Standardize the separator in name
Is your feature request related to a problem or challenge?
While working on replacing GetFieldAccess with get_field function, I got the name checking error that due to the mismatch of separator in creating name. I found that most of the name built with single comma and space, so I think we should switch others to this format .join(", ").
Error is like
External error: query failed: DataFusion error: Internal error: Input field name SUM(get_field(CASE WHEN get_field(t2.struct(t1.time,t1.load1,t1.load2,t1.host), Utf8("c3")) IS NOT NULL THEN t2.struct(t1.time,t1.load1,t1.load2,t1.host) END,Utf8("c2"))) does not match with the projection expression SUM(get_field(CASE WHEN get_field(t2.struct(t1.time,t1.load1,t1.load2,t1.host),Utf8("c3")) IS NOT NULL THEN t2.struct(t1.time,t1.load1,t1.load2,t1.host) END,Utf8("c2"))).
This was likely caused by a bug in DataFusion's code and we would welcome that you file an bug report in our issue tracker
[SQL] select t2."struct(t1.time,t1.load1,t1.load2,t1.host)"['c3'] as host, sum((case when t2."struct(t1.time,t1.load1,t1.load2,t1.host)"['c3'] is not null then t2."struct(t1.time,t1.load1,t1.load2,t1.host)" end)['c2']) from (select struct(time,load1,load2,host) from t1) t2 where t2."struct(t1.time,t1.load1,t1.load2,t1.host)"['c3'] IS NOT NULL group by t2."struct(t1.time,t1.load1,t1.load2,t1.host)"['c3'] order by host;
I got one with t2.struct(t1.time,t1.load1,t1.load2,t1.host), Utf8("c3") and another t2.struct(t1.time,t1.load1,t1.load2,t1.host),Utf8("c3"). That is why I think we should fix the separator.
Since the change is quite large, I would like to wait for more comment.
Describe the solution you'd like
No response
Describe alternatives you've considered
No response
Additional context
No response
You can find the error here with cargo test --test sqllogictests -- expr. Although, it is not on the main branch, but I think most of the changes are unrelated to the name checking issue.
I agree this makes sense. Thank you
Also why do we have all of these which are slightly different? It's really unmanageable:
-
Expr::schema_name -
Expr::canonical_name -
physical_name
Also why do we have all of these which are slightly different? It's really unmanageable:
Expr::schema_nameExpr::canonical_namephysical_name
schema_name is used in Schema.
canonical_name, I'm not sure. It seems we could remove this as well
physical_name, I think we might need another schema name for Physical Expr
I'm working on this naming issue #11782
Maybe we can just use the schema_name for physical_name? It sounds like a bug if they are different: https://github.com/apache/datafusion/pull/11977