[ https://issues.apache.org/jira/browse/FLINK-19799?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17221291#comment-17221291 ]
Stephan Ewen commented on FLINK-19799: -------------------------------------- Thanks, Steven. [~lzljs3620320] I think you mentioned something similar in a separate discussion. Is it fair to say that on the FileSourceEnumerator, the checkpoint state is in general a combination of the enumerator's state and the assigner's state? So if we generify it like that, would that solve issues? > Make FileSource extensible > -------------------------- > > Key: FLINK-19799 > URL: https://issues.apache.org/jira/browse/FLINK-19799 > Project: Flink > Issue Type: Improvement > Components: Connectors / FileSystem > Reporter: Stephan Ewen > Assignee: Stephan Ewen > Priority: Major > Fix For: 1.12.0 > > > The File System Source currently assumes all formats can represent their work > units as {{FileSourceSplit}}. If that is not the case, the formats cannot be > implemented using the {{FileSource}}. > We need to support extending the splits to carry additional information in > the splits, and to use that information when creating bulk readers and > handling split state. -- This message was sent by Atlassian Jira (v8.3.4#803005)