[
https://issues.apache.org/jira/browse/THRIFT-377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12688537#action_12688537
]
Joydeep Sen Sarma commented on THRIFT-377:
------------------------------------------
TFileTransport is quite different from TIOStream. The things that are
immediately visible are that it's framed (so allows for parallel processing and
fault recovery more easily), as well as that there's a support for concurrent
writers.
We have been using this internally - frankly i am not sure about all the
reasons why the publisher of the log wrote in this format - as a consumer from
java land - i have had to implement the same thing on the Java side. We have
had the Java read path code in production for a long time - but had little
reason to commit it to open source (no one else seemed to be using
TFileTransport).
The only reason for filing this now was that someone requested that we be able
to deal with TFileTransport in Hive (and that requires this Jira). Let me get
back to that person and see if they really want to use TFileTransport - or
would TIOStream suffice. If the latter does - we can close this.
> TFileTransport port in Java
> ---------------------------
>
> Key: THRIFT-377
> URL: https://issues.apache.org/jira/browse/THRIFT-377
> Project: Thrift
> Issue Type: New Feature
> Components: Library (Java)
> Reporter: Joydeep Sen Sarma
>
> there are environments in which data is being logged in TFileTransport
> (presumably C++ land) and data is desired to be consumed/imported into Hadoop.
> For this a Java version of TFileTransport is required. I am planning to post
> one that only reads the data - this code is already in use and tested and
> fulfils the current requirements.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
