static EC_OBJ EcLibRe_Compile( EC_OBJ stack, EcAny userdata )
{
const char *pattern;
EcInt patlen;
EC_OBJ opts = EC_NIL;
int options;
EcBool study = FALSE;
pcre *code;
pcre_extra *extra = NULL;
const char *errptr = NULL;
int erroffs = 0;
EC_OBJ res;
res = EcParseStackFunction( "re.compile", TRUE, stack, "s#|Ob", &pattern, &patlen, &opts, &study );
if (EC_ERRORP(res)) return res;
opts = obj2options( "re.compile", 2, opts, &options );
if (EC_ERRORP(opts)) return opts;
code = pcre_compile( pattern, options,
&errptr, &erroffs,
/* tableptr */ NULL );
if (! code) return EcReError( errptr, erroffs );
if (study)
{
extra = pcre_study( code,
0, /* currently pcre has no options for study */
&errptr );
if (! extra) return EcReError( errptr, -1 );
}
return make_regexp( code, extra, pattern, patlen, options );
}
struct regexp *
regexp_concat_n(struct info *info, int n, struct regexp **r) {
size_t len = 0;
char *pat, *p;
for (int i=0; i < n; i++)
if (r[i] != NULL)
len += strlen(r[i]->pattern->str) + strlen("()");
if (len == 0)
return NULL;
if (ALLOC_N(pat, len+1) < 0)
return NULL;
p = pat;
for (int i=0; i < n; i++) {
if (r[i] == NULL)
continue;
*p++ = '(';
p = stpcpy(p, r[i]->pattern->str);
*p++ = ')';
}
return make_regexp(info, pat);
}
struct regexp *
make_regexp_unescape(struct info *info, const char *pat, int nocase) {
char *p = unescape(pat, strlen(pat), NULL);
if (p == NULL)
return NULL;
return make_regexp(info, p, nocase);
}
void rule_for_word(char *word, char *token)
{
char regexp[STRING_LENGTH];
strcpy(regexp, word);
make_regexp(regexp);
printf("%s {\n yylval.str = newstr(yytext); return %s;\n}\n", regexp, token);
}
struct regexp *
regexp_concat(struct info *info, struct regexp *r1, struct regexp *r2) {
const char *p1 = r1->pattern->str;
const char *p2 = r2->pattern->str;
char *s;
if (asprintf(&s, "(%s)(%s)", p1, p2) == -1)
return NULL;
return make_regexp(info, s);
}
struct regexp *
regexp_maybe(struct info *info, struct regexp *r) {
const char *p;
char *s;
int ret;
if (r == NULL)
return NULL;
p = r->pattern->str;
ret = asprintf(&s, "(%s)?", p);
return (ret == -1) ? NULL : make_regexp(info, s, r->nocase);
}
struct regexp *
regexp_union_n(struct info *info, int n, struct regexp **r) {
size_t len = 0;
char *pat = NULL, *p, *expanded = NULL;
int nnocase = 0, npresent = 0;
int ret;
for (int i=0; i < n; i++)
if (r[i] != NULL) {
len += strlen(r[i]->pattern->str) + strlen("()|");
npresent += 1;
if (r[i]->nocase)
nnocase += 1;
}
bool mixedcase = nnocase > 0 && nnocase < npresent;
if (len == 0)
return NULL;
if (ALLOC_N(pat, len) < 0)
return NULL;
p = pat;
int added = 0;
for (int i=0; i < n; i++) {
if (r[i] == NULL)
continue;
if (added > 0)
*p++ = '|';
*p++ = '(';
if (mixedcase && r[i]->nocase) {
expanded = regexp_expand_nocase(r[i]);
ERR_BAIL(r[i]->info);
len += strlen(expanded) - strlen(r[i]->pattern->str);
ret = REALLOC_N(pat, len);
ERR_NOMEM(ret < 0, info);
p = pat + strlen(pat);
p = stpcpy(p, expanded);
FREE(expanded);
} else {
p = stpcpy(p, r[i]->pattern->str);
}
*p++ = ')';
added += 1;
}
*p = '\0';
return make_regexp(info, pat, nnocase == npresent);
error:
FREE(expanded);
FREE(pat);
return NULL;
}
static EC_OBJ regexp_copy( EC_OBJ obj, EcCopyType type )
{
pcre *code;
pcre_extra *extra;
EC_OBJ src, sobj;
EC_OBJ res;
const char *old_str;
EcInt old_slen;
int old_options;
const char *errptr = NULL;
int erroffs = 0;
/* compile a new regexp with the same string and options of the old one */
src = EC_REGEXPSRC(obj);
ASSERT( EC_ARRAYP(src) );
ASSERT( EC_ARRAYLEN(src) == 2 );
sobj = EcArrayGet( src, 0 );
res = EcArrayGet( src, 1 );
ASSERT( EC_STRINGP(sobj) );
ASSERT( EC_INUMP(res) );
old_str = EC_STRDATA(sobj);
old_slen = EC_STRLEN(sobj);
old_options = EC_INUM(res);
code = pcre_compile( old_str, old_options,
&errptr, &erroffs,
/* tableptr */ NULL );
if (! code) return EcReError( errptr, erroffs );
extra = NULL;
if (EC_PCREXTRA(obj))
{
extra = pcre_study( code,
0, /* currently pcre has no options for study */
&errptr );
if (! extra) return EcReError( errptr, -1 );
}
return make_regexp( code, extra,
old_str,
old_slen,
old_options );
}
struct regexp *
regexp_minus(struct info *info, struct regexp *r1, struct regexp *r2) {
const char *p1 = r1->pattern->str;
const char *p2 = r2->pattern->str;
struct regexp *result = NULL;
struct fa *fa = NULL, *fa1 = NULL, *fa2 = NULL;
int r;
char *s = NULL;
size_t s_len;
r = fa_compile(p1, strlen(p1), &fa1);
if (r != REG_NOERROR)
goto error;
r = fa_compile(p2, strlen(p2), &fa2);
if (r != REG_NOERROR)
goto error;
fa = fa_minus(fa1, fa2);
if (fa == NULL)
goto error;
r = fa_as_regexp(fa, &s, &s_len);
if (r < 0)
goto error;
if (s == NULL) {
/* FA is the empty set, which we can't represent as a regexp */
goto error;
}
result = make_regexp(info, s);
s = NULL;
done:
fa_free(fa);
fa_free(fa1);
fa_free(fa2);
free(s);
return result;
error:
unref(result, regexp);
goto done;
}
struct regexp *
regexp_minus(struct info *info, struct regexp *r1, struct regexp *r2) {
struct regexp *result = NULL;
struct fa *fa = NULL, *fa1 = NULL, *fa2 = NULL;
int r;
char *s = NULL;
size_t s_len;
fa1 = regexp_to_fa(r1);
ERR_BAIL(r1->info);
fa2 = regexp_to_fa(r2);
ERR_BAIL(r2->info);
fa = fa_minus(fa1, fa2);
if (fa == NULL)
goto error;
r = fa_as_regexp(fa, &s, &s_len);
if (r < 0)
goto error;
if (s == NULL) {
/* FA is the empty set, which we can't represent as a regexp */
goto error;
}
if (regexp_c_locale(&s, NULL) < 0)
goto error;
result = make_regexp(info, s, fa_is_nocase(fa));
s = NULL;
done:
fa_free(fa);
fa_free(fa1);
fa_free(fa2);
free(s);
return result;
error:
unref(result, regexp);
goto done;
}
struct regexp *
regexp_concat_n(struct info *info, int n, struct regexp **r) {
size_t len = 0;
char *pat = NULL, *p, *expanded = NULL;
int nnocase = 0, npresent = 0;
for (int i=0; i < n; i++)
if (r[i] != NULL) {
len += strlen(r[i]->pattern->str) + strlen("()");
npresent += 1;
if (r[i]->nocase)
nnocase += 1;
}
bool mixedcase = nnocase > 0 && nnocase < npresent;
if (len == 0)
return NULL;
len += 1;
if (ALLOC_N(pat, len) < 0)
return NULL;
p = pat;
for (int i=0; i < n; i++) {
if (r[i] == NULL)
continue;
*p++ = '(';
if (mixedcase && r[i]->nocase) {
p = append_expanded(r[i], &pat, p, &len);
ERR_BAIL(r[i]->info);
} else {
p = stpcpy(p, r[i]->pattern->str);
}
*p++ = ')';
}
*p = '\0';
return make_regexp(info, pat, nnocase == npresent);
error:
FREE(expanded);
FREE(pat);
return NULL;
}
struct regexp *
regexp_iter(struct info *info, struct regexp *r, int min, int max) {
const char *p;
char *s;
int ret = 0;
if (r == NULL)
return NULL;
p = r->pattern->str;
if ((min == 0 || min == 1) && max == -1) {
char q = (min == 0) ? '*' : '+';
ret = asprintf(&s, "(%s)%c", p, q);
} else if (min == max) {
ret = asprintf(&s, "(%s){%d}", p, min);
} else {
ret = asprintf(&s, "(%s){%d,%d}", p, min, max);
}
return (ret == -1) ? NULL : make_regexp(info, s, r->nocase);
}
struct regexp *make_regexp_literal(struct info *info, const char *text) {
char *pattern, *p;
/* Escape special characters in text since it should be taken
literally */
CALLOC(pattern, 2*strlen(text)+1);
p = pattern;
for (const char *t = text; *t != '\0'; t++) {
if ((*t == '\\') && t[1]) {
*p++ = *t++;
*p++ = *t;
} else if (strchr(".|{}[]()+*?", *t) != NULL) {
*p++ = '\\';
*p++ = *t;
} else {
*p++ = *t;
}
}
return make_regexp(info, pattern, 0);
}
/* Take a POSIX glob and turn it into a regexp. The regexp is constructed
* by doing the following translations of characters in the string:
* * -> [^/]*
* ? -> [^/]
* leave characters escaped with a backslash alone
* escape any of ".|{}()+^$" with a backslash
*
* Note that that ignores some of the finer points of globs, like
* complementation.
*/
struct regexp *make_regexp_from_glob(struct info *info, const char *glob) {
static const char *const star = "[^/]*";
static const char *const qmark = "[^/]";
static const char *const special = ".|{}()+^$";
int newlen = strlen(glob);
char *pat = NULL;
for (const char *s = glob; *s; s++) {
if (*s == '\\' && *(s+1))
s += 1;
else if (*s == '*')
newlen += strlen(star)-1;
else if (*s == '?')
newlen += strlen(qmark)-1;
else if (strchr(special, *s) != NULL)
newlen += 1;
}
if (ALLOC_N(pat, newlen + 1) < 0)
return NULL;
char *t = pat;
for (const char *s = glob; *s; s++) {
if (*s == '\\' && *(s+1)) {
*t++ = *s++;
*t++ = *s;
} else if (*s == '*') {
t = stpcpy(t, star);
} else if (*s == '?') {
t = stpcpy(t, qmark);
} else if (strchr(special, *s) != NULL) {
*t++ = '\\';
*t++ = *s;
} else {
*t++ = *s;
}
}
return make_regexp(info, pat, 0);
}
Regexp *
re_update(NODE *t)
{
NODE *t1;
if ((t->re_flags & CASE) == IGNORECASE) {
if ((t->re_flags & CONST) != 0) {
assert(t->type == Node_regex);
return t->re_reg;
}
t1 = force_string(tree_eval(t->re_exp));
if (t->re_text != NULL) {
if (cmp_nodes(t->re_text, t1) == 0) {
free_temp(t1);
return t->re_reg;
}
unref(t->re_text);
}
t->re_text = dupnode(t1);
free_temp(t1);
}
if (t->re_reg != NULL)
refree(t->re_reg);
if (t->re_cnt > 0)
t->re_cnt++;
if (t->re_cnt > 10)
t->re_cnt = 0;
if (t->re_text == NULL || (t->re_flags & CASE) != IGNORECASE) {
t1 = force_string(tree_eval(t->re_exp));
unref(t->re_text);
t->re_text = dupnode(t1);
free_temp(t1);
}
t->re_reg = make_regexp(t->re_text->stptr, t->re_text->stlen,
IGNORECASE, t->re_cnt);
t->re_flags &= ~CASE;
t->re_flags |= IGNORECASE;
return t->re_reg;
}
