expression: fix different types compare error #21338
Conversation
ti-srebot
added
the
first-time-contributor
|
Please follow PR Title Format:
Or if the count of mainly changed packages are more than 3, use
|
|
Please follow PR Title Format:
Or if the count of mainly changed packages are more than 3, use
|
|
Please follow PR Title Format:
Or if the count of mainly changed packages are more than 3, use
|
|
Please follow PR Title Format:
Or if the count of mainly changed packages are more than 3, use
|
…field's type is typebit
|
Please follow PR Title Format:
Or if the count of mainly changed packages are more than 3, use
|
|
Please follow PR Title Format:
Or if the count of mainly changed packages are more than 3, use
|
|
Please follow PR Title Format:
Or if the count of mainly changed packages are more than 3, use
|
zhaoxugang
changed the title
|
/run-all-tests |
zhaoxugang
changed the title
|
/run-all-tests |
zhaoxugang
changed the title
zhaoxugang
changed the title
|
/rebuild |
|
/run-all-tests |
1 similar comment
|
/run-all-tests |
|
PTAL @wshwsh12 |
expression/builtin.go
Outdated
| if fun, ok := res.(*ScalarFunction); (ok && fun.FuncName.L == ast.Cast) || res.GetType().Tp == mysql.TypeBit { | ||
| if evalType == types.ETInt || evalType == types.ETReal { | ||
| lfg, rfg := args[0].GetType().Flag, args[1].GetType().Flag | ||
| // set signedFlag to 0 when bit convert explicitly or implicitly to int or real |
There was a problem hiding this comment.
I can't understand the logic. In the case, all evalType.flag don't contain UnsignedFlag, I think. Could you provide an example?
drop table if exists t;
create table t(a bit(64), b double);
insert into t values(-21172, -11623);
select * from t where a < b;
I have a try, only add builtin_cast.goL487 and builtin_cast_vec.goL111, it also works. Is there any negligence in what I thought?
There was a problem hiding this comment.
I removed builtin.goL173,it not work,for this case i see the step:
1.Invoke GetAccurateCmpType() output ETReal
2.Invoke newBaseBuiltinFuncWithTp()
in this method the param a will produce the scalarFunction named cast,retFlag=160,arg's retFlag=32 by WrapWithCastAsReal()
so in builtin_cast_vec.goL111 the unsignedArgs0=true,the a will change to uint.
There was a problem hiding this comment.
Why the flag for column a is 32..? It doesn't has the unsigned flag when creating table.
There was a problem hiding this comment.
If cast a as double explicitly, it works...
[tidb]> select * from t where cast(a as double) < b;
+----------+--------+
| a | b |
+----------+--------+
| �������L | -11623 |
+----------+--------+
1 row in set (5.925 sec)
There was a problem hiding this comment.
in ddl_api.goL537 invoke col.Flag |= mysql.UnsignedFlag when create table,
you mean i should't change retFlag in generateRetFlag, but cast bit to double?
There was a problem hiding this comment.
I have a try in mysql8.0.22.. So the behavior in tidb is right? It maybe a bug in mysql8.0.21 and the previous version?
MySQL [test]> select version();
+-----------+
| version() |
+-----------+
| 8.0.22 |
+-----------+
1 row in set (0.000 sec)
MySQL [test]> drop table if exists t;
Query OK, 0 rows affected (0.006 sec)
MySQL [test]> create table t(a bit(64), b double);
Query OK, 0 rows affected (0.009 sec)
MySQL [test]> insert into t values(-21172, -11623);
Query OK, 1 row affected (0.002 sec)
MySQL [test]> select * from t where a < b;
Empty set (0.000 sec)
But in mysql8.0.21,
MySQL [test]> select version();
+-----------+
| version() |
+-----------+
| 8.0.21 |
+-----------+
1 row in set (0.000 sec)
MySQL [test]> drop table if exists t;
Query OK, 0 rows affected (0.007 sec)
MySQL [test]> create table t(a bit(64), b double);
Query OK, 0 rows affected (0.008 sec)
MySQL [test]> insert into t values(-21172, -11623);
Query OK, 1 row affected (0.001 sec)
MySQL [test]> select * from t where a < b;
+----------+--------+
| a | b |
+----------+--------+
| �������L | -11623 |
+----------+--------+
1 row in set (0.000 sec)
|
/run-all-test |
|
Please resolve conflicts. |
|
/merge |
ti-srebot
added
the
status/can-merge
|
/run-all-tests |
What problem does this PR solve?
Issue Number: close #20128
Problem Summary:
What is changed and how it works?
What's Changed:
fix different types compare error
How it Works:
Related changes
pingcap/docs/pingcap/docs-cn:Check List
Tests
Release note