[
https://issues.apache.org/jira/browse/AVRO-3590?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17603220#comment-17603220
]
Ryan Skraba commented on AVRO-3590:
-----------------------------------
Hello! Can you help clarify what's happening in this change? I think I'm
missing something!
How is it possible that the static initializer making
*{{SnappyCodec.INSTANCE}}* can see the snappy library and return an the
non-null Option and SnappyCodec, but the static initializer in
*{{CodecFactory}}* can't? Is there some sort of ClassLoader oddness happening
between the two?
If it's the case, it definitely should be fixed! Is there a corresponding
Flink Jira for example?
> Fix the problem that CodeFactory is null when specifying snappy format when
> java Client compresses data
> -------------------------------------------------------------------------------------------------------
>
> Key: AVRO-3590
> URL: https://issues.apache.org/jira/browse/AVRO-3590
> Project: Apache Avro
> Issue Type: Bug
> Components: java
> Affects Versions: 1.10.0
> Reporter: xingyuan cheng
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.10.0
>
> Attachments: image-2022-07-27-16-10-17-970.png,
> image-2022-07-27-16-10-33-410.png, image-2022-07-27-16-10-42-262.png,
> image-2022-07-27-16-11-46-218.png
>
> Time Spent: 50m
> Remaining Estimate: 0h
>
> When flink-avro writes data in batches based on the avro format, the general
> user will specify snappy compression. However, in the avro-1.10.0 version,
> the Codec and CodeFactory are returned by directly applying for the object,
> so that the user cannot obtain it Information for SnappyCodec.
> I refer to the practice of ```NullCodec```, changing it to a global variable
> way will work fine
>
> !image-2022-07-27-16-10-17-970.png!
>
> Before repair:
> !image-2022-07-27-16-10-42-262.png!
>
>
> after Repair:
>
> !image-2022-07-27-16-11-46-218.png!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
