[ 
https://issues.apache.org/jira/browse/PARQUET-1870?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17127963#comment-17127963
 ] 

Gabor Szadovszky commented on PARQUET-1870:
-------------------------------------------

We've already had a couple of discussions related to {{INT96}}. The main 
problem is that {{INT96}} is not only deprecated but never specified properly. 
We don't want to encourage anyone to use it so we don't want to implement any 
additional support for this type.

Meanwhile, I understand your problem. One way to workaround this issue is to 
not use {{parquet-avro}} but another binding (e.g. {{parquet-thrift}}, 
{{parquet-protobuf}} etc.) that handles {{INT96}}.  Though, I am not sure which 
one would handle it properly.
If you want to fix this in {{parquet-avro}}, it would be better to hide this 
new behavior behind a flag. (You may check the related parts of [this 
PR|https://github.com/apache/parquet-mr/pull/778] about introducing a new 
config.)

> Handle INT96 more gracefully in parquet-avro
> --------------------------------------------
>
>                 Key: PARQUET-1870
>                 URL: https://issues.apache.org/jira/browse/PARQUET-1870
>             Project: Parquet
>          Issue Type: Improvement
>          Components: parquet-avro
>    Affects Versions: 1.11.0
>            Reporter: Ben Watson
>            Priority: Minor
>
> The parquet-avro library does not support INT96 columns (PARQUET-323), and 
> any attempt to process a file containing such a column results in:
> {code:java}
> throw new IllegalArgumentException("INT96 not implemented and is 
> deprecated");{code}
> INT96 is still used in many legacy datasets, and so it would be useful to be 
> able to process Parquet files containing these records, even if the INT96 
> values themselves aren't rendered.
> The same functionality has already been re-added into parquet-pig 
> (PARQUET-1133).



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to