fix regexp expr bug: should consider collations of input params

src/sql/engine/expr/ob_expr_regexp.cpp

@@ -66,43 +66,51 @@ int ObExprRegexp::calc_result_type2(ObExprResType &type,
 {
   int ret = OB_SUCCESS;
   ObRawExpr * raw_expr = type_ctx.get_raw_expr();
+  ObCollationType res_cs_type = CS_TYPE_INVALID;
+  ObCollationLevel res_cs_level = CS_LEVEL_INVALID;
   CK(NULL != type_ctx.get_raw_expr());
   if (type1.is_null() || type2.is_null()) {
     type.set_int32();
     type.set_precision(DEFAULT_PRECISION_FOR_BOOL);
     type.set_scale(DEFAULT_SCALE_FOR_INTEGER);
+  } else if (OB_UNLIKELY(!is_type_valid(type1.get_type()) || !is_type_valid(type2.get_type()))) {
+    ret = OB_INVALID_ARGUMENT;
+    LOG_WARN("the param is not castable", K(ret), K(type1), K(type2));
+  } else if (OB_FAIL(ObCharset::aggregate_collation(type1.get_calc_collation_level(),
+                                              type1.get_calc_collation_type(),
+                                              type2.get_calc_collation_level(),
+                                              type2.get_calc_collation_type(),
+                                              res_cs_level,
+                                              res_cs_type))) {
+      LOG_WARN("fail to aggregate collation", K(ret), K(type1), K(type2));
   } else {
-    if (OB_UNLIKELY(!is_type_valid(type1.get_type()) || !is_type_valid(type2.get_type()))) {
-      ret = OB_INVALID_ARGUMENT;
-      LOG_WARN("the param is not castable", K(ret), K(type1), K(type2));
+    type.set_int32();
+    type.set_precision(DEFAULT_PRECISION_FOR_BOOL);
+    type.set_scale(DEFAULT_SCALE_FOR_INTEGER);
+    //why we set the calc collation type is utf16, because the ICU regexp engine is used uft16,
+    //we need convert it the need collation in advance, and no need to think about in regexp.
+    bool is_case_sensitive = ObCharset::is_bin_sort(res_cs_type);
+    bool need_utf8 = false;
+    type1.set_calc_type(ObVarcharType);
+    type1.set_calc_collation_level(type.get_collation_level());
+    type2.set_calc_type(ObVarcharType);
+    type2.set_calc_collation_level(type.get_collation_level());
+    if (OB_FAIL(ObExprRegexContext::check_need_utf8(raw_expr->get_param_expr(1), need_utf8))) {
+      LOG_WARN("fail to check need utf8", K(ret));
+    } else if (need_utf8) {
+      type2.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF8MB4_BIN : CS_TYPE_UTF8MB4_GENERAL_CI);
     } else {
-      type.set_int32();
-      type.set_precision(DEFAULT_PRECISION_FOR_BOOL);
-      type.set_scale(DEFAULT_SCALE_FOR_INTEGER);
-      //why we set the calc collation type is utf16, because the ICU regexp engine is used uft16,
-      //we need convert it the need collation in advance, and no need to think about in regexp.
-      bool is_case_sensitive = ObCharset::is_bin_sort(type1.get_calc_collation_type());
-      bool need_utf8 = false;
-      type1.set_calc_type(ObVarcharType);
-      type1.set_calc_collation_level(type.get_collation_level());
-      type2.set_calc_type(ObVarcharType);
-      type2.set_calc_collation_level(type.get_collation_level());
-      if (OB_FAIL(ObExprRegexContext::check_need_utf8(raw_expr->get_param_expr(1), need_utf8))) {
-        LOG_WARN("fail to check need utf8", K(ret));
-      } else if (need_utf8) {
-        type2.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF8MB4_BIN : CS_TYPE_UTF8MB4_GENERAL_CI);
-      } else {
-        type2.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF16_BIN : CS_TYPE_UTF16_GENERAL_CI);
-      }
-      need_utf8 = false;
-      if (OB_FAIL(ret)) {
-      } else if (OB_FAIL(ObExprRegexContext::check_need_utf8(raw_expr->get_param_expr(0), need_utf8))) {
-        LOG_WARN("fail to check need utf8", K(ret));
-      } else if (need_utf8) {
-        type1.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF8MB4_BIN : CS_TYPE_UTF8MB4_GENERAL_CI);
-      } else {
-        type1.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF16_BIN : CS_TYPE_UTF16_GENERAL_CI);
-      }
+      type2.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF16_BIN : CS_TYPE_UTF16_GENERAL_CI);
+    }
+
+    need_utf8 = false;
+    if (OB_FAIL(ret)) {
+    } else if (OB_FAIL(ObExprRegexContext::check_need_utf8(raw_expr->get_param_expr(0), need_utf8))) {
+      LOG_WARN("fail to check need utf8", K(ret));
+    } else if (need_utf8) {
+      type1.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF8MB4_BIN : CS_TYPE_UTF8MB4_GENERAL_CI);
+    } else {
+      type1.set_calc_collation_type(is_case_sensitive ? CS_TYPE_UTF16_BIN : CS_TYPE_UTF16_GENERAL_CI);
     }
   }
   return ret;

tools/deploy/mysql_test/test_suite/expr/r/mysql/func_regexp.result

@@ -6328,3 +6328,49 @@ select * from t1 where c1 regexp '^U';
 c1
 UNPRESS123
 UNPRESS456
+set names gbk;
+select 'a' collate gbk_bin regexp 'A';
+'a' collate gbk_bin regexp 'A'
+0
+select 'a' collate gbk_chinese_ci regexp 'A';
+'a' collate gbk_chinese_ci regexp 'A'
+1
+select 'a' regexp 'A' collate gbk_chinese_ci;
+'a' regexp 'A' collate gbk_chinese_ci
+1
+select 'a' regexp 'A' collate gbk_bin;
+'a' regexp 'A' collate gbk_bin
+0
+select 'a' collate gbk_bin regexp 'A' collate gbk_bin;
+'a' collate gbk_bin regexp 'A' collate gbk_bin
+0
+select 'a' collate gbk_chinese_ci regexp 'A' collate gbk_chinese_ci;
+'a' collate gbk_chinese_ci regexp 'A' collate gbk_chinese_ci
+1
+select 'a' collate gbk_bin regexp 'A' collate gbk_chinese_ci;
+ERROR HY000: Illegal mix of collations
+select 'a' collate gbk_chinese_ci regexp 'A' collate gbk_bin;
+ERROR HY000: Illegal mix of collations
+set names latin1;
+select 'a' collate latin1_bin regexp 'A';
+'a' collate latin1_bin regexp 'A'
+0
+select 'a' collate latin1_swedish_ci regexp 'A';
+'a' collate latin1_swedish_ci regexp 'A'
+1
+select 'a' regexp 'A' collate latin1_swedish_ci;
+'a' regexp 'A' collate latin1_swedish_ci
+1
+select 'a' regexp 'A' collate latin1_bin;
+'a' regexp 'A' collate latin1_bin
+0
+select 'a' collate latin1_bin regexp 'A' collate latin1_bin;
+'a' collate latin1_bin regexp 'A' collate latin1_bin
+0
+select 'a' collate latin1_swedish_ci regexp 'A' collate latin1_swedish_ci;
+'a' collate latin1_swedish_ci regexp 'A' collate latin1_swedish_ci
+1
+select 'a' collate latin1_bin regexp 'A' collate latin1_swedish_ci;
+ERROR HY000: Illegal mix of collations
+select 'a' collate latin1_swedish_ci regexp 'A' collate latin1_bin;
+ERROR HY000: Illegal mix of collations

tools/deploy/mysql_test/test_suite/expr/t/func_regexp.test

@@ -454,4 +454,31 @@ drop table if exists t1;
 create table t1(c1 blob);
 insert into t1 values('UNPRESS123');
 insert into t1 values('UNPRESS456');
-select * from t1 where c1 regexp '^U';
\ No newline at end of file
+select * from t1 where c1 regexp '^U';
+
+
+
+##bug48378677
+set names gbk;
+select 'a' collate gbk_bin regexp 'A';
+select 'a' collate gbk_chinese_ci regexp 'A';
+select 'a' regexp 'A' collate gbk_chinese_ci;
+select 'a' regexp 'A' collate gbk_bin;
+select 'a' collate gbk_bin regexp 'A' collate gbk_bin;
+select 'a' collate gbk_chinese_ci regexp 'A' collate gbk_chinese_ci;
+--error 1267
+select 'a' collate gbk_bin regexp 'A' collate gbk_chinese_ci;
+--error 1267
+select 'a' collate gbk_chinese_ci regexp 'A' collate gbk_bin;
+
+set names latin1;
+select 'a' collate latin1_bin regexp 'A';
+select 'a' collate latin1_swedish_ci regexp 'A';
+select 'a' regexp 'A' collate latin1_swedish_ci;
+select 'a' regexp 'A' collate latin1_bin;
+select 'a' collate latin1_bin regexp 'A' collate latin1_bin;
+select 'a' collate latin1_swedish_ci regexp 'A' collate latin1_swedish_ci;
+--error 1267
+select 'a' collate latin1_bin regexp 'A' collate latin1_swedish_ci;
+--error 1267
+select 'a' collate latin1_swedish_ci regexp 'A' collate latin1_bin;
\ No newline at end of file