It seems like we need a multiple, case insensitive map to store metadata? ------------------
Liang Zhang (John) Apache ShardingSphere & Dubbo tsubasaotl <[email protected]> 于2019年4月8日周一 下午7:44写道： > Hi everybody, > When I try to fix issue#2152[1], I can't find the best way to solve it. > Because the case of the label will change with the database type. > > > For my test, in MySQL, the result is defined by the case of the query > field in SQL. > example: > select id from tb --> id > select ID from tb --> ID > > > In Oracle, whatever the field is in SQL, the result must be upper case. > example: > select id from tb --> ID (will cause issue#2152[1]) > select ID from tb --> ID > > > > And in pg, the result is lower case if SQL without double quotes > example: > select id from tb --> id > select ID from tb --> id (will cause issue#2152[1]) > > > > But if the field in SQL with double quotes, the result will change > example: > select "ID" from tb --> ID > > > There are two solutions, > 1. Cache the real labels got from JDBC, and ignore case when get index > from label. > 2. Do upper or lower for Cached labels according to database type. > > > The first solution may lose field for some SQL like 'select xxx as STATUS, > status from tb' . > The second solution may return error result for method > `QueryResultMetaData.getColumnLabel`. > > > Any advise or new solution for this issue? > > > [1] https://github.com/apache/incubator-shardingsphere/issues/2152 > > > ------------------ > Yi Yang (Sion) > Apache ShardingSphere
