[ 
https://issues.apache.org/jira/browse/ARROW-8657?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17096866#comment-17096866
 ] 

Wes McKinney edited comment on ARROW-8657 at 4/30/20, 6:37 PM:
---------------------------------------------------------------

For the record, I think we need to introduce a new flag to toggle the use of 
newer logical types and associated casting/metadata behavior, and leave the 
1.0/2.0 flag for its intended use, i.e. the DataPageV1 vs DataPageV2

So my suggested fix is:

* Add the new flag that is separate from switching version 1.0/2.0
* Revert the behavior in Python of version='2.0' to use DataPageV1, **but issue 
a FutureWarning to get people to use the new flag**
* In a future release (maybe 2 releases from now), {{version='2.0'}} will again 
write DataPageV2


was (Author: wesmckinn):
For the record, I think we need to introduce a new flag to toggle the use of 
newer logical types and associated casting/metadata behavior, and leave the 
1.0/2.0 flag for its intended use, i.e. the DataPageV1 vs DataPageV2

So my suggested fix is:

* Add the new flag that is separate from switching version 1.0/2.0
* Revert the behavior in Python of version='2.0' to use DataPageV1, **but make 
a future warning to get people to use the new flag**
* In a future release (maybe 2 releases from now), {{version='2.0'}} will again 
write DataPageV2

> [Python][C++][Parquet] Forward compatibility issue from 0.16 to 0.17 when 
> using version='2.0'
> ---------------------------------------------------------------------------------------------
>
>                 Key: ARROW-8657
>                 URL: https://issues.apache.org/jira/browse/ARROW-8657
>             Project: Apache Arrow
>          Issue Type: Bug
>          Components: C++, Python
>    Affects Versions: 0.17.0
>            Reporter: Pierre Belzile
>            Priority: Major
>
> With the recent release of 0.17, the ParquetVersion is used to define the 
> logical type interpretation of fields and the selection of the DataPage 
> format.
> As a result all parquet files that were created with ParquetVersion::V2 to 
> get features such as unsigned int32s, timestamps with nanosecond resolution, 
> etc are now unreadable. That's TBs of data in my case.
> Those two concerns should be separated. Given that that DataPageV2 pages were 
> not written prior to 0.17 and in order to allow reading existing files, the 
> existing version property should continue to operate as in 0.16 and inform 
> the logical type mapping.
> Some consideration should be given to issue a release 0.17.1.
>  



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

Reply via email to