf;
1936 w->prev = prev_walker;
1937 for (i = 0; i < array->csize; i++) {
1938 hi = array->items[i];
1940 strcpy(w->end, hi->name);
1947 static int hashwalk_next(var *v)
1949 walker_list *w = v->x.walker;
1951 if (w->cur >= w->end) {
1952 walker_list *prev_walker = w->prev;
1954 debug_printf_walker("end of iteration, free(walker@%p:%p), prev_walker:%p\n", &v->x.walker, w, prev_walker);
1956 v->x.walker = prev_walker;
1960 setvar_s(v, nextword(&w->cur));
1964 /* evaluate node, return 1 when result is true, 0 otherwise */
1965 static int ptest(node *pattern)
1967 /* ptest__v is "static": to save stack space? */
1968 return istrue(evaluate(pattern, &G.ptest__v));
1971 /* read next record from stream rsm into a variable v */
1972 static int awk_getline(rstream *rsm, var *v)
1975 regmatch_t pmatch[2];
1976 int size, a, p, pp = 0;
1977 int fd, so, eo, r, rp;
1980 debug_printf_eval("entered %s()\n", __func__);
1982 /* we're using our own buffer since we need access to accumulating
1985 fd = fileno(rsm->F);
1990 c = (char) rsplitter.n.info;
1994 m = qrealloc(m, 256, &size);
2001 if ((rsplitter.n.info & OPCLSMASK) == OC_REGEXP) {
2002 if (regexec(icase ? rsplitter.n.r.ire : rsplitter.n.l.re,
2003 b, 1, pmatch, 0) == 0) {
2004 so = pmatch[0].rm_so;
2005 eo = pmatch[0].rm_eo;
2009 } else if (c != '\0') {
2010 s = strchr(b+pp, c);
2012 s = memchr(b+pp, '\0', p - pp);
2019 while (b[rp] == '\n')
2021 s = strstr(b+rp, "\n\n");
2024 while (b[eo] == '\n')
2033 memmove(m, m+a, p+1);
2038 m = qrealloc(m, a+p+128, &size);
2041 p += safe_read(fd, b+p, size-p-1);
2045 setvar_i(intvar[ERRNO], errno);
2054 c = b[so]; b[so] = '\0';
2058 c = b[eo]; b[eo] = '\0';
2059 setvar_s(intvar[RT], b+so);
2068 debug_printf_eval("returning from %s(): %d\n", __func__, r);
2073 static int fmt_num(char *b, int size, const char *format, double n, int int_as_int)
2077 const char *s = format;
2079 if (int_as_int && n == (long long)n) {
2080 r = snprintf(b, size, "%lld", (long long)n);
2082 do { c = *s; } while (c && *++s);
2083 if (strchr("diouxX", c)) {
2084 r = snprintf(b, size, format, (int)n);
2085 } else if (strchr("eEfgG", c)) {
2086 r = snprintf(b, size, format, n);
2088 syntax_error(EMSG_INV_FMT);
2094 /* formatted output into an allocated buffer, return ptr to buffer */
2095 static char *awk_printf(node *n)
2100 int i, j, incr, bsize;
2105 fmt = f = xstrdup(getvar_s(evaluate(nextarg(&n), v)));
2110 while (*f && (*f != '%' || *++f == '%'))
2112 while (*f && !isalpha(*f)) {
2114 syntax_error("%*x formats are not supported");
2118 incr = (f - s) + MAXVARFMT;
2119 b = qrealloc(b, incr + i, &bsize);
2125 arg = evaluate(nextarg(&n), v);
2128 if (c == 'c' || !c) {
2129 i += sprintf(b+i, s, is_numeric(arg) ?
2130 (char)getvar_i(arg) : *getvar_s(arg));
2131 } else if (c == 's') {
2133 b = qrealloc(b, incr+i+strlen(s1), &bsize);
2134 i += sprintf(b+i, s, s1);
2136 i += fmt_num(b+i, incr, s, getvar_i(arg), FALSE);
2140 /* if there was an error while sprintf, return value is negative */
2147 b = xrealloc(b, i + 1);
2152 /* Common substitution routine.
2153 * Replace (nm)'th substring of (src) that matches (rn) with (repl),
2154 * store result into (dest), return number of substitutions.
2155 * If nm = 0, replace all matches.
2156 * If src or dst is NULL, use $0.
2157 * If subexp != 0, enable subexpression matching (\1-\9).
2159 static int awk_sub(node *rn, const char *repl, int nm, var *src, var *dest, int subexp)
2163 int match_no, residx, replen, resbufsize;
2165 regmatch_t pmatch[10];
2166 regex_t sreg, *regex;
2172 regex = as_regex(rn, &sreg);
2173 sp = getvar_s(src ? src : intvar[F0]);
2174 replen = strlen(repl);
2175 while (regexec(regex, sp, 10, pmatch, regexec_flags) == 0) {
2176 int so = pmatch[0].rm_so;
2177 int eo = pmatch[0].rm_eo;
2179 //bb_error_msg("match %u: [%u,%u] '%s'%p", match_no+1, so, eo, sp,sp);
2180 resbuf = qrealloc(resbuf, residx + eo + replen, &resbufsize);
2181 memcpy(resbuf + residx, sp, eo);
2183 if (++match_no >= nm) {
2188 residx -= (eo - so);
2190 for (s = repl; *s; s++) {
2191 char c = resbuf[residx++] = *s;
2196 if (c == '&' || (subexp && c >= '0' && c <= '9')) {
2198 residx -= ((nbs + 3) >> 1);
2205 resbuf[residx++] = c;
2207 int n = pmatch[j].rm_eo - pmatch[j].rm_so;
2208 resbuf = qrealloc(resbuf, residx + replen + n, &resbufsize);
2209 memcpy(resbuf + residx, sp + pmatch[j].rm_so, n);
2217 regexec_flags = REG_NOTBOL;
2222 /* Empty match (e.g. "b*" will match anywhere).
2223 * Advance by one char. */
2225 //gsub(/\<b*/,"") on "abc" will reach this point, advance to "bc"
2226 //... and will erroneously match "b" even though it is NOT at the word start.
2227 //we need REG_NOTBOW but it does not exist...
2228 //TODO: if EXTRA_COMPAT=y, use GNU matching and re_search,
2229 //it should be able to do it correctly.
2230 /* Subtle: this is safe only because
2231 * qrealloc allocated at least one extra byte */
2232 resbuf[residx] = *sp;
2240 resbuf = qrealloc(resbuf, residx + strlen(sp), &resbufsize);
2241 strcpy(resbuf + residx, sp);
2243 //bb_error_msg("end sp:'%s'%p", sp,sp);
2244 setvar_p(dest ? dest : intvar[F0], resbuf);
2250 static NOINLINE int do_mktime(const char *ds)
2255 /*memset(&then, 0, sizeof(then)); - not needed */
2256 then.tm_isdst = -1; /* default is unknown */
2258 /* manpage of mktime says these fields are ints,
2259 * so we can sscanf stuff directly into them */
2260 count = sscanf(ds, "%u %u %u %u %u %u %d",
2261 &then.tm_year, &then.tm_mon, &then.tm_mday,
2262 &then.tm_hour, &then.tm_min, &then.tm_sec,
2266 || (unsigned)then.tm_mon < 1
2267 || (unsigned)then.tm_year < 1900
2273 then.tm_year -= 1900;
2275 return mktime(&then);
2278 static NOINLINE var *exec_builtin(node *op, var *res)
2280 #define tspl (G.exec_builtin__tspl)
2286 regmatch_t pmatch[2];
2295 isr = info = op->info;
2298 av[2] = av[3] = NULL;
2299 for (i = 0; i < 4 && op; i++) {
2300 an[i] = nextarg(&op);
2301 if (isr & 0x09000000)
2302 av[i] = evaluate(an[i], &tv[i]);
2303 if (isr & 0x08000000)
2304 as[i] = getvar_s(av[i]);
2309 if ((uint32_t)nargs < (info >> 30))
2310 syntax_error(EMSG_TOO_FEW_ARGS);
2316 if (ENABLE_FEATURE_AWK_LIBM)
2317 setvar_i(res, atan2(getvar_i(av[0]), getvar_i(av[1])));
2319 syntax_error(EMSG_NO_MATH);
2326 spl = (an[2]->info & OPCLSMASK) == OC_REGEXP ?
2327 an[2] : mk_splitter(getvar_s(evaluate(an[2], &tv[2])), &tspl);
2332 n = awk_split(as[0], spl, &s);
2334 clear_array(iamarray(av[1]));
2335 for (i = 1; i <= n; i++)
2336 setari_u(av[1], i, nextword(&s));
2346 i = getvar_i(av[1]) - 1;
2351 n = (nargs > 2) ? getvar_i(av[2]) : l-i;
2354 s = xstrndup(as[0]+i, n);
2359 /* Bitwise ops must assume that operands are unsigned. GNU Awk 3.1.5:
2360 * awk '{ print or(-1,1) }' gives "4.29497e+09", not "-2.xxxe+09" */
2362 setvar_i(res, getvar_i_int(av[0]) & getvar_i_int(av[1]));
2366 setvar_i(res, ~getvar_i_int(av[0]));
2370 setvar_i(res, getvar_i_int(av[0]) << getvar_i_int(av[1]));
2374 setvar_i(res, getvar_i_int(av[0]) | getvar_i_int(av[1]));
2378 setvar_i(res, getvar_i_int(av[0]) >> getvar_i_int(av[1]));
2382 setvar_i(res, getvar_i_int(av[0]) ^ getvar_i_int(av[1]));
2388 s1 = s = xstrdup(as[0]);
2390 //*s1 = (info == B_up) ? toupper(*s1) : tolower(*s1);
2391 if ((unsigned char)((*s1 | 0x20) - 'a') <= ('z' - 'a'))
2392 *s1 = (info == B_up) ? (*s1 & 0xdf) : (*s1 | 0x20);
2402 l = strlen(as[0]) - ll;
2403 if (ll > 0 && l >= 0) {
2405 char *s = strstr(as[0], as[1]);
2407 n = (s - as[0]) + 1;
2409 /* this piece of code is terribly slow and
2410 * really should be rewritten
2412 for (i = 0; i <= l; i++) {
2413 if (strncasecmp(as[0]+i, as[1], ll) == 0) {
2425 tt = getvar_i(av[1]);
2428 //s = (nargs > 0) ? as[0] : "%a %b %d %H:%M:%S %Z %Y";
2429 i = strftime(g_buf, MAXVARFMT,
2430 ((nargs > 0) ? as[0] : "%a %b %d %H:%M:%S %Z %Y"),
2433 setvar_s(res, g_buf);
2437 setvar_i(res, do_mktime(as[0]));
2441 re = as_regex(an[1], &sreg);
2442 n = regexec(re, as[0], 1, pmatch, 0);
2447 pmatch[0].rm_so = 0;
2448 pmatch[0].rm_eo = -1;
2450 setvar_i(newvar("RSTART"), pmatch[0].rm_so);
2451 setvar_i(newvar("RLENGTH"), pmatch[0].rm_eo - pmatch[0].rm_so);
2452 setvar_i(res, pmatch[0].rm_so);
2458 awk_sub(an[0], as[1], getvar_i(av[2]), av[3], res, TRUE);
2462 setvar_i(res, awk_sub(an[0], as[1], 0, av[2], av[2], FALSE));
2466 setvar_i(res, awk_sub(an[0], as[1], 1, av[2], av[2], FALSE));
2476 * Evaluate node - the heart of the program. Supplied with subtree
2477 * and place where to store result. returns ptr to result.
2479 #define XC(n) ((n) >> 8)
2481 static var *evaluate(node *op, var *res)
2483 /* This procedure is recursive so we should count every byte */
2484 #define fnargs (G.evaluate__fnargs)
2485 /* seed is initialized to 1 */
2486 #define seed (G.evaluate__seed)
2487 #define sreg (G.evaluate__sreg)
2492 return setvar_s(res, NULL);
2494 debug_printf_eval("entered %s()\n", __func__);
2502 } L = L; /* for compiler */
2513 opn = (opinfo & OPNMASK);
2514 g_lineno = op->lineno;
2516 debug_printf_eval("opinfo:%08x opn:%08x\n", opinfo, opn);
2518 /* "delete" is special:
2519 * "delete array[var--]" must evaluate index expr only once,
2520 * must not evaluate it in "execute inevitable things" part.
2522 if (XC(opinfo & OPCLSMASK) == XC(OC_DELETE)) {
2523 uint32_t info = op1->info & OPCLSMASK;
2526 debug_printf_eval("DELETE\n");
2527 if (info == OC_VAR) {
2529 } else if (info == OC_FNARG) {
2530 v = &fnargs[op1->l.aidx];
2532 syntax_error(EMSG_NOT_ARRAY);
2534 if (op1->r.n) { /* array ref? */
2536 s = getvar_s(evaluate(op1->r.n, v1));
2537 hash_remove(iamarray(v), s);
2539 clear_array(iamarray(v));
2544 /* execute inevitable things */
2545 if (opinfo & OF_RES1)
2546 L.v = evaluate(op1, v1);
2547 if (opinfo & OF_RES2)
2548 R.v = evaluate(op->r.n, v1+1);
2549 if (opinfo & OF_STR1) {
2550 L.s = getvar_s(L.v);
2551 debug_printf_eval("L.s:'%s'\n", L.s);
2553 if (opinfo & OF_STR2) {
2554 R.s = getvar_s(R.v);
2555 debug_printf_eval("R.s:'%s'\n", R.s);
2557 if (opinfo & OF_NUM1) {
2558 L_d = getvar_i(L.v);
2559 debug_printf_eval("L_d:%f\n", L_d);
2562 debug_printf_eval("switch(0x%x)\n", XC(opinfo & OPCLSMASK));
2563 switch (XC(opinfo & OPCLSMASK)) {
2565 /* -- iterative node type -- */
2569 if ((op1->info & OPCLSMASK) == OC_COMMA) {
2570 /* it's range pattern */
2571 if ((opinfo & OF_CHECKED) || ptest(op1->l.n)) {
2572 op->info |= OF_CHECKED;
2573 if (ptest(op1->r.n))
2574 op->info &= ~OF_CHECKED;
2580 op = ptest(op1) ? op->a.n : op->r.n;
2584 /* just evaluate an expression, also used as unconditional jump */
2588 /* branch, used in if-else and various loops */
2590 op = istrue(L.v) ? op->a.n : op->r.n;
2593 /* initialize for-in loop */
2594 case XC( OC_WALKINIT ):
2595 hashwalk_init(L.v, iamarray(R.v));
2598 /* get next array item */
2599 case XC( OC_WALKNEXT ):
2600 op = hashwalk_next(L.v) ? op->a.n : op->r.n;
2603 case XC( OC_PRINT ):
2604 case XC( OC_PRINTF ): {
2608 rstream *rsm = newfile(R.s);
2611 rsm->F = popen(R.s, "w");
2613 bb_perror_msg_and_die("popen");
2616 rsm->F = xfopen(R.s, opn=='w' ? "w" : "a");
2622 if ((opinfo & OPCLSMASK) == OC_PRINT) {
2624 fputs(getvar_s(intvar[F0]), F);
2627 var *v = evaluate(nextarg(&op1), v1);
2628 if (v->type & VF_NUMBER) {
2629 fmt_num(g_buf, MAXVARFMT, getvar_s(intvar[OFMT]),
2633 fputs(getvar_s(v), F);
2637 fputs(getvar_s(intvar[OFS]), F);
2640 fputs(getvar_s(intvar[ORS]), F);
2642 } else { /* OC_PRINTF */
2643 char *s = awk_printf(op1);
2651 /* case XC( OC_DELETE ): - moved to happen before arg evaluation */
2653 case XC( OC_NEWSOURCE ):
2654 g_progname = op->l.new_progname;
2657 case XC( OC_RETURN ):
2661 case XC( OC_NEXTFILE ):
2672 /* -- recursive node type -- */
2675 debug_printf_eval("VAR\n");
2677 if (L.v == intvar[NF])
2681 case XC( OC_FNARG ):
2682 debug_printf_eval("FNARG[%d]\n", op->l.aidx);
2683 L.v = &fnargs[op->l.aidx];
2685 res = op->r.n ? findvar(iamarray(L.v), R.s) : L.v;
2689 setvar_i(res, hash_search(iamarray(R.v), L.s) ? 1 : 0);
2692 case XC( OC_REGEXP ):
2694 L.s = getvar_s(intvar[F0]);
2697 case XC( OC_MATCH ):
2701 regex_t *re = as_regex(op1, &sreg);
2702 int i = regexec(re, L.s, 0, NULL, 0);
2705 setvar_i(res, (i == 0) ^ (opn == '!'));
2710 debug_printf_eval("MOVE\n");
2711 /* if source is a temporary string, jusk relink it to dest */
2712 //Disabled: if R.v is numeric but happens to have cached R.v->string,
2713 //then L.v ends up being a string, which is wrong
2714 // if (R.v == v1+1 && R.v->string) {
2715 // res = setvar_p(L.v, R.v->string);
2716 // R.v->string = NULL;
2718 res = copyvar(L.v, R.v);
2722 case XC( OC_TERNARY ):
2723 if ((op->r.n->info & OPCLSMASK) != OC_COLON)
2724 syntax_error(EMSG_POSSIBLE_ERROR);
2725 res = evaluate(istrue(L.v) ? op->r.n->l.n : op->r.n->r.n, res);
2728 case XC( OC_FUNC ): {
2730 const char *sv_progname;
2732 /* The body might be empty, still has to eval the args */
2733 if (!op->r.n->info && !op->r.f->body.first)
2734 syntax_error(EMSG_UNDEF_FUNC);
2736 vbeg = v = nvalloc(op->r.f->nargs + 1);
2738 var *arg = evaluate(nextarg(&op1), v1);
2740 v->type |= VF_CHILD;
2742 if (++v - vbeg >= op->r.f->nargs)
2748 sv_progname = g_progname;
2750 res = evaluate(op->r.f->body.first, res);
2752 g_progname = sv_progname;
2759 case XC( OC_GETLINE ):
2760 case XC( OC_PGETLINE ): {
2767 if ((opinfo & OPCLSMASK) == OC_PGETLINE) {
2768 rsm->F = popen(L.s, "r");
2769 rsm->is_pipe = TRUE;
2771 rsm->F = fopen_for_read(L.s); /* not xfopen! */
2776 iF = next_input_file();
2780 if (!rsm || !rsm->F) {
2781 setvar_i(intvar[ERRNO], errno);
2789 i = awk_getline(rsm, R.v);
2790 if (i > 0 && !op1) {
2791 incvar(intvar[FNR]);
2798 /* simple builtins */
2799 case XC( OC_FBLTIN ): {
2800 double R_d = R_d; /* for compiler */
2804 R_d = (long long)L_d;
2808 R_d = (double)rand() / (double)RAND_MAX;
2812 if (ENABLE_FEATURE_AWK_LIBM) {
2818 if (ENABLE_FEATURE_AWK_LIBM) {
2824 if (ENABLE_FEATURE_AWK_LIBM) {
2830 if (ENABLE_FEATURE_AWK_LIBM) {
2836 if (ENABLE_FEATURE_AWK_LIBM) {
2841 syntax_error(EMSG_NO_MATH);
2846 seed = op1 ? (unsigned)L_d : (unsigned)time(NULL);
2855 debug_printf_eval("length: L.s:'%s'\n", L.s);
2857 L.s = getvar_s(intvar[F0]);
2858 debug_printf_eval("length: L.s='%s'\n", L.s);
2860 else if (L.v->type & VF_ARRAY) {
2861 R_d = L.v->x.array->nel;
2862 debug_printf_eval("length: array_len:%d\n", L.v->x.array->nel);
2870 R_d = (ENABLE_FEATURE_ALLOW_EXEC && L.s && *L.s)
2871 ? (system(L.s) >> 8) : 0;
2877 } else if (L.s && *L.s) {
2878 rstream *rsm = newfile(L.s);
2888 rsm = (rstream *)hash_search(fdhash, L.s);
2889 debug_printf_eval("OC_FBLTIN F_cl rsm:%p\n", rsm);
2891 debug_printf_eval("OC_FBLTIN F_cl "
2892 "rsm->is_pipe:%d, ->F:%p\n",
2893 rsm->is_pipe, rsm->F);
2894 /* Can be NULL if open failed. Example:
2895 * getline line <"doesnt_exist";
2896 * close("doesnt_exist"); <--- here rsm->F is NULL
2899 err = rsm->is_pipe ? pclose(rsm->F) : fclose(rsm->F);
2901 hash_remove(fdhash, L.s);
2904 setvar_i(intvar[ERRNO], errno);
2913 case XC( OC_BUILTIN ):
2914 res = exec_builtin(op, res);
2917 case XC( OC_SPRINTF ):
2918 setvar_p(res, awk_printf(op1));
2921 case XC( OC_UNARY ): {
2924 Ld = R_d = getvar_i(R.v);
2951 case XC( OC_FIELD ): {
2952 int i = (int)getvar_i(R.v);
2954 syntax_error(EMSG_NEGATIVE_FIELD);
2961 res = &Fields[i - 1];
2966 /* concatenation (" ") and index joining (",") */
2967 case XC( OC_CONCAT ):
2968 case XC( OC_COMMA ): {
2969 const char *sep = "";
2970 if ((opinfo & OPCLSMASK) == OC_COMMA)
2971 sep = getvar_s(intvar[SUBSEP]);
2972 setvar_p(res, xasprintf("%s%s%s", L.s, sep, R.s));
2977 setvar_i(res, istrue(L.v) ? ptest(op->r.n) : 0);
2981 setvar_i(res, istrue(L.v) ? 1 : ptest(op->r.n));
2984 case XC( OC_BINARY ):
2985 case XC( OC_REPLACE ): {
2986 double R_d = getvar_i(R.v);
2987 debug_printf_eval("BINARY/REPLACE: R_d:%f opn:%c\n", R_d, opn);
3000 syntax_error(EMSG_DIV_BY_ZERO);
3004 if (ENABLE_FEATURE_AWK_LIBM)
3005 L_d = pow(L_d, R_d);
3007 syntax_error(EMSG_NO_MATH);
3011 syntax_error(EMSG_DIV_BY_ZERO);
3012 L_d -= (long long)(L_d / R_d) * R_d;
3015 debug_printf_eval("BINARY/REPLACE result:%f\n", L_d);
3016 res = setvar_i(((opinfo & OPCLSMASK) == OC_BINARY) ? res : L.v, L_d);
3020 case XC( OC_COMPARE ): {
3021 int i = i; /* for compiler */
3024 if (is_numeric(L.v) && is_numeric(R.v)) {
3025 Ld = getvar_i(L.v) - getvar_i(R.v);
3027 const char *l = getvar_s(L.v);
3028 const char *r = getvar_s(R.v);
3029 Ld = icase ? strcasecmp(l, r) : strcmp(l, r);
3031 switch (opn & 0xfe) {
3042 setvar_i(res, (i == 0) ^ (opn & 1));
3047 syntax_error(EMSG_POSSIBLE_ERROR);
3050 if ((opinfo & OPCLSMASK) <= SHIFT_TIL_THIS)
3052 if ((opinfo & OPCLSMASK) >= RECUR_FROM_THIS)
3059 debug_printf_eval("returning from %s(): %p\n", __func__, res);
3067 /* -------- main & co. -------- */
3069 static int awk_exit(int r)
3080 evaluate(endseq.first, &tv);
3083 /* waiting for children */
3084 for (i = 0; i < fdhash->csize; i++) {
3085 hi = fdhash->items[i];
3087 if (hi->data.rs.F && hi->data.rs.is_pipe)
3088 pclose(hi->data.rs.F);
3096 /* if expr looks like "var=value", perform assignment and return 1,
3097 * otherwise return 0 */
3098 static int is_assignment(const char *expr)
3102 if (!isalnum_(*expr) || (val = strchr(expr, '=')) == NULL) {
3106 exprc = xstrdup(expr);
3107 val = exprc + (val - expr);
3110 unescape_string_in_place(val);
3111 setvar_u(newvar(exprc), val);
3116 /* switch to next input file */
3117 static rstream *next_input_file(void)
3119 #define rsm (G.next_input_file__rsm)
3120 #define files_happen (G.next_input_file__files_happen)
3123 const char *fname, *ind;
3128 rsm.pos = rsm.adv = 0;
3131 if (getvar_i(intvar[ARGIND])+1 >= getvar_i(intvar[ARGC])) {
3138 ind = getvar_s(incvar(intvar[ARGIND]));
3139 fname = getvar_s(findvar(iamarray(intvar[ARGV]), ind));
3140 if (fname && *fname && !is_assignment(fname)) {
3141 F = xfopen_stdin(fname);
3146 files_happen = TRUE;
3147 setvar_s(intvar[FILENAME], fname);
3154 int awk_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
3155 int awk_main(int argc UNUSED_PARAM, char **argv)
3159 llist_t *list_v = NULL;
3160 llist_t *list_f = NULL;
3161 #if ENABLE_FEATURE_AWK_GNU_EXTENSIONS
3162 llist_t *list_e = NULL;
3168 char *vnames = (char *)vNames; /* cheat */
3169 char *vvalues = (char *)vValues;
3173 /* Undo busybox.c, or else strtod may eat ','! This breaks parsing:
3174 * $1,$2 == '$1,' '$2', NOT '$1' ',' '$2' */
3175 if (ENABLE_LOCALE_SUPPORT)
3176 setlocale(LC_NUMERIC, "C");
3180 /* allocate global buffer */
3181 g_buf = xmalloc(MAXVARFMT + 1);
3183 vhash = hash_init();
3184 ahash = hash_init();
3185 fdhash = hash_init();
3186 fnhash = hash_init();
3188 /* initialize variables */
3189 for (i = 0; *vnames; i++) {
3190 intvar[i] = v = newvar(nextword(&vnames));
3191 if (*vvalues != '\377')
3192 setvar_s(v, nextword(&vvalues));
3196 if (*vnames == '*') {
3197 v->type |= VF_SPECIAL;
3202 handle_special(intvar[FS]);
3203 handle_special(intvar[RS]);
3205 newfile("/dev/stdin")->F = stdin;
3206 newfile("/dev/stdout")->F = stdout;
3207 newfile("/dev/stderr")->F = stderr;
3209 /* Huh, people report that sometimes environ is NULL. Oh well. */
3210 if (environ) for (envp = environ; *envp; envp++) {
3211 /* environ is writable, thus we don't strdup it needlessly */
3213 char *s1 = strchr(s, '=');
3216 /* Both findvar and setvar_u take const char*
3217 * as 2nd arg -> environment is not trashed */
3218 setvar_u(findvar(iamarray(intvar[ENVIRON]), s), s1 + 1);
3222 opt = getopt32(argv, OPTSTR_AWK, &opt_F, &list_v, &list_f, IF_FEATURE_AWK_GNU_EXTENSIONS(&list_e,) NULL);
3226 bb_error_msg("warning: option -W is ignored");
3228 unescape_string_in_place(opt_F);
3229 setvar_s(intvar[FS], opt_F);
3232 if (!is_assignment(llist_pop(&list_v)))
3239 g_progname = llist_pop(&list_f);
3240 from_file = xfopen_stdin(g_progname);
3241 /* one byte is reserved for some trick in next_token */
3242 for (i = j = 1; j > 0; i += j) {
3243 s = xrealloc(s, i + 4096);
3244 j = fread(s + i, 1, 4094, from_file);
3248 parse_program(s + 1);
3251 g_progname = "cmd. line";
3252 #if ENABLE_FEATURE_AWK_GNU_EXTENSIONS
3254 parse_program(llist_pop(&list_e));
3257 if (!(opt & (OPT_f | OPT_e))) {
3260 parse_program(*argv++);
3263 /* fill in ARGV array */
3264 setari_u(intvar[ARGV], 0, "awk");
3267 setari_u(intvar[ARGV], ++i, *argv++);
3268 setvar_i(intvar[ARGC], i + 1);
3270 evaluate(beginseq.first, &tv);
3271 if (!mainseq.first && !endseq.first)
3272 awk_exit(EXIT_SUCCESS);
3274 /* input file could already be opened in BEGIN block */
3276 iF = next_input_file();
3278 /* passing through input files */
3281 setvar_i(intvar[FNR], 0);
3283 while ((i = awk_getline(iF, intvar[F0])) > 0) {
3286 incvar(intvar[FNR]);
3287 evaluate(mainseq.first, &tv);
3294 syntax_error(strerror(errno));
3296 iF = next_input_file();
3299 awk_exit(EXIT_SUCCESS);