[
https://issues.apache.org/jira/browse/BEAM-12791?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17476357#comment-17476357
]
Kenneth Knowles commented on BEAM-12791:
----------------------------------------
This is a great request. There is a new feature in Beam that allows this kind
of dynamic read to be executed in a scalable way. So far it has been used to
enhance a few IOs but there is plenty more to do. I'll ping [~lcwik] and
[~chamikara] to make a note of the feature request.
It is worth noting that if each individual Datastore query is small, it is OK
to do it yourself in a DoFn using the Datastore client library. The main
limitation is that each read from Datastore cannot be split, unless you
implement it as a "splittable DoFn".
> Allow Queries from PCollection in
> datastore.v1new.datastoreio.ReadFromDatastore
> -------------------------------------------------------------------------------
>
> Key: BEAM-12791
> URL: https://issues.apache.org/jira/browse/BEAM-12791
> Project: Beam
> Issue Type: New Feature
> Components: io-py-gcp
> Reporter: Willi Schinmeyer
> Priority: P2
>
> As far as I can tell, I currently have to provide the Datastore Query when I
> initially create my pipeline. However, I would like to only fetch Entities
> corresponding to the data currently being processed by my pipeline. In other
> words, I would like to pass a PCollection of Queries (or at least Query
> Filters) to ReadFromDatastore.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
