[
https://issues.apache.org/jira/browse/CALCITE-497?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14236221#comment-14236221
]
Julian Hyde commented on CALCITE-497:
-------------------------------------
This is related to CALCITE-356, which is also about resolving long identifiers
like a.b.c.d.e.f.
What happens if a is a schema and b is a table not used in the FROM clause?
What happens if a is a table alias and there is a column-group named b and also
a column called b in a column-group z?
I think it is worth solving this and CALCITE-356 simultaneously, with lots and
lots of tests for the corner cases.
I think we should model column-groups as records, and allow qualifying field
names to be omitted if the inner field is unique. The alternative, adding
qualifiers to the names in the list of RelDataTypeField that come from
Table.getRowType(RelDataTypeFactory).getFieldList(), would complicate a lot of
things.
> Support optional qualifier for column name references
> -----------------------------------------------------
>
> Key: CALCITE-497
> URL: https://issues.apache.org/jira/browse/CALCITE-497
> Project: Calcite
> Issue Type: Improvement
> Reporter: James Taylor
> Assignee: Julian Hyde
> Labels: phoenix
>
> Some systems, such as HBase, have a named group to which columns belong (i.e.
> a column family in this case). To disambiguate column references (i.e. if the
> same column name is used in different column families), it's useful to allow
> an optional qualifier. This is similar to the RECORD field mechanism that
> Calcite already supports, but in this case the record name would be optional,
> since it's not necessary if the column names are unique. For example:
> http://phoenix.apache.org/language/index.html#column_ref
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
