xref: /freebsd/bin/expr/expr.y (revision 3642298923e528d795e3a30ec165d2b469e28b40)
1 %{
2 /*-
3  * Written by Pace Willisson (pace@blitz.com)
4  * and placed in the public domain.
5  *
6  * Largely rewritten by J.T. Conklin (jtc@wimsey.com)
7  *
8  * $FreeBSD$
9  */
10 
11 #include <sys/types.h>
12 
13 #include <ctype.h>
14 #include <err.h>
15 #include <errno.h>
16 #include <inttypes.h>
17 #include <limits.h>
18 #include <locale.h>
19 #include <stdio.h>
20 #include <stdlib.h>
21 #include <string.h>
22 #include <regex.h>
23 #include <unistd.h>
24 
25 /*
26  * POSIX specifies a specific error code for syntax errors.  We exit
27  * with this code for all errors.
28  */
29 #define	ERR_EXIT	2
30 
31 enum valtype {
32 	integer, numeric_string, string
33 } ;
34 
35 struct val {
36 	enum valtype type;
37 	union {
38 		char *s;
39 		intmax_t i;
40 	} u;
41 } ;
42 
43 struct val *result;
44 
45 int		chk_div(intmax_t, intmax_t);
46 int		chk_minus(intmax_t, intmax_t, intmax_t);
47 int		chk_plus(intmax_t, intmax_t, intmax_t);
48 int		chk_times(intmax_t, intmax_t, intmax_t);
49 void		free_value(struct val *);
50 int		is_zero_or_null(struct val *);
51 int		isstring(struct val *);
52 struct val	*make_integer(intmax_t);
53 struct val	*make_str(const char *);
54 struct val	*op_and(struct val *, struct val *);
55 struct val	*op_colon(struct val *, struct val *);
56 struct val	*op_div(struct val *, struct val *);
57 struct val	*op_eq(struct val *, struct val *);
58 struct val	*op_ge(struct val *, struct val *);
59 struct val	*op_gt(struct val *, struct val *);
60 struct val	*op_le(struct val *, struct val *);
61 struct val	*op_lt(struct val *, struct val *);
62 struct val	*op_minus(struct val *, struct val *);
63 struct val	*op_ne(struct val *, struct val *);
64 struct val	*op_or(struct val *, struct val *);
65 struct val	*op_plus(struct val *, struct val *);
66 struct val	*op_rem(struct val *, struct val *);
67 struct val	*op_times(struct val *, struct val *);
68 intmax_t	to_integer(struct val *);
69 void		to_string(struct val *);
70 int		yyerror(const char *);
71 int		yylex(void);
72 int		yyparse(void);
73 
74 static int	eflag;
75 char **av;
76 %}
77 
78 %union
79 {
80 	struct val *val;
81 }
82 
83 %left <val> '|'
84 %left <val> '&'
85 %left <val> '=' '>' '<' GE LE NE
86 %left <val> '+' '-'
87 %left <val> '*' '/' '%'
88 %left <val> ':'
89 
90 %token <val> TOKEN
91 %type <val> start expr
92 
93 %%
94 
95 start: expr { result = $$; }
96 
97 expr:	TOKEN
98 	| '(' expr ')' { $$ = $2; }
99 	| expr '|' expr { $$ = op_or ($1, $3); }
100 	| expr '&' expr { $$ = op_and ($1, $3); }
101 	| expr '=' expr { $$ = op_eq ($1, $3); }
102 	| expr '>' expr { $$ = op_gt ($1, $3); }
103 	| expr '<' expr { $$ = op_lt ($1, $3); }
104 	| expr GE expr  { $$ = op_ge ($1, $3); }
105 	| expr LE expr  { $$ = op_le ($1, $3); }
106 	| expr NE expr  { $$ = op_ne ($1, $3); }
107 	| expr '+' expr { $$ = op_plus ($1, $3); }
108 	| expr '-' expr { $$ = op_minus ($1, $3); }
109 	| expr '*' expr { $$ = op_times ($1, $3); }
110 	| expr '/' expr { $$ = op_div ($1, $3); }
111 	| expr '%' expr { $$ = op_rem ($1, $3); }
112 	| expr ':' expr { $$ = op_colon ($1, $3); }
113 	;
114 
115 
116 %%
117 
118 struct val *
119 make_integer(intmax_t i)
120 {
121 	struct val *vp;
122 
123 	vp = (struct val *) malloc (sizeof (*vp));
124 	if (vp == NULL) {
125 		errx(ERR_EXIT, "malloc() failed");
126 	}
127 
128 	vp->type = integer;
129 	vp->u.i  = i;
130 	return vp;
131 }
132 
133 struct val *
134 make_str(const char *s)
135 {
136 	struct val *vp;
137 	char *ep;
138 
139 	vp = (struct val *) malloc (sizeof (*vp));
140 	if (vp == NULL || ((vp->u.s = strdup (s)) == NULL)) {
141 		errx(ERR_EXIT, "malloc() failed");
142 	}
143 
144 	/*
145 	 * Previously we tried to scan the string to see if it ``looked like''
146 	 * an integer (erroneously, as it happened).  Let strtoimax() do the
147 	 * dirty work.  We could cache the value, except that we are using
148 	 * a union and need to preserve the original string form until we
149 	 * are certain that it is not needed.
150 	 *
151 	 * IEEE Std.1003.1-2001 says:
152 	 * /integer/ An argument consisting only of an (optional) unary minus
153 	 *	     followed by digits.
154 	 *
155 	 * This means that arguments which consist of digits followed by
156 	 * non-digits MUST NOT be considered integers.  strtoimax() will
157 	 * figure this out for us.
158 	 */
159 	if (eflag)
160 		(void)strtoimax(s, &ep, 10);
161 	else
162 		(void)strtol(s, &ep, 10);
163 
164 	if (*ep != '\0')
165 		vp->type = string;
166 	else
167 		vp->type = numeric_string;
168 
169 	return vp;
170 }
171 
172 
173 void
174 free_value(struct val *vp)
175 {
176 	if (vp->type == string || vp->type == numeric_string)
177 		free (vp->u.s);
178 }
179 
180 
181 intmax_t
182 to_integer(struct val *vp)
183 {
184 	intmax_t i;
185 
186 	if (vp->type == integer)
187 		return 1;
188 
189 	if (vp->type == string)
190 		return 0;
191 
192 	/* vp->type == numeric_string, make it numeric */
193 	errno = 0;
194 	if (eflag) {
195 		i  = strtoimax(vp->u.s, (char **)NULL, 10);
196 		if (errno == ERANGE)
197 			err(ERR_EXIT, NULL);
198 	} else {
199 		i = strtol(vp->u.s, (char **)NULL, 10);
200 	}
201 
202 	free (vp->u.s);
203 	vp->u.i = i;
204 	vp->type = integer;
205 	return 1;
206 }
207 
208 void
209 to_string(struct val *vp)
210 {
211 	char *tmp;
212 
213 	if (vp->type == string || vp->type == numeric_string)
214 		return;
215 
216 	/*
217 	 * log_10(x) ~= 0.3 * log_2(x).  Rounding up gives the number
218 	 * of digits; add one each for the sign and terminating null
219 	 * character, respectively.
220 	 */
221 #define	NDIGITS(x) (3 * (sizeof(x) * CHAR_BIT) / 10 + 1 + 1 + 1)
222 	tmp = malloc(NDIGITS(vp->u.i));
223 	if (tmp == NULL)
224 		errx(ERR_EXIT, "malloc() failed");
225 
226 	sprintf(tmp, "%jd", vp->u.i);
227 	vp->type = string;
228 	vp->u.s  = tmp;
229 }
230 
231 
232 int
233 isstring(struct val *vp)
234 {
235 	/* only TRUE if this string is not a valid integer */
236 	return (vp->type == string);
237 }
238 
239 
240 int
241 yylex(void)
242 {
243 	char *p;
244 
245 	if (*av == NULL)
246 		return (0);
247 
248 	p = *av++;
249 
250 	if (strlen (p) == 1) {
251 		if (strchr ("|&=<>+-*/%:()", *p))
252 			return (*p);
253 	} else if (strlen (p) == 2 && p[1] == '=') {
254 		switch (*p) {
255 		case '>': return (GE);
256 		case '<': return (LE);
257 		case '!': return (NE);
258 		}
259 	}
260 
261 	yylval.val = make_str (p);
262 	return (TOKEN);
263 }
264 
265 int
266 is_zero_or_null(struct val *vp)
267 {
268 	if (vp->type == integer) {
269 		return (vp->u.i == 0);
270 	} else {
271 		return (*vp->u.s == 0 || (to_integer (vp) && vp->u.i == 0));
272 	}
273 	/* NOTREACHED */
274 }
275 
276 int
277 main(int argc, char *argv[])
278 {
279 	int c;
280 
281 	setlocale (LC_ALL, "");
282 	if (getenv("EXPR_COMPAT") != NULL
283 	    || check_utility_compat("expr")) {
284 		av = argv + 1;
285 		eflag = 1;
286 	} else {
287 		while ((c = getopt(argc, argv, "e")) != -1)
288 			switch (c) {
289 			case 'e':
290 				eflag = 1;
291 				break;
292 
293 			default:
294 				fprintf(stderr,
295 				    "usage: expr [-e] expression\n");
296 				exit(ERR_EXIT);
297 			}
298 		av = argv + optind;
299 	}
300 
301 	yyparse();
302 
303 	if (result->type == integer)
304 		printf("%jd\n", result->u.i);
305 	else
306 		printf("%s\n", result->u.s);
307 
308 	return (is_zero_or_null(result));
309 }
310 
311 int
312 yyerror(const char *s __unused)
313 {
314 	errx(ERR_EXIT, "syntax error");
315 }
316 
317 
318 struct val *
319 op_or(struct val *a, struct val *b)
320 {
321 	if (is_zero_or_null (a)) {
322 		free_value (a);
323 		return (b);
324 	} else {
325 		free_value (b);
326 		return (a);
327 	}
328 }
329 
330 struct val *
331 op_and(struct val *a, struct val *b)
332 {
333 	if (is_zero_or_null (a) || is_zero_or_null (b)) {
334 		free_value (a);
335 		free_value (b);
336 		return (make_integer ((intmax_t)0));
337 	} else {
338 		free_value (b);
339 		return (a);
340 	}
341 }
342 
343 struct val *
344 op_eq(struct val *a, struct val *b)
345 {
346 	struct val *r;
347 
348 	if (isstring (a) || isstring (b)) {
349 		to_string (a);
350 		to_string (b);
351 		r = make_integer ((intmax_t)(strcoll (a->u.s, b->u.s) == 0));
352 	} else {
353 		(void)to_integer(a);
354 		(void)to_integer(b);
355 		r = make_integer ((intmax_t)(a->u.i == b->u.i));
356 	}
357 
358 	free_value (a);
359 	free_value (b);
360 	return r;
361 }
362 
363 struct val *
364 op_gt(struct val *a, struct val *b)
365 {
366 	struct val *r;
367 
368 	if (isstring (a) || isstring (b)) {
369 		to_string (a);
370 		to_string (b);
371 		r = make_integer ((intmax_t)(strcoll (a->u.s, b->u.s) > 0));
372 	} else {
373 		(void)to_integer(a);
374 		(void)to_integer(b);
375 		r = make_integer ((intmax_t)(a->u.i > b->u.i));
376 	}
377 
378 	free_value (a);
379 	free_value (b);
380 	return r;
381 }
382 
383 struct val *
384 op_lt(struct val *a, struct val *b)
385 {
386 	struct val *r;
387 
388 	if (isstring (a) || isstring (b)) {
389 		to_string (a);
390 		to_string (b);
391 		r = make_integer ((intmax_t)(strcoll (a->u.s, b->u.s) < 0));
392 	} else {
393 		(void)to_integer(a);
394 		(void)to_integer(b);
395 		r = make_integer ((intmax_t)(a->u.i < b->u.i));
396 	}
397 
398 	free_value (a);
399 	free_value (b);
400 	return r;
401 }
402 
403 struct val *
404 op_ge(struct val *a, struct val *b)
405 {
406 	struct val *r;
407 
408 	if (isstring (a) || isstring (b)) {
409 		to_string (a);
410 		to_string (b);
411 		r = make_integer ((intmax_t)(strcoll (a->u.s, b->u.s) >= 0));
412 	} else {
413 		(void)to_integer(a);
414 		(void)to_integer(b);
415 		r = make_integer ((intmax_t)(a->u.i >= b->u.i));
416 	}
417 
418 	free_value (a);
419 	free_value (b);
420 	return r;
421 }
422 
423 struct val *
424 op_le(struct val *a, struct val *b)
425 {
426 	struct val *r;
427 
428 	if (isstring (a) || isstring (b)) {
429 		to_string (a);
430 		to_string (b);
431 		r = make_integer ((intmax_t)(strcoll (a->u.s, b->u.s) <= 0));
432 	} else {
433 		(void)to_integer(a);
434 		(void)to_integer(b);
435 		r = make_integer ((intmax_t)(a->u.i <= b->u.i));
436 	}
437 
438 	free_value (a);
439 	free_value (b);
440 	return r;
441 }
442 
443 struct val *
444 op_ne(struct val *a, struct val *b)
445 {
446 	struct val *r;
447 
448 	if (isstring (a) || isstring (b)) {
449 		to_string (a);
450 		to_string (b);
451 		r = make_integer ((intmax_t)(strcoll (a->u.s, b->u.s) != 0));
452 	} else {
453 		(void)to_integer(a);
454 		(void)to_integer(b);
455 		r = make_integer ((intmax_t)(a->u.i != b->u.i));
456 	}
457 
458 	free_value (a);
459 	free_value (b);
460 	return r;
461 }
462 
463 int
464 chk_plus(intmax_t a, intmax_t b, intmax_t r)
465 {
466 
467 	/* sum of two positive numbers must be positive */
468 	if (a > 0 && b > 0 && r <= 0)
469 		return 1;
470 	/* sum of two negative numbers must be negative */
471 	if (a < 0 && b < 0 && r >= 0)
472 		return 1;
473 	/* all other cases are OK */
474 	return 0;
475 }
476 
477 struct val *
478 op_plus(struct val *a, struct val *b)
479 {
480 	struct val *r;
481 
482 	if (!to_integer(a) || !to_integer(b)) {
483 		errx(ERR_EXIT, "non-numeric argument");
484 	}
485 
486 	if (eflag) {
487 		r = make_integer(a->u.i + b->u.i);
488 		if (chk_plus(a->u.i, b->u.i, r->u.i)) {
489 			errx(ERR_EXIT, "overflow");
490 		}
491 	} else
492 		r = make_integer((long)a->u.i + (long)b->u.i);
493 
494 	free_value (a);
495 	free_value (b);
496 	return r;
497 }
498 
499 int
500 chk_minus(intmax_t a, intmax_t b, intmax_t r)
501 {
502 
503 	/* special case subtraction of INTMAX_MIN */
504 	if (b == INTMAX_MIN) {
505 		if (a >= 0)
506 			return 1;
507 		else
508 			return 0;
509 	}
510 	/* this is allowed for b != INTMAX_MIN */
511 	return chk_plus (a, -b, r);
512 }
513 
514 struct val *
515 op_minus(struct val *a, struct val *b)
516 {
517 	struct val *r;
518 
519 	if (!to_integer(a) || !to_integer(b)) {
520 		errx(ERR_EXIT, "non-numeric argument");
521 	}
522 
523 	if (eflag) {
524 		r = make_integer(a->u.i - b->u.i);
525 		if (chk_minus(a->u.i, b->u.i, r->u.i)) {
526 			errx(ERR_EXIT, "overflow");
527 		}
528 	} else
529 		r = make_integer((long)a->u.i - (long)b->u.i);
530 
531 	free_value (a);
532 	free_value (b);
533 	return r;
534 }
535 
536 int
537 chk_times(intmax_t a, intmax_t b, intmax_t r)
538 {
539 	/* special case: first operand is 0, no overflow possible */
540 	if (a == 0)
541 		return 0;
542 	/* cerify that result of division matches second operand */
543 	if (r / a != b)
544 		return 1;
545 	return 0;
546 }
547 
548 struct val *
549 op_times(struct val *a, struct val *b)
550 {
551 	struct val *r;
552 
553 	if (!to_integer(a) || !to_integer(b)) {
554 		errx(ERR_EXIT, "non-numeric argument");
555 	}
556 
557 	if (eflag) {
558 		r = make_integer(a->u.i * b->u.i);
559 		if (chk_times(a->u.i, b->u.i, r->u.i)) {
560 			errx(ERR_EXIT, "overflow");
561 		}
562 	} else
563 		r = make_integer((long)a->u.i * (long)b->u.i);
564 
565 	free_value (a);
566 	free_value (b);
567 	return (r);
568 }
569 
570 int
571 chk_div(intmax_t a, intmax_t b)
572 {
573 	/* div by zero has been taken care of before */
574 	/* only INTMAX_MIN / -1 causes overflow */
575 	if (a == INTMAX_MIN && b == -1)
576 		return 1;
577 	/* everything else is OK */
578 	return 0;
579 }
580 
581 struct val *
582 op_div(struct val *a, struct val *b)
583 {
584 	struct val *r;
585 
586 	if (!to_integer(a) || !to_integer(b)) {
587 		errx(ERR_EXIT, "non-numeric argument");
588 	}
589 
590 	if (b->u.i == 0) {
591 		errx(ERR_EXIT, "division by zero");
592 	}
593 
594 	if (eflag) {
595 		r = make_integer(a->u.i / b->u.i);
596 		if (chk_div(a->u.i, b->u.i)) {
597 			errx(ERR_EXIT, "overflow");
598 		}
599 	} else
600 		r = make_integer((long)a->u.i / (long)b->u.i);
601 
602 	free_value (a);
603 	free_value (b);
604 	return r;
605 }
606 
607 struct val *
608 op_rem(struct val *a, struct val *b)
609 {
610 	struct val *r;
611 
612 	if (!to_integer(a) || !to_integer(b)) {
613 		errx(ERR_EXIT, "non-numeric argument");
614 	}
615 
616 	if (b->u.i == 0) {
617 		errx(ERR_EXIT, "division by zero");
618 	}
619 
620 	if (eflag)
621 		r = make_integer(a->u.i % b->u.i);
622 	        /* chk_rem necessary ??? */
623 	else
624 		r = make_integer((long)a->u.i % (long)b->u.i);
625 
626 	free_value (a);
627 	free_value (b);
628 	return r;
629 }
630 
631 struct val *
632 op_colon(struct val *a, struct val *b)
633 {
634 	regex_t rp;
635 	regmatch_t rm[2];
636 	char errbuf[256];
637 	int eval;
638 	struct val *v;
639 
640 	/* coerce to both arguments to strings */
641 	to_string(a);
642 	to_string(b);
643 
644 	/* compile regular expression */
645 	if ((eval = regcomp (&rp, b->u.s, 0)) != 0) {
646 		regerror (eval, &rp, errbuf, sizeof(errbuf));
647 		errx(ERR_EXIT, "%s", errbuf);
648 	}
649 
650 	/* compare string against pattern */
651 	/* remember that patterns are anchored to the beginning of the line */
652 	if (regexec(&rp, a->u.s, (size_t)2, rm, 0) == 0 && rm[0].rm_so == 0) {
653 		if (rm[1].rm_so >= 0) {
654 			*(a->u.s + rm[1].rm_eo) = '\0';
655 			v = make_str (a->u.s + rm[1].rm_so);
656 
657 		} else {
658 			v = make_integer ((intmax_t)(rm[0].rm_eo - rm[0].rm_so));
659 		}
660 	} else {
661 		if (rp.re_nsub == 0) {
662 			v = make_integer ((intmax_t)0);
663 		} else {
664 			v = make_str ("");
665 		}
666 	}
667 
668 	/* free arguments and pattern buffer */
669 	free_value (a);
670 	free_value (b);
671 	regfree (&rp);
672 
673 	return v;
674 }
675