MDEV-8362 dash '-' is not recognized in charset armscii8 on select where query

This commit is contained in:
Alexander Barkov 2015-07-14 12:00:05 +04:00
parent 657f8a8d60
commit 75931feabe
3 changed files with 62 additions and 1 deletions

View file

@ -1787,3 +1787,24 @@ SELECT * FROM t2 WHERE a=(SELECT a FROM t1) AND a=_LATIN1'x';
ERROR 21000: Subquery returns more than 1 row
DROP TABLE t2;
DROP TABLE t1;
#
# Start of 10.1 tests
#
#
# MDEV-8362 dash '-' is not recognized in charset armscii8 on select where query
#
CREATE TABLE t1 (a VARCHAR(64) CHARACTER SET armscii8);
INSERT INTO t1 VALUES ('abc-def');
SELECT * FROM t1 WHERE a='abc-def';
a
abc-def
SELECT * FROM t1 WHERE a LIKE 'abc%';
a
abc-def
DROP TABLE t1;
SELECT HEX(CONVERT(_utf8 0x2728292C2D2E USING armscii8));
HEX(CONVERT(_utf8 0x2728292C2D2E USING armscii8))
2728292C2D2E
#
# End of 10.1 tests
#

View file

@ -331,3 +331,22 @@ CREATE TABLE t2 (a BINARY(1));
SELECT * FROM t2 WHERE a=(SELECT a FROM t1) AND a=_LATIN1'x';
DROP TABLE t2;
DROP TABLE t1;
--echo #
--echo # Start of 10.1 tests
--echo #
--echo #
--echo # MDEV-8362 dash '-' is not recognized in charset armscii8 on select where query
--echo #
CREATE TABLE t1 (a VARCHAR(64) CHARACTER SET armscii8);
INSERT INTO t1 VALUES ('abc-def');
SELECT * FROM t1 WHERE a='abc-def';
SELECT * FROM t1 WHERE a LIKE 'abc%';
DROP TABLE t1;
SELECT HEX(CONVERT(_utf8 0x2728292C2D2E USING armscii8));
--echo #
--echo # End of 10.1 tests
--echo #

View file

@ -1303,7 +1303,28 @@ create_fromuni(struct charset_info_st *cs,
if (wc >= idx[i].uidx.from && wc <= idx[i].uidx.to && wc)
{
int ofs= wc - idx[i].uidx.from;
tab[ofs]= ch;
if (!tab[ofs] || tab[ofs] > 0x7F) /* Prefer ASCII*/
{
/*
Some character sets can have double encoding. For example,
in ARMSCII8, the following characters are encoded twice:
Encoding#1 Encoding#2 Unicode Character Name
---------- ---------- ------- --------------
0x27 0xFF U+0027 APOSTROPHE
0x28 0xA5 U+0028 LEFT PARENTHESIS
0x29 0xA4 U+0029 RIGHT PARENTHESIS
0x2C 0xAB U+002C COMMA
0x2D 0xAC U+002D HYPHEN-MINUS
0x2E 0xA9 U+002E FULL STOP
That is, both 0x27 and 0xFF convert to Unicode U+0027.
When converting back from Unicode to ARMSCII,
we prefer the ASCII range, that is we want U+0027
to convert to 0x27 rather than to 0xFF.
*/
tab[ofs]= ch;
}
}
}
}