[
https://issues.apache.org/jira/browse/PARQUET-911?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15972447#comment-15972447
]
Itai Incze commented on PARQUET-911:
------------------------------------
Been working on this a bit, and I have a working prototype for building nested
structs, without proper handling of repetition yet.
What I'm wondering about right now before submitting a PR is what the API
should be like -
in {{arrow/reader.h}} there's the method ReadColumn which refers to a flattened
schema. Does it make sense to keep such an API side by side with a nested
"column" read method?
On one hand, in the context of reading parquet, it might sound useful. On the
other hand the whole notion of a "column" is not exactly an arrow terminology.
This reflects on other questions, for example - in {{ReadTable}}, what do the
{{column_indices}} mean? indices to the schema tree leaves or the direct top
level fields of the schema?
cc [~wesmckinn]
> C++: Support nested structs in parquet_arrow
> --------------------------------------------
>
> Key: PARQUET-911
> URL: https://issues.apache.org/jira/browse/PARQUET-911
> Project: Parquet
> Issue Type: Improvement
> Components: parquet-cpp
> Reporter: Uwe L. Korn
>
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
