Changeset: eacb1f23e00e for MonetDB
URL: https://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=eacb1f23e00e
Modified Files:
sql/backends/monet5/sql_round_impl.h
sql/server/sql_decimal.c
sql/server/sql_decimal.h
sql/server/sql_parser.y
Branch: Oct2020-merged-Jun2020
Log Message:
Stricter decimal parsing.
diffs (truncated from 317 to 300 lines):
diff --git a/sql/backends/monet5/sql_round_impl.h
b/sql/backends/monet5/sql_round_impl.h
--- a/sql/backends/monet5/sql_round_impl.h
+++ b/sql/backends/monet5/sql_round_impl.h
@@ -309,48 +309,42 @@ nil_2dec(TYPE *res, const void *val, con
static inline str
str_2dec_body(TYPE *res, const str val, const int d, const int sc)
{
- char *s = val;
- char *dot, *end;
+ char *s;
int digits;
int scale;
BIG value;
- dot = strchr(s, '.');
- if (dot != NULL) {
- s = strip_extra_zeros(s);
- digits = _strlen(s) - 1;
- scale = _strlen(dot + 1);
- } else {
- digits = _strlen(s);
- scale = 0;
- }
- end = NULL;
+ if (*d < 0 || *d >= (int) (sizeof(scales) / sizeof(scales[0])))
+ throw(SQL, STRING(TYPE), SQLSTATE(42000) "Decimal (%s) doesn't
have format (%d.%d)", *val, *d, *sc);
+
+ s = *val;
+
+ int has_errors;
value = 0;
- if (digits < 0)
- throw(SQL, STRING(TYPE), SQLSTATE(42000) "Decimal (%s) doesn't
have format (%d.%d)", s, d, sc);
- if (d < 0 || (size_t) d >= sizeof(scales) / sizeof(scales[0]))
- throw(SQL, STRING(TYPE), SQLSTATE(42000) "Decimal (%s) doesn't
have format (%d.%d)", s, d, sc);
+ // s = strip_extra_zeros(s);
+
+ value = decimal_from_str(s, &digits, &scale, &has_errors);
+ if (has_errors)
+ throw(SQL, STRING(TYPE), SQLSTATE(42000) "Decimal (%s) doesn't
have format (%d.%d)", *val, *d, *sc);
- value = decimal_from_str(s, &end);
- if (*s == '+' || *s == '-')
- digits--;
- if (scale < sc) {
+ // handle situations where the de facto scale is different from the
formal scale.
+ if (scale < *sc) {
/* the current scale is too small, increase it by adding 0's */
- int dff = sc - scale; /* CANNOT be 0! */
+ int dff = *sc - scale; /* CANNOT be 0! */
if (dff >= MAX_SCALE)
- throw(SQL, STRING(TYPE), SQLSTATE(42000) "Rounding of
decimal (%s) doesn't fit format (%d.%d)", s, d, sc);
+ throw(SQL, STRING(TYPE), SQLSTATE(42000) "Rounding of
decimal (%s) doesn't fit format (%d.%d)", *val, *d, *sc);
value *= scales[dff];
scale += dff;
digits += dff;
- } else if (scale > sc) {
+ } else if (scale > *sc) {
/* the current scale is too big, decrease it by correctly
rounding */
/* we should round properly, and check for overflow (res >=
10^digits+scale) */
- int dff = scale - sc; /* CANNOT be 0 */
+ int dff = scale - *sc; /* CANNOT be 0 */
if (dff >= MAX_SCALE)
- throw(SQL, STRING(TYPE), SQLSTATE(42000) "Rounding of
decimal (%s) doesn't fit format (%d.%d)", s, d, sc);
+ throw(SQL, STRING(TYPE), SQLSTATE(42000) "Rounding of
decimal (%s) doesn't fit format (%d.%d)", *val, *d, *sc);
BIG rnd = scales[dff] >> 1;
@@ -361,11 +355,13 @@ str_2dec_body(TYPE *res, const str val,
value /= scales[dff];
scale -= dff;
digits -= dff;
- if (value >= scales[d] || value <= -scales[d])
- throw(SQL, STRING(TYPE), SQLSTATE(42000) "Rounding of
decimal (%s) doesn't fit format (%d.%d)", s, d, sc);
+ if (value >= scales[*d] || value <= -scales[*d]) {
+ throw(SQL, STRING(TYPE), SQLSTATE(42000) "Rounding of
decimal (%s) doesn't fit format (%d.%d)", *val, *d, *sc);
+ }
}
- if (value <= -scales[d] || value >= scales[d] || *end)
- throw(SQL, STRING(TYPE), SQLSTATE(42000) "Decimal (%s) doesn't
have format (%d.%d)", s, d, sc);
+ if (value <= -scales[*d] || value >= scales[*d]) {
+ throw(SQL, STRING(TYPE), SQLSTATE(42000) "Decimal (%s) doesn't
have format (%d.%d)", *val, *d, *sc);
+ }
*res = (TYPE) value;
return MAL_SUCCEED;
}
diff --git a/sql/server/sql_decimal.c b/sql/server/sql_decimal.c
--- a/sql/server/sql_decimal.c
+++ b/sql/server/sql_decimal.c
@@ -10,46 +10,97 @@
#include "sql_decimal.h"
-#ifdef HAVE_HGE
-hge
-#else
-lng
-#endif
-decimal_from_str(char *dec, char **end)
+
+DEC_TPE
+decimal_from_str(char *dec, int* digits, int* scale, int* has_errors)
{
#ifdef HAVE_HGE
- hge res = 0;
- const hge max0 = GDK_hge_max / 10, max1 = GDK_hge_max % 10;
+ const hge max0 = GDK_hge_max / 10, max1 = GDK_hge_max % 10;
#else
- lng res = 0;
- const lng max0 = GDK_lng_max / 10, max1 = GDK_lng_max % 10;
+ const lng max0 = GDK_lng_max / 10, max1 = GDK_lng_max % 10;
#endif
- int neg = 0, seen_dot = 0;
+
+ assert(digits);
+ assert(scale);
+ assert(has_errors);
+ DEC_TPE res = 0;
+ *has_errors = 0;
+
+ int _digits = 0;
+ int _scale = 0;
+
+// preceding whitespace:
+ int neg = 0;
while(isspace((unsigned char) *dec))
dec++;
+
+// optional sign:
if (*dec == '-') {
neg = 1;
dec++;
} else if (*dec == '+') {
dec++;
}
- for (; *dec && (isdigit((unsigned char) *dec) || *dec == '.'); dec++) {
- if (*dec != '.') {
- if (res > max0 || (res == max0 && *dec - '0' > max1))
- break;
- res *= 10;
- res += *dec - '0';
- } else if (seen_dot) {
- break; /* dot cannot appear twice */
- } else {
- seen_dot = 1;
- }
+
+// optional fractional separator first opportunity
+ if (*dec == '.') { // case: (+|-).456
+fractional_sep_first_opp:
+ dec++;
+ _digits++; // add one to digits for single implicit preceding
0, e.g. (+|-)0.456
+ goto trailing_digits;
+ }
+
+// preceding_digits:
+ if (!isdigit((unsigned char) *dec)) {
+ *has_errors = 1;
+ goto end_state;
+ }
+ while (*dec == '0'){
+ // skip leading zeros in preceding digits.
+ dec++;
+ if (*dec == '.')
+ goto fractional_sep_first_opp;
}
+ for (; *dec && (isdigit((unsigned char) *dec)); dec++) {
+ if (res > max0 || (res == max0 && *dec - '0' > max1))
+ break;
+ res *= 10;
+ res += *dec - '0';
+ _digits++;
+ }
+
+// optional fractional separator second opportunity
+ if (*dec == '.') // case: (+|-)123.(456)
+ dec++;
+ else // case: (+|-)123
+ goto trailing_whitespace;
+
+trailing_digits:
+ if (!isdigit((unsigned char) *dec))
+ goto trailing_whitespace;
+ for (; *dec && (isdigit((unsigned char) *dec)); dec++) {
+ if (res > max0 || (res == max0 && *dec - '0' > max1))
+ break;
+ res *= 10;
+ res += *dec - '0';
+ _scale++;
+ }
+ _digits += _scale;
+
+trailing_whitespace:
while(isspace((unsigned char) *dec))
dec++;
- if (end)
- *end = dec;
+
+end_state:
+ /* When the string cannot be parsed up to and including the null
terminator,
+ * the string is an invalid decimal representation. */
+ if (*dec != 0)
+ *has_errors = 1;
+
+ *digits = _digits;
+ *scale = _scale;
+
if (neg)
return -res;
else
diff --git a/sql/server/sql_decimal.h b/sql/server/sql_decimal.h
--- a/sql/server/sql_decimal.h
+++ b/sql/server/sql_decimal.h
@@ -14,19 +14,14 @@
#include "gdk.h"
#ifdef HAVE_HGE
-extern hge decimal_from_str(char *dec, char **end);
-extern char * decimal_to_str(sql_allocator *sa, hge v, sql_subtype *t);
+#define DEC_TPE hge
#else
-extern lng decimal_from_str(char *dec, char **end);
-extern char * decimal_to_str(sql_allocator *sa, lng v, sql_subtype *t);
+#define DEC_TPE lng
#endif
-#ifdef HAVE_HGE
-extern hge
-#else
-extern lng
-#endif
-scale2value(int scale);
+extern DEC_TPE decimal_from_str(char *dec, int* digits, int* scale, int*
has_errors);
+extern char * decimal_to_str(sql_allocator *sa, DEC_TPE v, sql_subtype *t);
+DEC_TPE scale2value(int scale);
#endif /* _SQL_DECIMAL_H */
diff --git a/sql/server/sql_parser.y b/sql/server/sql_parser.y
--- a/sql/server/sql_parser.y
+++ b/sql/server/sql_parser.y
@@ -4710,37 +4710,43 @@ literal:
}
| INTNUM
{ char *s = sa_strdup(SA, $1);
- char *dot = strchr(s, '.');
- int digits = _strlen(s) - 1;
- int scale = digits - (int) (dot-s);
- sql_subtype t;
-
- if (digits <= 0)
- digits = 1;
- if (digits <= MAX_DEC_DIGITS) {
-#ifdef HAVE_HGE
- hge value = decimal_from_str(s, NULL);
-#else
- lng value = decimal_from_str(s, NULL);
-#endif
-
- if (*s == '+' || *s == '-')
- digits --;
- sql_find_subtype(&t, "decimal", digits, scale );
- $$ = _newAtomNode( atom_dec(SA, &t, value));
- } else {
- char *p = $1;
- double val;
-
- errno = 0;
- val = strtod($1,&p);
- if (p == $1 || is_dbl_nil(val) || (errno == ERANGE &&
(val < -1 || val > 1))) {
- sqlformaterror(m, SQLSTATE(22003) "Double value
too large or not a number (%s)", $1);
+ int digits;
+ int scale;
+ int has_errors;
+ sql_subtype t;
+
+ DEC_TPE value = decimal_from_str(s, &digits, &scale,
&has_errors);
+
+ if (has_errors) {
+ char *msg = sql_message(SQLSTATE(22003) "Double
value too large or not a number (%s)", $1);
+
+ yyerror(m, msg);
+ _DELETE(msg);
$$ = NULL;
YYABORT;
}
- sql_find_subtype(&t, "double", 51, 0 );
- $$ = _newAtomNode(atom_float(SA, &t, val));
+
+ if (digits <= MAX_DEC_DIGITS) {
+ double val = strtod($1,NULL);
+ sql_find_subtype(&t, "decimal", digits, scale );
+ $$ = _newAtomNode( atom_dec(SA, &t, value,
val));
+ }
+ else {
+ char *p = $1;
_______________________________________________
checkin-list mailing list
checkin-list@monetdb.org
https://www.monetdb.org/mailman/listinfo/checkin-list
