[
https://issues.apache.org/jira/browse/DERBY-4789?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Rick Hillegas updated DERBY-4789:
---------------------------------
Attachment: derby-4789-01-ab-alwaysForTableFunctions.diff
Attaching derby-4789-01-ab-alwaysForTableFunctions.diff. This patch enables the
bulk-insert optimization when the source stream of an INSERT statement includes
a table function. I am running regression tests now.
This patch does the following:
1) At the end of optimization of an INSERT statement, bulk-insert is turned on
if a user-defined table function is found in the source stream of the INSERT.
2) No changes are made to bulk-insert itself. New conglomerates are created
only if the target table is empty
The patch is tested using an INSERT from a table function and an INSERT from a
view which wraps a table function. In both cases, bulk-insert occurs if the
target table is empty.
Touches the following files:
------------
A
java/engine/org/apache/derby/impl/sql/compile/HasTableFunctionVisitor.java
M java/engine/org/apache/derby/impl/sql/compile/FromVTI.java
M java/engine/org/apache/derby/impl/sql/compile/HasNodeVisitor.java
Creates a new Visitor which reports whether the AST contains a user-defined
table function.
------------
M java/engine/org/apache/derby/impl/sql/compile/InsertNode.java
Logic to support (1).
------------
M
java/testing/org/apache/derbyTesting/functionTests/tests/lang/TableFunctionTest.java
New test cases.
> Always apply the bulk-insert optimization when inserting from a table
> function.
> -------------------------------------------------------------------------------
>
> Key: DERBY-4789
> URL: https://issues.apache.org/jira/browse/DERBY-4789
> Project: Derby
> Issue Type: Improvement
> Components: SQL
> Reporter: Rick Hillegas
> Attachments: derby-4789-01-ab-alwaysForTableFunctions.diff
>
>
> Inserting from a table function is a lot like importing from a file:
> 1) Derby has limited visibility into the size of the external data source.
> 2) The user is often trying to import a large data set.
> The import procedures assume that Derby should always apply the bulk-insert
> optimization when importing from a file. The same assumption seems reasonable
> whenever a table function appears in the source stream of an INSERT.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
