Changeset: c70c5c521c4d for MonetDB
URL: https://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=c70c5c521c4d
Modified Files:
monetdb5/mal/Tests/tst201.malC
monetdb5/modules/mal/pcre.c
Branch: Jun2020
Log Message:
Also don't use PCRE for patterns without _ not ending in %.
diffs (truncated from 366 to 300 lines):
diff --git a/monetdb5/mal/Tests/tst201.malC b/monetdb5/mal/Tests/tst201.malC
--- a/monetdb5/mal/Tests/tst201.malC
+++ b/monetdb5/mal/Tests/tst201.malC
@@ -61,14 +61,14 @@ exit MALException;
v45:= algebra.slice(b,v43,v44);
v46:= io.print(v45);
- v75:= algebra.likeselect(b, nil:bat[:oid], "ox", nil:str, false, false);
+ v75:= algebra.likeselect(b, nil:bat[:oid], "ox", "", false, false);
v76:= io.print(v75);
- v78:= algebra.likeselect(b, nil:bat[:oid], "fo", nil:str, false, false);
+ v78:= algebra.likeselect(b, nil:bat[:oid], "fo", "", false, false);
v79:= io.print(v78);
- v81:= algebra.likeselect(b, nil:bat[:oid], "fox", nil:str, false,
false);
+ v81:= algebra.likeselect(b, nil:bat[:oid], "fox", "", false, false);
v82:= io.print(v81);
- v84:= algebra.likeselect(b, nil:bat[:oid], " fox", nil:str, false,
false);
+ v84:= algebra.likeselect(b, nil:bat[:oid], " fox", "", false, false);
v85:= io.print(v84);
diff --git a/monetdb5/modules/mal/pcre.c b/monetdb5/modules/mal/pcre.c
--- a/monetdb5/modules/mal/pcre.c
+++ b/monetdb5/modules/mal/pcre.c
@@ -85,13 +85,14 @@ mal_export str ILIKEjoin(bat *r1, bat *r
mal_export str ILIKEjoin1(bat *r1, bat *r2, const bat *lid, const bat *rid,
const bat *slid, const bat *srid, const bit *nil_matches, const lng *estimate);
/* current implementation assumes simple %keyword% [keyw%]* */
-typedef struct RE {
+struct RE {
char *k;
uint32_t *w;
- bool search;
+ bool search:1,
+ atend:1;
size_t len;
struct RE *n;
-} RE;
+};
/* We cannot use strcasecmp and strncasecmp since they work byte for
* byte and don't deal with multibyte encodings (such as UTF-8).
@@ -233,32 +234,32 @@ myucslen(const uint32_t *ucs)
return i;
}
-static inline int
-mywstrncasecmp(const char *restrict s1, const uint32_t *restrict s2, size_t n2)
+static inline bool
+mywstrncaseeq(const char *restrict s1, const uint32_t *restrict s2, size_t n2,
bool atend)
{
uint32_t c1;
while (n2 > 0) {
size_t nn1 = utfc8touc(&c1, s1);
if (nn1 == 0 || nn1 == (size_t) -1)
- return -(*s2 != 0);
+ return (*s2 == 0);
if (*s2 == 0)
- return 1;
+ return false;
if (nn1 == (size_t) -1 || nn1 == (size_t) -2)
- return 0; /* actually an error that shouldn't
happen */
+ return true; /* actually an error that shouldn't
happen */
#if SIZEOF_WCHAR_T == 2
if (c1 > 0xFFFF || *s2 > 0xFFFF) {
if (c1 != *s2)
- return c1 - *s2;
+ return false;
} else
#endif
if (towlower((wint_t) c1) != towlower((wint_t) *s2))
- return towlower((wint_t) c1) - towlower((wint_t) *s2);
+ return false;
s1 += nn1;
n2--;
s2++;
}
- return 0;
+ return !atend || *s1 == 0;
}
static inline int
@@ -316,12 +317,12 @@ mywstrcasecmp(const char *restrict s1, c
}
static inline const char *
-mywstrcasestr(const char *restrict haystack, const uint32_t *restrict wneedle)
+mywstrcasestr(const char *restrict haystack, const uint32_t *restrict wneedle,
bool atend)
{
size_t nlen = myucslen(wneedle);
if (nlen == 0)
- return haystack;
+ return atend ? haystack + strlen(haystack) : haystack;
size_t hlen = strlen(haystack);
@@ -347,7 +348,7 @@ mywstrcasestr(const char *restrict hayst
break;
h += j;
}
- if (i == nlen)
+ if (i == nlen && (!atend || haystack[h] == 0))
return haystack;
haystack += step;
hlen -= step;
@@ -362,31 +363,26 @@ static bool
re_simple(const char *pat, unsigned char esc)
{
bool escaped = false;
- bool percatend = false;
if (pat == 0)
- return 0;
+ return false;
if (*pat == '%') {
- percatend = true;
pat++;
}
while (*pat) {
- percatend = false;
if (escaped) {
escaped = false;
} else if ((unsigned char) *pat == esc) {
escaped = true;
} else if (*pat == '_') {
- return 0;
- } else if (*pat == '%') {
- percatend = true;
+ return false;
}
pat++;
}
- return percatend;
+ return true;
}
-static bool
+static inline bool
is_strcmpable(const char *pat, const char *esc)
{
if (pat[strcspn(pat, "%_")])
@@ -395,15 +391,17 @@ is_strcmpable(const char *pat, const cha
}
static inline bool
-re_match_ignore(const char *s, RE *pattern)
+re_match_ignore(const char *restrict s, const struct RE *restrict pattern)
{
- RE *r;
+ const struct RE *r;
for (r = pattern; r; r = r->n) {
if (*r->w == 0 && (r->search || *s == 0))
return true;
if (!*s ||
- (r->search ? (s = mywstrcasestr(s, r->w)) == NULL :
mywstrncasecmp(s, r->w, r->len) != 0))
+ (r->search
+ ? (s = mywstrcasestr(s, r->w, r->atend)) == NULL
+ : !mywstrncaseeq(s, r->w, r->len, r->atend)))
return false;
s += r->len;
}
@@ -411,15 +409,22 @@ re_match_ignore(const char *s, RE *patte
}
static inline bool
-re_match_no_ignore(const char *s, RE *pattern)
+re_match_no_ignore(const char *restrict s, const struct RE *restrict pattern)
{
- RE *r;
+ const struct RE *r;
+ size_t l;
for (r = pattern; r; r = r->n) {
if (*r->k == 0 && (r->search || *s == 0))
return true;
if (!*s ||
- (r->search ? (s = strstr(s, r->k)) == NULL : strncmp(s,
r->k, r->len) != 0))
+ (r->search
+ ? (r->atend
+ ? (l = strlen(s)) < r->len || strcmp(s + l -
r->len, r->k) != 0
+ : (s = strstr(s, r->k)) == NULL)
+ : (r->atend
+ ? strcmp(s, r->k) != 0
+ : strncmp(s, r->k, r->len) != 0)))
return false;
s += r->len;
}
@@ -427,13 +432,13 @@ re_match_no_ignore(const char *s, RE *pa
}
static void
-re_destroy(RE *p)
+re_destroy(struct RE *p)
{
if (p) {
GDKfree(p->k);
GDKfree(p->w);
do {
- RE *n = p->n;
+ struct RE *n = p->n;
GDKfree(p);
p = n;
@@ -446,15 +451,15 @@ re_destroy(RE *p)
* fields in the first structure are allocated, whereas in all
* subsequent structures the fields point into the allocated buffer of
* the first. */
-static RE *
+static struct RE *
re_create(const char *pat, bool caseignore, uint32_t esc)
{
- RE *r = (RE*)GDKmalloc(sizeof(RE)), *n = r;
+ struct RE *r = GDKmalloc(sizeof(struct RE)), *n = r;
bool escaped = false;
if (r == NULL)
return NULL;
- *r = (struct RE) {.search = false};
+ *r = (struct RE) {.atend = true};
while (esc != '%' && *pat == '%') {
pat++; /* skip % */
@@ -473,25 +478,28 @@ re_create(const char *pat, bool caseigno
while (*wp) {
if (escaped) {
*wq++ = *wp;
+ n->len++;
escaped = false;
} else if (*wp == esc) {
escaped = true;
} else if (*wp == '%') {
- n->len = (size_t) (wq - n->w);
+ n->atend = false;
while (wp[1] == '%')
wp++;
if (wp[1]) {
- n = n->n = GDKmalloc(sizeof(RE));
+ n = n->n = GDKmalloc(sizeof(struct RE));
if (n == NULL)
goto bailout;
- *n = (struct RE) {.search = true, .w =
wp + 1};
+ *n = (struct RE) {.search = true,
.atend = true, .w = wp + 1};
}
*wq++ = 0;
} else {
*wq++ = *wp;
+ n->len++;
}
wp++;
}
+ *wq = 0;
} else {
char *p, *q;
if ((p = GDKstrdup(pat)) == NULL) {
@@ -503,25 +511,28 @@ re_create(const char *pat, bool caseigno
while (*p) {
if (escaped) {
*q++ = *p;
+ n->len++;
escaped = false;
} else if ((unsigned char) *p == esc) {
escaped = true;
} else if (*p == '%') {
- n->len = (size_t) (q - n->k);
+ n->atend = false;
while (p[1] == '%')
p++;
if (p[1]) {
- n = n->n = GDKmalloc(sizeof(RE));
+ n = n->n = GDKmalloc(sizeof(struct RE));
if (n == NULL)
goto bailout;
- *n = (struct RE) {.search = true, .k =
p + 1};
+ *n = (struct RE) {.search = true,
.atend = true, .k = p + 1};
}
*q++ = 0;
} else {
*q++ = *p;
+ n->len++;
}
p++;
}
+ *q = 0;
}
return r;
bailout:
@@ -713,7 +724,7 @@ re_likeselect(BAT **bnp, BAT *b, BAT *s,
BUN p, q;
oid o, off;
const char *v;
- RE *re = NULL;
+ struct RE *re = NULL;
uint32_t *wpat = NULL;
assert(ATOMstorage(b->ttype) == TYPE_str);
@@ -765,14 +776,14 @@ re_likeselect(BAT **bnp, BAT *b, BAT *s,
_______________________________________________
checkin-list mailing list
checkin-list@monetdb.org
https://www.monetdb.org/mailman/listinfo/checkin-list
