[
https://issues.apache.org/jira/browse/AVRO-893?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13114363#comment-13114363
]
Douglas Creager commented on AVRO-893:
--------------------------------------
I actually have an implementation of this in my github tracking branch, which I
haven't merged into SVN yet. I assume you're using HEAD, and not any of the
1.5 releases, since those don't have the new value_t API in them. I'll try to
get these features merged into SVN this week, but in the meantime, you could
try building the latest on my github branch:
https://github.com/dcreager/avro/tree/dcreager-mods
You use the same {{avro_file_writer_t}} type from before, but there's a new
{{avro_file_writer_append_value}} function that takes in an {{avro_value_t}}
instance.
> New value_t API does not write out spec compatible files - header, metadata
> and blocks are missing.
> ---------------------------------------------------------------------------------------------------
>
> Key: AVRO-893
> URL: https://issues.apache.org/jira/browse/AVRO-893
> Project: Avro
> Issue Type: Improvement
> Components: c
> Affects Versions: 1.5.4
> Environment: N/A
> Reporter: Lucas Martin-King
> Labels: file
>
> We have switched over to the new value_t API for our application, as it
> provides more AVRO types that we need (enum, array, etc), however we have
> noticed that there isn't a way to write out a spec compatible AVRO file.
> The old API was able to write out a header as well as the schema, the finally
> each data block (and sync marker).
> I could possibly try and re-implement this myself, however I am unsure as to
> the design (how it would fit into the new datum API).
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
