diff --git a/mysql-test/r/ctype_utf8.result b/mysql-test/r/ctype_utf8.result index 64c693a292a..48ff82aeeb2 100644 --- a/mysql-test/r/ctype_utf8.result +++ b/mysql-test/r/ctype_utf8.result @@ -955,6 +955,10 @@ char_length(a) length(a) a 2 4 ан drop table t1; set names utf8; +select 'andre%' like 'andreñ%' escape 'ñ'; +'andre%' like 'andreñ%' escape 'ñ' +1 +set names utf8; select 'a\\' like 'a\\'; 'a\\' like 'a\\' 1 diff --git a/mysql-test/r/func_like.result b/mysql-test/r/func_like.result index ac8e5eda8e8..7e6fedb9403 100644 --- a/mysql-test/r/func_like.result +++ b/mysql-test/r/func_like.result @@ -158,3 +158,10 @@ DROP TABLE t1; select _cp866'aaaaaaaaa' like _cp866'%aaaa%' collate cp866_bin; _cp866'aaaaaaaaa' like _cp866'%aaaa%' collate cp866_bin 1 +set names koi8r; +select 'andre%' like 'andreÊ%' escape 'Ê'; +'andre%' like 'andreÊ%' escape 'Ê' +1 +select _cp1251'andre%' like convert('andreÊ%' using cp1251) escape 'Ê'; +_cp1251'andre%' like convert('andreÊ%' using cp1251) escape 'Ê' +1 diff --git a/mysql-test/t/ctype_utf8.test b/mysql-test/t/ctype_utf8.test index ce259f465d9..041451272d4 100644 --- a/mysql-test/t/ctype_utf8.test +++ b/mysql-test/t/ctype_utf8.test @@ -810,6 +810,12 @@ alter table t1 modify a char(2) character set utf8; select char_length(a), length(a), a from t1 order by a; drop table t1; +# +# Bugs#12611 +# ESCAPE + LIKE do not work when the escape char is a multibyte one +# +set names utf8; +select 'andre%' like 'andreñ%' escape 'ñ'; # # Bugs#11754: SET NAMES utf8 followed by SELECT "A\\" LIKE "A\\" returns 0 diff --git a/mysql-test/t/func_like.test b/mysql-test/t/func_like.test index 684d7032038..4e1183afeff 100644 --- a/mysql-test/t/func_like.test +++ b/mysql-test/t/func_like.test @@ -96,4 +96,21 @@ DROP TABLE t1; # select _cp866'aaaaaaaaa' like _cp866'%aaaa%' collate cp866_bin; +# +# Check 8bit escape character +# +set names koi8r; +select 'andre%' like 'andreÊ%' escape 'Ê'; + +# Check 8bit escape character with charset conversion: +# For "a LIKE b ESCAPE c" expressions, +# escape character is converted into the operation character set, +# which is result of aggregation of character sets of "a" and "b". +# "c" itself doesn't take part in aggregation, because its collation +# doesn't matter, escape character is always compared binary. +# In the example below, escape character is converted from koi8r into cp1251: +# +select _cp1251'andre%' like convert('andreÊ%' using cp1251) escape 'Ê'; + +# # End of 4.1 tests diff --git a/sql/item_cmpfunc.cc b/sql/item_cmpfunc.cc index cc2849ff7e6..31bfa9ba3b7 100644 --- a/sql/item_cmpfunc.cc +++ b/sql/item_cmpfunc.cc @@ -2811,7 +2811,42 @@ bool Item_func_like::fix_fields(THD *thd, Item **ref) { /* If we are on execution stage */ String *escape_str= escape_item->val_str(&tmp_value1); - escape= escape_str ? *(escape_str->ptr()) : '\\'; + if (escape_str) + { + CHARSET_INFO *cs= cmp.cmp_collation.collation; + if (use_mb(cs)) + { + my_wc_t wc; + int rc= cs->cset->mb_wc(cs, &wc, + (const uchar*) escape_str->ptr(), + (const uchar*) escape_str->ptr() + + escape_str->length()); + escape= (int) (rc > 0 ? wc : '\\'); + } + else + { + /* + In the case of 8bit character set, we pass native + code instead of Unicode code as "escape" argument. + Convert to "cs" if charset of escape differs. + */ + uint32 unused; + if (escape_str->needs_conversion(escape_str->length(), + escape_str->charset(), cs, &unused)) + { + char ch; + uint errors; + uint32 cnvlen= copy_and_convert(&ch, 1, cs, escape_str->ptr(), + escape_str->length(), + escape_str->charset(), &errors); + escape= cnvlen ? ch : '\\'; + } + else + escape= *(escape_str->ptr()); + } + } + else + escape= '\\'; /* We could also do boyer-more for non-const items, but as we would have to diff --git a/sql/item_cmpfunc.h b/sql/item_cmpfunc.h index 1915dbaabf6..0609adce7f0 100644 --- a/sql/item_cmpfunc.h +++ b/sql/item_cmpfunc.h @@ -966,7 +966,7 @@ class Item_func_like :public Item_bool_func2 Item *escape_item; public: - char escape; + int escape; Item_func_like(Item *a,Item *b, Item *escape_arg) :Item_bool_func2(a,b), canDoTurboBM(FALSE), pattern(0), pattern_len(0),