[
https://issues.apache.org/jira/browse/BEAM-14393?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17532453#comment-17532453
]
Chamikara Madhusanka Jayalath commented on BEAM-14393:
------------------------------------------------------
Thanks. Is this optimization something we can do before Beam 3.0.0 ? We don't
have to wait for the next major version unless this somehow breaks one of the
major public APIs (for example, FileSystem API). Doesn't sound like that's the
case.
> Obtain metadata field at once in file system's IO connectors
> ------------------------------------------------------------
>
> Key: BEAM-14393
> URL: https://issues.apache.org/jira/browse/BEAM-14393
> Project: Beam
> Issue Type: Improvement
> Components: io-py-common
> Reporter: Yi Hu
> Priority: P2
> Fix For: 3.0.0
>
>
> This tasks involves refactoring and improvements of IO connectors' file
> metadata related methods (GcsIO, S3IO, BlobIO, hadoop).
> Currently, we have individual methods like size, last_updated, checksum, and
> others. Each one would make a HTTP request in order to get the specific
> metadata field. If one needs to gather multiple metadata fields, then every
> specific method are called and making multiple requests under the hood.
> Actually, the HTTP response contains multiple file metadata fields but each
> time only one field is collected and others are discarded.
> We should have a public method that returns a named tuple which contains
> multiple file metadata fields. In its implementation it only makes one
> request, as existing methods for single metadata field.
>
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
