[
https://issues.apache.org/jira/browse/FLINK-5571?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15853889#comment-15853889
]
ASF GitHub Bot commented on FLINK-5571:
---------------------------------------
Github user godfreyhe commented on the issue:
https://github.com/apache/flink/pull/3176
Hi @wuchong , thanks for this review.
I think `UserDefinedScalarFunctionTest` is not enough, because `open` ,
`close` methods and job parameters are strongly depend on runtime, so it's
better to test them in integration tests.
I will add a test to verify Java ScalarFunction later.
> add open and close methods for UserDefinedFunction in TableAPI & SQL
> --------------------------------------------------------------------
>
> Key: FLINK-5571
> URL: https://issues.apache.org/jira/browse/FLINK-5571
> Project: Flink
> Issue Type: New Feature
> Components: Table API & SQL
> Reporter: godfrey he
> Assignee: godfrey he
>
> Currently, a User Defined Function (UDF) in table API & SQL works on zero,
> one, or multiple values in custom evaluation method. Many UDFs need more
> complex features, e.g. report metrics, get parameters from job configuration,
> or get extra data from distribute cache file, etc. Adding open and close
> methods in UserDefinedFunction class can solve this problem. The code cloud
> look like:
> {code}
> trait UserDefinedFunction {
> def open(context: UDFContext): Unit = {}
> def close(): Unit = {}
> }
> {code}
> UDFContext contains the information about metric reporters, job parameters,
> distribute cache, etc. The code cloud look like:
> {code}
> class UDFContext(context: RuntimeContext) {
> def getMetricGroup: MetricGroup = ???
> def getDistributedCacheFile(name: String): File = ???
> def getJobParameter(key: String, default: String): String = ???
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
