[
https://issues.apache.org/jira/browse/FLINK-4147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15361434#comment-15361434
]
Aljoscha Krettek commented on FLINK-4147:
-----------------------------------------
+1 though there is a tradeoff. Not having special-case source methods on
{{StreamExecutionEnvironment}} makes it harder to discover sources. This is the
same argument as the one between a Flink-style API and a Beam-style API. Where
in Flink, all transformation methods are readily discoverable on {{DataStream}}
while the Beam {{PCollection}} only has an {{.apply()}} method and operations
are separate.
> Consider moving the file sources from the StreamExecutionEnvironment to the
> flink-connector-filesystem
> ------------------------------------------------------------------------------------------------------
>
> Key: FLINK-4147
> URL: https://issues.apache.org/jira/browse/FLINK-4147
> Project: Flink
> Issue Type: Sub-task
> Components: DataStream API
> Reporter: Robert Metzger
> Fix For: 2.0.0
>
>
> The StreamExecutionEnvironment has many convenience methods for reading
> static files.
> However, it seems that this makes it hard to maintain Flink's APIs and that
> not so many users are using the file sources.
> The filesystem connector is the logical location for persistent file reading.
> I suggest to move those sources there.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
