[
https://issues.apache.org/jira/browse/AVRO-1182?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13687223#comment-13687223
]
Daniel Russel commented on AVRO-1182:
-------------------------------------
> Is it possible to address that problem? I understand it is not
> straightforward because of no-copy input streams.
The code does (attempt to) address this problem.
Testing it systematically is a little trickier I think as I would have to force
the read block size to match the sync interval. So I would need (header_size +
sync_interval) to be equal to the input buffer size. I presume, for a given
record, the header size can be assumed to be constant over time. And the sync
interval I can specify. The input buffer size isn't exposed though, so the test
is a bit brittle to that default value changing.
I could do a more stochastic test where I make the sync size really small and
then read small records from the file. One of the syncs is likely to hit an
input buffer boundary and I can check that none are missed. Do you think that
is enough?
> DataFileReader missing seek, sync methods
> -----------------------------------------
>
> Key: AVRO-1182
> URL: https://issues.apache.org/jira/browse/AVRO-1182
> Project: Avro
> Issue Type: Improvement
> Components: c++
> Affects Versions: 1.7.3
> Reporter: Daniel Russel
> Attachments: add_seek
>
>
> The DataFileReader is missing the seek and sync methods that are found in the
> java version making it hard to navigate a file except in a linear fashion.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
