[
https://issues.apache.org/jira/browse/TAJO-30?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13814657#comment-13814657
]
Dmitriy V. Ryaboy commented on TAJO-30:
---------------------------------------
Hi folks,
Just popping in to confirm that you do *not* need to take on a thrift
dependency to use Parquet.
Although internally the parquet format is described as thrift, the dependency
is not forced on consumers; it's an implementation details that does not leak
out of the Parquet APIs. Furthermore, we "shade" all the thrift classes, so
even if you do take a thrift dependency later on, and if it's a conflicting
version of thrift (some API-incompatible change or something like that), things
will just keep on working.
I think so far, we've managed to not have any non-parquet dependencies at all
for the core packages (when we do pull things in, we shade them; and we are
very picky about what we do pull in).
> Parquet Integration
> -------------------
>
> Key: TAJO-30
> URL: https://issues.apache.org/jira/browse/TAJO-30
> Project: Tajo
> Issue Type: New Feature
> Reporter: Hyunsik Choi
> Assignee: Dongmin Yu
> Labels: Parquet
>
> Parquet is very promising file format developed by twitter. We need to
> investigate the applicability of Parquet. If possible, we implement Parquet
> port.
> http://parquet.io/
--
This message was sent by Atlassian JIRA
(v6.1#6144)
