xref: /netbsd/bin/ksh/expr.c (revision bf9ec67e)
1 /*	$NetBSD: expr.c,v 1.4 1999/10/20 15:49:15 hubertf Exp $	*/
2 
3 /*
4  * Korn expression evaluation
5  */
6 /*
7  * todo: better error handling: if in builtin, should be builtin error, etc.
8  */
9 
10 #include "sh.h"
11 #include <ctype.h>
12 
13 
14 /* The order of these enums is constrained by the order of opinfo[] */
15 enum token {
16 	/* some (long) unary operators */
17 	O_PLUSPLUS = 0, O_MINUSMINUS,
18 	/* binary operators */
19 	O_EQ, O_NE,
20 	/* assignments are assumed to be in range O_ASN .. O_BORASN */
21 	O_ASN, O_TIMESASN, O_DIVASN, O_MODASN, O_PLUSASN, O_MINUSASN,
22 	       O_LSHIFTASN, O_RSHIFTASN, O_BANDASN, O_BXORASN, O_BORASN,
23 	O_LSHIFT, O_RSHIFT,
24 	O_LE, O_GE, O_LT, O_GT,
25 	O_LAND,
26 	O_LOR,
27 	O_TIMES, O_DIV, O_MOD,
28 	O_PLUS, O_MINUS,
29 	O_BAND,
30 	O_BXOR,
31 	O_BOR,
32 	O_TERN,
33 	O_COMMA,
34 	/* things after this aren't used as binary operators */
35 	/* unary that are not also binaries */
36 	O_BNOT, O_LNOT,
37 	/* misc */
38 	OPEN_PAREN, CLOSE_PAREN, CTERN,
39 	/* things that don't appear in the opinfo[] table */
40 	VAR, LIT, END, BAD
41     };
42 #define IS_BINOP(op) (((int)op) >= (int)O_EQ && ((int)op) <= (int)O_COMMA)
43 #define IS_ASSIGNOP(op)	((int)(op) >= (int)O_ASN && (int)(op) <= (int)O_BORASN)
44 
45 enum prec {
46 	P_PRIMARY = 0,		/* VAR, LIT, (), ~ ! - + */
47 	P_MULT,			/* * / % */
48 	P_ADD,			/* + - */
49 	P_SHIFT,		/* << >> */
50 	P_RELATION,		/* < <= > >= */
51 	P_EQUALITY,		/* == != */
52 	P_BAND,			/* & */
53 	P_BXOR,			/* ^ */
54 	P_BOR,			/* | */
55 	P_LAND,			/* && */
56 	P_LOR,			/* || */
57 	P_TERN,			/* ?: */
58 	P_ASSIGN,		/* = *= /= %= += -= <<= >>= &= ^= |= */
59 	P_COMMA			/* , */
60     };
61 #define MAX_PREC	P_COMMA
62 
63 struct opinfo {
64 	char		name[4];
65 	int		len;	/* name length */
66 	enum prec	prec;	/* precidence: lower is higher */
67 };
68 
69 /* Tokens in this table must be ordered so the longest are first
70  * (eg, += before +).  If you change something, change the order
71  * of enum token too.
72  */
73 static const struct opinfo opinfo[] = {
74 		{ "++",	 2, P_PRIMARY },	/* before + */
75 		{ "--",	 2, P_PRIMARY },	/* before - */
76 		{ "==",	 2, P_EQUALITY },	/* before = */
77 		{ "!=",	 2, P_EQUALITY },	/* before ! */
78 		{ "=",	 1, P_ASSIGN },		/* keep assigns in a block */
79 		{ "*=",	 2, P_ASSIGN },
80 		{ "/=",	 2, P_ASSIGN },
81 		{ "%=",	 2, P_ASSIGN },
82 		{ "+=",	 2, P_ASSIGN },
83 		{ "-=",	 2, P_ASSIGN },
84 		{ "<<=", 3, P_ASSIGN },
85 		{ ">>=", 3, P_ASSIGN },
86 		{ "&=",	 2, P_ASSIGN },
87 		{ "^=",	 2, P_ASSIGN },
88 		{ "|=",	 2, P_ASSIGN },
89 		{ "<<",	 2, P_SHIFT },
90 		{ ">>",	 2, P_SHIFT },
91 		{ "<=",	 2, P_RELATION },
92 		{ ">=",	 2, P_RELATION },
93 		{ "<",	 1, P_RELATION },
94 		{ ">",	 1, P_RELATION },
95 		{ "&&",	 2, P_LAND },
96 		{ "||",	 2, P_LOR },
97 		{ "*",	 1, P_MULT },
98 		{ "/",	 1, P_MULT },
99 		{ "%",	 1, P_MULT },
100 		{ "+",	 1, P_ADD },
101 		{ "-",	 1, P_ADD },
102 		{ "&",	 1, P_BAND },
103 		{ "^",	 1, P_BXOR },
104 		{ "|",	 1, P_BOR },
105 		{ "?",	 1, P_TERN },
106 		{ ",",	 1, P_COMMA },
107 		{ "~",	 1, P_PRIMARY },
108 		{ "!",	 1, P_PRIMARY },
109 		{ "(",	 1, P_PRIMARY },
110 		{ ")",	 1, P_PRIMARY },
111 		{ ":",	 1, P_PRIMARY },
112 		{ "",	 0, P_PRIMARY } /* end of table */
113 	    };
114 
115 
116 typedef struct expr_state Expr_state;
117 struct expr_state {
118 	const char *expression;		/* expression being evaluated */
119 	const char *tokp;		/* lexical position */
120 	enum token  tok;		/* token from token() */
121 	int	    noassign;		/* don't do assigns (for ?:,&&,||) */
122 	struct tbl *val;		/* value from token() */
123 	struct tbl *evaling;		/* variable that is being recursively
124 					 * expanded (EXPRINEVAL flag set)
125 					 */
126 };
127 
128 enum error_type { ET_UNEXPECTED, ET_BADLIT, ET_RECURSIVE,
129 		  ET_LVALUE, ET_RDONLY, ET_STR };
130 
131 static void        evalerr  ARGS((Expr_state *es, enum error_type type,
132 				  const char *str)) GCC_FUNC_ATTR(noreturn);
133 static struct tbl *evalexpr ARGS((Expr_state *es, enum prec prec));
134 static void        token    ARGS((Expr_state *es));
135 static struct tbl *do_ppmm  ARGS((Expr_state *es, enum token op,
136 				  struct tbl *vasn, bool_t is_prefix));
137 static void	   assign_check ARGS((Expr_state *es, enum token op,
138 				      struct tbl *vasn));
139 static struct tbl *tempvar  ARGS((void));
140 static struct tbl *intvar   ARGS((Expr_state *es, struct tbl *vp));
141 
142 /*
143  * parse and evalute expression
144  */
145 int
146 evaluate(expr, rval, error_ok)
147 	const char *expr;
148 	long *rval;
149 	int error_ok;
150 {
151 	struct tbl v;
152 	int ret;
153 
154 	v.flag = DEFINED|INTEGER;
155 	v.type = 0;
156 	ret = v_evaluate(&v, expr, error_ok);
157 	*rval = v.val.i;
158 	return ret;
159 }
160 
161 /*
162  * parse and evalute expression, storing result in vp.
163  */
164 int
165 v_evaluate(vp, expr, error_ok)
166 	struct tbl *vp;
167 	const char *expr;
168 	volatile int error_ok;
169 {
170 	struct tbl *v;
171 	Expr_state curstate;
172 	Expr_state * const es = &curstate;
173 	int i;
174 
175 	/* save state to allow recursive calls */
176 	curstate.expression = curstate.tokp = expr;
177 	curstate.noassign = 0;
178 	curstate.evaling = (struct tbl *) 0;
179 
180 	newenv(E_ERRH);
181 	i = ksh_sigsetjmp(e->jbuf, 0);
182 	if (i) {
183 		/* Clear EXPRINEVAL in of any variables we were playing with */
184 		if (curstate.evaling)
185 			curstate.evaling->flag &= ~EXPRINEVAL;
186 		quitenv();
187 		if (i == LAEXPR) {
188 			if (error_ok == KSH_RETURN_ERROR)
189 				return 0;
190 			errorf(null);
191 		}
192 		unwind(i);
193 		/*NOTREACHED*/
194 	}
195 
196 	token(es);
197 #if 1 /* ifdef-out to disallow empty expressions to be treated as 0 */
198 	if (es->tok == END) {
199 		es->tok = LIT;
200 		es->val = tempvar();
201 	}
202 #endif /* 0 */
203 	v = intvar(es, evalexpr(es, MAX_PREC));
204 
205 	if (es->tok != END)
206 		evalerr(es, ET_UNEXPECTED, (char *) 0);
207 
208 	if (vp->flag & INTEGER)
209 		setint_v(vp, v);
210 	else
211 		/* can fail if readony */
212 		setstr(vp, str_val(v), error_ok);
213 
214 	quitenv();
215 
216 	return 1;
217 }
218 
219 static void
220 evalerr(es, type, str)
221 	Expr_state *es;
222 	enum error_type type;
223 	const char *str;
224 {
225 	char tbuf[2];
226 	const char *s;
227 
228 	switch (type) {
229 	case ET_UNEXPECTED:
230 		switch (es->tok) {
231 		case VAR:
232 			s = es->val->name;
233 			break;
234 		case LIT:
235 			s = str_val(es->val);
236 			break;
237 		case END:
238 			s = "end of expression";
239 			break;
240 		case BAD:
241 			tbuf[0] = *es->tokp;
242 			tbuf[1] = '\0';
243 			s = tbuf;
244 			break;
245 		default:
246 			s = opinfo[(int)es->tok].name;
247 		}
248 		warningf(TRUE, "%s: unexpected `%s'", es->expression, s);
249 		break;
250 
251 	case ET_BADLIT:
252 		warningf(TRUE, "%s: bad number `%s'", es->expression, str);
253 		break;
254 
255 	case ET_RECURSIVE:
256 		warningf(TRUE, "%s: expression recurses on parameter `%s'",
257 			es->expression, str);
258 		break;
259 
260 	case ET_LVALUE:
261 		warningf(TRUE, "%s: %s requires lvalue",
262 			es->expression, str);
263 		break;
264 
265 	case ET_RDONLY:
266 		warningf(TRUE, "%s: %s applied to read only variable",
267 			es->expression, str);
268 		break;
269 
270 	default: /* keep gcc happy */
271 	case ET_STR:
272 		warningf(TRUE, "%s: %s", es->expression, str);
273 		break;
274 	}
275 	unwind(LAEXPR);
276 }
277 
278 static struct tbl *
279 evalexpr(es, prec)
280 	Expr_state *es;
281 	enum prec prec;
282 {
283 	struct tbl *vl, UNINITIALIZED(*vr), *vasn;
284 	enum token op;
285 	long UNINITIALIZED(res);
286 
287 	if (prec == P_PRIMARY) {
288 		op = es->tok;
289 		if (op == O_BNOT || op == O_LNOT || op == O_MINUS
290 		    || op == O_PLUS)
291 		{
292 			token(es);
293 			vl = intvar(es, evalexpr(es, P_PRIMARY));
294 			if (op == O_BNOT)
295 				vl->val.i = ~vl->val.i;
296 			else if (op == O_LNOT)
297 				vl->val.i = !vl->val.i;
298 			else if (op == O_MINUS)
299 				vl->val.i = -vl->val.i;
300 			/* op == O_PLUS is a no-op */
301 		} else if (op == OPEN_PAREN) {
302 			token(es);
303 			vl = evalexpr(es, MAX_PREC);
304 			if (es->tok != CLOSE_PAREN)
305 				evalerr(es, ET_STR, "missing )");
306 			token(es);
307 		} else if (op == O_PLUSPLUS || op == O_MINUSMINUS) {
308 			token(es);
309 			vl = do_ppmm(es, op, es->val, TRUE);
310 			token(es);
311 		} else if (op == VAR || op == LIT) {
312 			vl = es->val;
313 			token(es);
314 		} else {
315 			evalerr(es, ET_UNEXPECTED, (char *) 0);
316 			/*NOTREACHED*/
317 		}
318 		if (es->tok == O_PLUSPLUS || es->tok == O_MINUSMINUS) {
319 			vl = do_ppmm(es, es->tok, vl, FALSE);
320 			token(es);
321 		}
322 		return vl;
323 	}
324 	vl = evalexpr(es, ((int) prec) - 1);
325 	for (op = es->tok; IS_BINOP(op) && opinfo[(int) op].prec == prec;
326 		op = es->tok)
327 	{
328 		token(es);
329 		vasn = vl;
330 		if (op != O_ASN) /* vl may not have a value yet */
331 			vl = intvar(es, vl);
332 		if (IS_ASSIGNOP(op)) {
333 			assign_check(es, op, vasn);
334 			vr = intvar(es, evalexpr(es, P_ASSIGN));
335 		} else if (op != O_TERN && op != O_LAND && op != O_LOR)
336 			vr = intvar(es, evalexpr(es, ((int) prec) - 1));
337 		if ((op == O_DIV || op == O_MOD || op == O_DIVASN
338 		     || op == O_MODASN) && vr->val.i == 0)
339 		{
340 			if (es->noassign)
341 				vr->val.i = 1;
342 			else
343 				evalerr(es, ET_STR, "zero divisor");
344 		}
345 		switch ((int) op) {
346 		case O_TIMES:
347 		case O_TIMESASN:
348 			res = vl->val.i * vr->val.i;
349 			break;
350 		case O_DIV:
351 		case O_DIVASN:
352 			res = vl->val.i / vr->val.i;
353 			break;
354 		case O_MOD:
355 		case O_MODASN:
356 			res = vl->val.i % vr->val.i;
357 			break;
358 		case O_PLUS:
359 		case O_PLUSASN:
360 			res = vl->val.i + vr->val.i;
361 			break;
362 		case O_MINUS:
363 		case O_MINUSASN:
364 			res = vl->val.i - vr->val.i;
365 			break;
366 		case O_LSHIFT:
367 		case O_LSHIFTASN:
368 			res = vl->val.i << vr->val.i;
369 			break;
370 		case O_RSHIFT:
371 		case O_RSHIFTASN:
372 			res = vl->val.i >> vr->val.i;
373 			break;
374 		case O_LT:
375 			res = vl->val.i < vr->val.i;
376 			break;
377 		case O_LE:
378 			res = vl->val.i <= vr->val.i;
379 			break;
380 		case O_GT:
381 			res = vl->val.i > vr->val.i;
382 			break;
383 		case O_GE:
384 			res = vl->val.i >= vr->val.i;
385 			break;
386 		case O_EQ:
387 			res = vl->val.i == vr->val.i;
388 			break;
389 		case O_NE:
390 			res = vl->val.i != vr->val.i;
391 			break;
392 		case O_BAND:
393 		case O_BANDASN:
394 			res = vl->val.i & vr->val.i;
395 			break;
396 		case O_BXOR:
397 		case O_BXORASN:
398 			res = vl->val.i ^ vr->val.i;
399 			break;
400 		case O_BOR:
401 		case O_BORASN:
402 			res = vl->val.i | vr->val.i;
403 			break;
404 		case O_LAND:
405 			if (!vl->val.i)
406 				es->noassign++;
407 			vr = intvar(es, evalexpr(es, ((int) prec) - 1));
408 			res = vl->val.i && vr->val.i;
409 			if (!vl->val.i)
410 				es->noassign--;
411 			break;
412 		case O_LOR:
413 			if (vl->val.i)
414 				es->noassign++;
415 			vr = intvar(es, evalexpr(es, ((int) prec) - 1));
416 			res = vl->val.i || vr->val.i;
417 			if (vl->val.i)
418 				es->noassign--;
419 			break;
420 		case O_TERN:
421 			{
422 				int e = vl->val.i != 0;
423 				if (!e)
424 					es->noassign++;
425 				vl = evalexpr(es, MAX_PREC);
426 				if (!e)
427 					es->noassign--;
428 				if (es->tok != CTERN)
429 					evalerr(es, ET_STR, "missing :");
430 				token(es);
431 				if (e)
432 					es->noassign++;
433 				vr = evalexpr(es, P_TERN);
434 				if (e)
435 					es->noassign--;
436 				vl = e ? vl : vr;
437 			}
438 			break;
439 		case O_ASN:
440 			res = vr->val.i;
441 			break;
442 		case O_COMMA:
443 			res = vr->val.i;
444 			break;
445 		}
446 		if (IS_ASSIGNOP(op)) {
447 			vr->val.i = res;
448 			if (vasn->flag & INTEGER)
449 				setint_v(vasn, vr);
450 			else
451 				setint(vasn, res);
452 			vl = vr;
453 		} else if (op != O_TERN)
454 			vl->val.i = res;
455 	}
456 	return vl;
457 }
458 
459 static void
460 token(es)
461 	Expr_state *es;
462 {
463 	const char *cp;
464 	int c;
465 	char *tvar;
466 
467 	/* skip white space */
468 	for (cp = es->tokp; (c = *cp), isspace(c); cp++)
469 		;
470 	es->tokp = cp;
471 
472 	if (c == '\0')
473 		es->tok = END;
474 	else if (letter(c)) {
475 		for (; letnum(c); c = *cp)
476 			cp++;
477 		if (c == '[') {
478 			int len;
479 
480 			len = array_ref_len(cp);
481 			if (len == 0)
482 				evalerr(es, ET_STR, "missing ]");
483 			cp += len;
484 		}
485 #ifdef KSH
486 		else if (c == '(' /*)*/ ) {
487 		    /* todo: add math functions (all take single argument):
488 		     * abs acos asin atan cos cosh exp int log sin sinh sqrt
489 		     * tan tanh
490 		     */
491 		    ;
492 		}
493 #endif /* KSH */
494 		if (es->noassign) {
495 			es->val = tempvar();
496 			es->val->flag |= EXPRLVALUE;
497 		} else {
498 			tvar = str_nsave(es->tokp, cp - es->tokp, ATEMP);
499 			es->val = global(tvar);
500 			afree(tvar, ATEMP);
501 		}
502 		es->tok = VAR;
503 	} else if (digit(c)) {
504 		for (; c != '_' && (letnum(c) || c == '#'); c = *cp++)
505 			;
506 		tvar = str_nsave(es->tokp, --cp - es->tokp, ATEMP);
507 		es->val = tempvar();
508 		es->val->flag &= ~INTEGER;
509 		es->val->type = 0;
510 		es->val->val.s = tvar;
511 		if (setint_v(es->val, es->val) == NULL)
512 			evalerr(es, ET_BADLIT, tvar);
513 		afree(tvar, ATEMP);
514 		es->tok = LIT;
515 	} else {
516 		int i, n0;
517 
518 		for (i = 0; (n0 = opinfo[i].name[0]); i++)
519 			if (c == n0
520 			    && strncmp(cp, opinfo[i].name, opinfo[i].len) == 0)
521 			{
522 				es->tok = (enum token) i;
523 				cp += opinfo[i].len;
524 				break;
525 			}
526 		if (!n0)
527 			es->tok = BAD;
528 	}
529 	es->tokp = cp;
530 }
531 
532 /* Do a ++ or -- operation */
533 static struct tbl *
534 do_ppmm(es, op, vasn, is_prefix)
535 	Expr_state *es;
536 	enum token op;
537 	struct tbl *vasn;
538 	bool_t is_prefix;
539 {
540 	struct tbl *vl;
541 	int oval;
542 
543 	assign_check(es, op, vasn);
544 
545 	vl = intvar(es, vasn);
546 	oval = op == O_PLUSPLUS ? vl->val.i++ : vl->val.i--;
547 	if (vasn->flag & INTEGER)
548 		setint_v(vasn, vl);
549 	else
550 		setint(vasn, vl->val.i);
551 	if (!is_prefix)		/* undo the inc/dec */
552 		vl->val.i = oval;
553 
554 	return vl;
555 }
556 
557 static void
558 assign_check(es, op, vasn)
559 	Expr_state *es;
560 	enum token op;
561 	struct tbl *vasn;
562 {
563 	if (vasn->name[0] == '\0' && !(vasn->flag & EXPRLVALUE))
564 		evalerr(es, ET_LVALUE, opinfo[(int) op].name);
565 	else if (vasn->flag & RDONLY)
566 		evalerr(es, ET_RDONLY, opinfo[(int) op].name);
567 }
568 
569 static struct tbl *
570 tempvar()
571 {
572 	register struct tbl *vp;
573 
574 	vp = (struct tbl*) alloc(sizeof(struct tbl), ATEMP);
575 	vp->flag = ISSET|INTEGER;
576 	vp->type = 0;
577 	vp->areap = ATEMP;
578 	vp->val.i = 0;
579 	vp->name[0] = '\0';
580 	return vp;
581 }
582 
583 /* cast (string) variable to temporary integer variable */
584 static struct tbl *
585 intvar(es, vp)
586 	Expr_state *es;
587 	struct tbl *vp;
588 {
589 	struct tbl *vq;
590 
591 	/* try to avoid replacing a temp var with another temp var */
592 	if (vp->name[0] == '\0'
593 	    && (vp->flag & (ISSET|INTEGER|EXPRLVALUE)) == (ISSET|INTEGER))
594 		return vp;
595 
596 	vq = tempvar();
597 	if (setint_v(vq, vp) == NULL) {
598 		if (vp->flag & EXPRINEVAL)
599 			evalerr(es, ET_RECURSIVE, vp->name);
600 		es->evaling = vp;
601 		vp->flag |= EXPRINEVAL;
602 		v_evaluate(vq, str_val(vp), KSH_UNWIND_ERROR);
603 		vp->flag &= ~EXPRINEVAL;
604 		es->evaling = (struct tbl *) 0;
605 	}
606 	return vq;
607 }
608