[
https://issues.apache.org/jira/browse/DRILL-6403?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16471138#comment-16471138
]
Paul Rogers commented on DRILL-6403:
------------------------------------
The two schema builders serve much different uses. The one in {{RowSet}}
handles all types and can handle nested structures. It is designed for fluent
use in tests.
The one in {{record}} is quite simple and appears to handle only top-level
fields that can be expressed as a {{MaterializedField}}, leaving it to the
caller to create the {{MaterializedField}}. Oddly, this older version appears
to be used in only one place: {{MergingReceiver.innerNext()}}.
Here is a suggestion: if/when operators make use of the new {{ResultSetLoader}}
classes, then can be migrated to use the {{RowSet}} schema builder. Actually,
since {{ResultSetLoader}} has schema building built in (fields can be added on
the fly), the use of a separate schema builder may no longer be necessary
except in tests and special cases.
> There are two SchemaBuilders that do the same thing. Consolidate them.
> ----------------------------------------------------------------------
>
> Key: DRILL-6403
> URL: https://issues.apache.org/jira/browse/DRILL-6403
> Project: Apache Drill
> Issue Type: Improvement
> Reporter: Timothy Farkas
> Assignee: Timothy Farkas
> Priority: Major
>
> There is org.apache.drill.test.rowSet.schema.SchemaBuilder used for testing
> and org.apache.drill.exec.record.SchemaBuilder used in Drill. They basically
> do the same thing. We should combine them and have one schema builder.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
