Changeset: b0c02c519446 for MonetDB
URL: https://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=b0c02c519446
Modified Files:
monetdb5/modules/atoms/str.c
monetdb5/modules/atoms/str.h
monetdb5/modules/kernel/batstr.c
sql/backends/monet5/sql_cast.c
sql/backends/monet5/sql_result.c
Branch: alloc-less-str
Log Message:
Added string buffer for more string functions and cleanup
diffs (truncated from 746 to 300 lines):
diff --git a/monetdb5/modules/atoms/str.c b/monetdb5/modules/atoms/str.c
--- a/monetdb5/modules/atoms/str.c
+++ b/monetdb5/modules/atoms/str.c
@@ -3209,8 +3209,23 @@ UTF8_offset(char *restrict s, int n)
return s;
}
+/* The batstr module functions use a single buffer to avoid malloc/free
overhead.
+ Note the buffer should be always large enough to hold null strings, so less
testing will be required */
+#define CHECK_STR_BUFFER_LENGTH(buf, buflen, nextlen, op) \
+ do { \
+ if (nextlen > *buflen) { \
+ int newlen = nextlen + 1024; \
+ str newbuf = GDKmalloc(newlen); \
+ if (!newbuf) \
+ throw(MAL, op, SQLSTATE(HY013)
MAL_MALLOC_FAIL); \
+ GDKfree(*buf); \
+ *buf = newbuf; \
+ *buflen = newlen; \
+ } \
+ } while (0)
+
static str
-convertCase(BAT *from, BAT *to, str *res, const char *src, const char *malfunc)
+convertCase(BAT *from, BAT *to, str *buf, int *buflen, const char *src, const
char *malfunc)
{
size_t len = strlen(src);
char *dst;
@@ -3218,14 +3233,17 @@ convertCase(BAT *from, BAT *to, str *res
bool lower_to_upper = from == UTF8_toUpperFrom;
if (strNil(src)) {
- *res = GDKstrdup(str_nil);
+ strcpy(*buf, str_nil);
+ return MAL_SUCCEED;
} else {
- if (BAThash(from) != GDK_SUCCEED ||
- (*res = GDKmalloc(len + 1)) == NULL) {
+ Hash *h;
+ int nextlen = len + 1;
+ if (BAThash(from) != GDK_SUCCEED)
throw(MAL, malfunc, SQLSTATE(HY013) MAL_MALLOC_FAIL);
- }
- Hash *h = from->thash;
- dst = *res;
+ CHECK_STR_BUFFER_LENGTH(buf, buflen, nextlen, malfunc);
+
+ h = from->thash;
+ dst = *buf;
while (src < end) {
int c;
@@ -3251,29 +3269,22 @@ convertCase(BAT *from, BAT *to, str *res
}
}
}
- if (dst + UTF8_CHARLEN(c) > *res + len) {
+ if (dst + UTF8_CHARLEN(c) > *buf + len) {
/* doesn't fit, so allocate more space;
* also allocate enough for the rest of the
* source */
- size_t off = dst - *res;
+ size_t off = dst - *buf;
+ int nextlen = (len += 4 + (end - src)) + 1;
- dst = GDKrealloc(*res, (len += 4 + (end - src))
+ 1);
- if (dst == NULL) {
- /* if realloc fails, original buffer is
still
- * allocated, so free it */
- GDKfree(*res);
- throw(MAL, malfunc, SQLSTATE(HY013)
MAL_MALLOC_FAIL);
- }
- *res = dst;
- dst = *res + off;
+ CHECK_STR_BUFFER_LENGTH(buf, buflen, nextlen,
malfunc);
+ dst = *buf + off;
}
UTF8_PUTCHAR(c, dst);
}
*dst = 0;
+ return MAL_SUCCEED;
}
- if (*res != NULL)
- return MAL_SUCCEED;
- illegal:
+illegal:
throw(MAL, malfunc, SQLSTATE(42000) "Illegal Unicode code point");
}
@@ -3323,7 +3334,7 @@ STRlike(const char *s, const char *pat,
return *t == 0 && *p == 0;
}
-str
+static str
STRlikewrap(bit *ret, const str *s, const str *pat, const str *esc)
{
if (strNil(*s) || strNil(*pat) || strNil(*esc))
@@ -3333,7 +3344,7 @@ STRlikewrap(bit *ret, const str *s, cons
return MAL_SUCCEED;
}
-str
+static str
STRlikewrap2(bit *ret, const str *s, const str *pat)
{
if (strNil(*s) || strNil(*pat))
@@ -3343,7 +3354,7 @@ STRlikewrap2(bit *ret, const str *s, con
return MAL_SUCCEED;
}
-str
+static str
STRtostr(str *res, const str *src)
{
if( *src == 0)
@@ -3369,7 +3380,7 @@ str_length(const char *s)
return (int) l;
}
-str
+static str
STRLength(int *res, const str *arg1)
{
*res = str_length(*arg1);
@@ -3388,26 +3399,13 @@ str_bytes(const char *s)
return (int) l;
}
-str
+static str
STRBytes(int *res, const str *arg1)
{
*res = str_bytes(*arg1);
return MAL_SUCCEED;
}
-#define CHECK_BUFFER_LENGTH(buf, buflen, nextlen, op) \
- do { \
- if (nextlen > *buflen) { \
- int newlen = nextlen + 1024; \
- str newbuf = GDKmalloc(newlen); \
- if (!newbuf) \
- throw(MAL, op, SQLSTATE(HY013)
MAL_MALLOC_FAIL); \
- GDKfree(*buf); \
- *buf = newbuf; \
- *buflen = newlen; \
- } \
- } while (0)
-
str
str_tail(str *buf, int *buflen, const char *s, int off)
{
@@ -3425,12 +3423,12 @@ str_tail(str *buf, int *buflen, const ch
}
str tail = UTF8_strtail(s, off);
int nextlen = (int) strlen(tail) + 1;
- CHECK_BUFFER_LENGTH(buf, buflen, nextlen, "str.tail");
+ CHECK_STR_BUFFER_LENGTH(buf, buflen, nextlen, "str.tail");
strcpy(*buf, tail);
return MAL_SUCCEED;
}
-str
+static str
STRTail(str *res, const str *arg1, const int *offset)
{
int buflen = INITIAL_STR_BUFFER_LENGTH;
@@ -3475,7 +3473,7 @@ str_Sub_String(str *buf, int *buflen, co
}
s = UTF8_strtail(s, off);
len = UTF8_strtail(s, l) - s;
- CHECK_BUFFER_LENGTH(buf, buflen, len + 1, "str.substring");
+ CHECK_STR_BUFFER_LENGTH(buf, buflen, len + 1, "str.substring");
strcpy_len(*buf, s, (size_t)(len + 1));
return MAL_SUCCEED;
}
@@ -3505,7 +3503,7 @@ str_from_wchr(str *buf, int *buflen, int
strcpy(*buf, str_nil);
return MAL_SUCCEED;
}
- CHECK_BUFFER_LENGTH(buf, buflen, 5, "str.unicode");
+ CHECK_STR_BUFFER_LENGTH(buf, buflen, 5, "str.unicode");
str s = *buf;
UTF8_PUTCHAR(c, s);
*s = 0;
@@ -3514,7 +3512,7 @@ illegal:
throw(MAL, "str.unicode", SQLSTATE(42000) "Illegal Unicode code point");
}
-str
+static str
STRFromWChr(str *res, const int *c)
{
int buflen = INITIAL_STR_BUFFER_LENGTH;
@@ -3552,7 +3550,7 @@ illegal:
throw(MAL, "str.unicodeAt", SQLSTATE(42000) "Illegal Unicode code
point");
}
-str
+static str
STRWChrAt(int *res, const str *arg1, const int *at)
{
return str_wchr_at(res, *arg1, *at);
@@ -3568,7 +3566,7 @@ str_is_prefix(const char *s, const char
return strncmp(s, prefix, strlen(prefix)) == 0;
}
-str
+static str
STRPrefix(bit *res, const str *arg1, const str *arg2)
{
*res = str_is_prefix(*arg1, *arg2);
@@ -3593,7 +3591,7 @@ str_is_suffix(const char *s, const char
}
/* returns whether arg1 ends with arg2 */
-str
+static str
STRSuffix(bit *res, const str *arg1, const str *arg2)
{
*res = str_is_suffix(*arg1, *arg2);
@@ -3601,15 +3599,51 @@ STRSuffix(bit *res, const str *arg1, con
}
str
+str_lower(str *buf, int *buflen, const char *s)
+{
+ return convertCase(UTF8_toLowerFrom, UTF8_toLowerTo, buf, buflen, s,
"str.lower");
+}
+
+static str
STRLower(str *res, const str *arg1)
{
- return convertCase(UTF8_toLowerFrom, UTF8_toLowerTo, res, *arg1,
"str.lower");
+ int buflen = INITIAL_STR_BUFFER_LENGTH;
+ str buf = GDKmalloc(buflen), msg;
+
+ *res = NULL;
+ if (!buf)
+ throw(SQL, "str.lower", SQLSTATE(HY013) MAL_MALLOC_FAIL);
+ msg = str_lower(&buf, &buflen, *arg1);
+ if (!msg && !(*res = GDKstrdup(buf))) {
+ msg = createException(MAL, "str.lower", SQLSTATE(HY013)
MAL_MALLOC_FAIL);
+ }
+
+ GDKfree(buf);
+ return msg;
}
str
+str_upper(str *buf, int *buflen, const char *s)
+{
+ return convertCase(UTF8_toUpperFrom, UTF8_toUpperTo, buf, buflen, s,
"str.upper");
+}
+
+static str
STRUpper(str *res, const str *arg1)
{
- return convertCase(UTF8_toUpperFrom, UTF8_toUpperTo, res, *arg1,
"str.upper");
+ int buflen = INITIAL_STR_BUFFER_LENGTH;
+ str buf = GDKmalloc(buflen), msg;
+
+ *res = NULL;
+ if (!buf)
+ throw(SQL, "str.upper", SQLSTATE(HY013) MAL_MALLOC_FAIL);
+ msg = str_upper(&buf, &buflen, *arg1);
+ if (!msg && !(*res = GDKstrdup(buf))) {
+ msg = createException(MAL, "str.upper", SQLSTATE(HY013)
MAL_MALLOC_FAIL);
+ }
+
+ GDKfree(buf);
+ return msg;
}
/* find first occurrence of needle in haystack */
@@ -3781,67 +3815,112 @@ const int whitespace[] = {
};
#define NSPACES (sizeof(whitespace) / sizeof(whitespace[0]))
-/* remove all whitespace from either side of arg1 */
str
-STRStrip(str *res, const str *arg1)
+str_strip(str *buf, int *buflen, const char *s)
{
- const char *s = *arg1;
- size_t len;
- size_t n;
-
if (strNil(s)) {
- *res = GDKstrdup(str_nil);
+ strcpy(*buf, str_nil);
+ return MAL_SUCCEED;
} else {
- len = strlen(s);
- n = lstrip(s, len, whitespace, NSPACES);
+ size_t len = strlen(s);
+ size_t n = lstrip(s, len, whitespace, NSPACES);
s += n;
len -= n;
_______________________________________________
checkin-list mailing list
[email protected]
https://www.monetdb.org/mailman/listinfo/checkin-list
