sbase/expr.c

285 lines
6.9 KiB
C
Raw Normal View History

/* See LICENSE file for copyright and license details. */
#include <inttypes.h>
#include <stdio.h>
#include <string.h>
2015-02-26 04:12:14 +00:00
#include "utf.h"
#include "util.h"
2015-02-26 04:12:14 +00:00
/* token types for lexing/parsing
* single character operators represent themselves */
enum {
VAL = CHAR_MAX + 1, GE, LE, NE
};
typedef struct {
2015-02-26 04:12:14 +00:00
char *s; /* iff s is NULL, Val is an integer */
intmax_t n;
} Val;
static size_t intlen;
static void
enan(Val v)
{
if (v.s)
enprintf(2, "syntax error: expected integer got `%s'\n", v.s);
}
static void
ezero(intmax_t n)
{
2014-11-16 19:29:29 +00:00
if (n == 0)
enprintf(2, "division by zero\n");
}
2015-02-26 04:12:14 +00:00
static char *
valstr(Val val, char *buf, size_t bufsiz)
{
if (val.s)
return val.s;
snprintf(buf, bufsiz, "%"PRIdMAX, val.n);
return buf;
}
static int
valcmp(Val a, Val b)
{
char buf1[intlen], buf2[intlen];
char *astr = valstr(a, buf1, sizeof(buf1));
char *bstr = valstr(b, buf2, sizeof(buf2));
if (!a.s && !b.s)
return (a.n > b.n) - (a.n < b.n);
return strcmp(astr, bstr);
}
/* match vstr against BRE vregx (treat both values as strings)
* if there is at least one subexpression \(...\)
* then return the text matched by it \1 (empty string for no match)
* else return number of characters matched (0 for no match)
*/
static Val
match(Val vstr, Val vregx)
{
regex_t re;
regmatch_t matches[2];
char buf1[intlen], buf2[intlen];
char *str = valstr(vstr, buf1, sizeof(buf1));
char *regx = valstr(vregx, buf2, sizeof(buf2));;
char anchreg[strlen(regx) + 2];
/* expr(1p) "all patterns are anchored to the beginning of the string" */
snprintf(anchreg, sizeof(anchreg), "^%s", regx);
enregcomp(3, &re, anchreg, 0);
if (regexec(&re, str, 2, matches, 0)) {
regfree(&re);
return (Val){ (re.re_nsub ? "" : NULL), 0 };
}
if (re.re_nsub) {
intmax_t d;
char *s = str + matches[1].rm_so, *p = str + matches[1].rm_eo;
regfree(&re);
*p = '\0';
d = strtoimax(s, &p, 10);
if (*s && !*p) /* string matched by subexpression is an integer */
return (Val){ NULL, d };
/* FIXME? string is never free()d, worth fixing?
* need to allocate as it could be in buf1 instead of vstr.s */
return (Val){ enstrdup(3, s), 0 };
}
regfree(&re);
str += matches[0].rm_so;
return (Val){ NULL, utfnlen(str, matches[0].rm_eo - matches[0].rm_so) };
}
/* ops points to a stack of operators, opp points to one past the last op
* vals points to a stack of values , valp points to one past the last val
* guaranteed that opp != ops
* ops is unused here, but still included for parity with vals
* pop operator, pop two values, apply operator, push result
*/
2014-10-16 11:29:02 +00:00
static void
2015-02-26 04:12:14 +00:00
doop(int *ops, int **opp, Val *vals, Val **valp)
{
Val ret, a, b;
2015-02-26 04:12:14 +00:00
int op;
/* For an operation, we need a valid operator
* and two values on the stack */
if ((*opp)[-1] == '(')
enprintf(2, "syntax error: extra (\n");
2015-02-26 04:12:14 +00:00
if (*valp - vals < 2)
enprintf(2, "syntax error: missing expression or extra operator\n");
a = (*valp)[-2];
b = (*valp)[-1];
2015-02-26 04:12:14 +00:00
op = (*opp)[-1];
2015-02-26 04:12:14 +00:00
switch (op) {
case '|':
2015-02-26 04:12:14 +00:00
if ( a.s && *a.s) ret = (Val){ a.s , 0 };
else if (!a.s && a.n) ret = (Val){ NULL, a.n };
else if ( b.s && *b.s) ret = (Val){ b.s , 0 };
else ret = (Val){ NULL, b.n };
break;
case '&':
2015-02-26 04:12:14 +00:00
if (((a.s && *a.s) || a.n) && ((b.s && *b.s) || b.n))
ret = a;
2015-02-26 04:12:14 +00:00
else
ret = (Val){ NULL, 0 };
break;
2015-02-26 04:12:14 +00:00
case '=': ret = (Val){ NULL, valcmp(a, b) == 0 }; break;
case '>': ret = (Val){ NULL, valcmp(a, b) > 0 }; break;
2015-02-26 04:12:14 +00:00
case GE : ret = (Val){ NULL, valcmp(a, b) >= 0 }; break;
case '<': ret = (Val){ NULL, valcmp(a, b) < 0 }; break;
2015-02-26 04:12:14 +00:00
case LE : ret = (Val){ NULL, valcmp(a, b) <= 0 }; break;
case NE : ret = (Val){ NULL, valcmp(a, b) != 0 }; break;
2015-02-26 04:12:14 +00:00
case '+': enan(a); enan(b); ret = (Val){ NULL, a.n + b.n }; break;
case '-': enan(a); enan(b); ret = (Val){ NULL, a.n - b.n }; break;
case '*': enan(a); enan(b); ret = (Val){ NULL, a.n * b.n }; break;
case '/': enan(a); enan(b); ezero(b.n); ret = (Val){ NULL, a.n / b.n }; break;
case '%': enan(a); enan(b); ezero(b.n); ret = (Val){ NULL, a.n % b.n }; break;
case ':': ret = match(a, b); break;
}
(*valp)[-2] = ret;
(*opp)--;
(*valp)--;
}
2015-02-26 04:12:14 +00:00
/* retrn the type of the next token, s
* if it is a value, place the value in v for use by parser
*/
static int
2015-02-26 04:12:14 +00:00
lex(char *s, Val *v)
{
intmax_t d;
2015-02-26 04:12:14 +00:00
char *p, *ops = "|&=><+-*/%():";
/* clean integer */
2015-02-26 04:12:14 +00:00
d = strtoimax(s, &p, 10);
if (*s && !*p) {
*v = (Val){ NULL, d };
return VAL;
}
/* one-char operand */
2015-02-26 04:12:14 +00:00
if (*s && !s[1] && strchr(ops, *s))
return *s;
/* two-char operand */
2015-02-26 04:12:14 +00:00
if (!strcmp(s, ">=")) return GE;
if (!strcmp(s, "<=")) return LE;
if (!strcmp(s, "!=")) return NE;
/* nothing matched, treat as string */
2015-02-26 04:12:14 +00:00
*v = (Val){ s, 0 };
return VAL;
}
2015-02-26 04:12:14 +00:00
/* using shunting-yard to convert from infix to rpn
* https://en.wikipedia.org/wiki/Shunting-yard_algorithm
* instead of creating rpn output to evaluate later, evaluate it immediately as
* it is created
* vals is the value stack, valp points to one past last value on the stack
* ops is the operator stack, opp points to one past last op on the stack
*/
static int
2014-11-16 19:29:29 +00:00
parse(char **expr, int exprlen)
{
2015-02-26 04:12:14 +00:00
Val vals[exprlen], *valp = vals;
int ops[exprlen], *opp = ops;
int i, lasttype = 0;
char prec[] = { /* precedence of operators */
['|'] = 1,
['&'] = 2,
['='] = 3, ['>'] = 3, [GE] = 3, ['<'] = 3, [LE] = 3, [NE] = 3,
['+'] = 4, ['-'] = 4,
['*'] = 5, ['/'] = 5, ['%'] = 5,
[':'] = 6,
};
2014-11-16 19:29:29 +00:00
for (i = 0; i < exprlen; i++) {
2015-02-26 04:12:14 +00:00
Val v;
int type = lex(expr[i], &v);
2014-11-16 19:29:29 +00:00
switch (type) {
2014-11-16 19:29:29 +00:00
case VAL:
2015-02-26 04:12:14 +00:00
*valp++ = v;
2014-11-16 19:29:29 +00:00
break;
case '(':
*opp++ = '(';
break;
case ')':
2014-11-16 19:29:29 +00:00
if (lasttype == '(')
enprintf(2, "syntax error: empty ( )\n");
2015-02-26 04:12:14 +00:00
while (opp > ops && opp[-1] != '(')
doop(ops, &opp, vals, &valp);
if (opp == ops)
enprintf(2, "syntax error: extra )\n");
2014-11-16 19:29:29 +00:00
opp--;
break;
2015-02-26 04:12:14 +00:00
default: /* operator */
2014-11-16 19:29:29 +00:00
if (prec[lasttype])
enprintf(2, "syntax error: extra operator\n");
2015-02-26 04:12:14 +00:00
while (opp > ops && prec[opp[-1]] >= prec[type])
doop(ops, &opp, vals, &valp);
2014-11-16 19:29:29 +00:00
*opp++ = type;
break;
}
2014-11-16 19:29:29 +00:00
lasttype = type;
}
2015-02-26 04:12:14 +00:00
while (opp > ops)
doop(ops, &opp, vals, &valp);
2015-02-26 04:12:14 +00:00
if (valp == vals)
enprintf(2, "syntax error: missing expression\n");
2015-02-26 04:12:14 +00:00
if (--valp != vals)
enprintf(2, "syntax error: extra expression\n");
2014-11-16 19:29:29 +00:00
if (valp->s)
printf("%s\n", valp->s);
else
2014-11-16 19:29:29 +00:00
printf("%"PRIdMAX"\n", valp->n);
2014-11-16 19:29:29 +00:00
return (valp->s && *valp->s) || valp->n;
}
2015-02-26 04:12:14 +00:00
/* the only way to get usage() is if the user didn't supply -- and expression
* begins with a -
* expr(1p): "... the conforming application must employ the -- construct ...
* if there is any chance the first operand might be a negative integer (or any
* string with a leading minus"
*/
2014-11-16 19:29:29 +00:00
static void
usage(void)
{
2015-02-26 04:12:14 +00:00
enprintf(3, "usage: %s [--] expression\n"
"note : the -- is mandatory if expression begins with a -\n", argv0);
}
int
2014-11-16 19:29:29 +00:00
main(int argc, char *argv[])
{
2014-11-16 19:29:29 +00:00
intmax_t n = INTMAX_MIN;
/* Get the maximum number of digits (+ sign) */
2015-02-26 04:12:14 +00:00
for (intlen = (n < 0); n; n /= 10, ++intlen)
;
2014-11-16 19:29:29 +00:00
ARGBEGIN {
2015-02-26 18:07:02 +00:00
default:
usage();
2014-11-16 19:29:29 +00:00
} ARGEND;
2014-11-16 19:29:29 +00:00
return !parse(argv, argc);
}