[
https://issues.apache.org/jira/browse/HADOOP-3315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12593051#action_12593051
]
Owen O'Malley commented on HADOOP-3315:
---------------------------------------
I like the record count being in the tail. If we are saving the first key of
each block, we should probably also save the last key of the file too.
I would think this is just a new class that people migrate too. For most
applications, I would expect it to be an easy transition.
Srikanth, it is useful to have the magic bytes at the front so that commands
like "file" on unix can work. It is just a constant 4 bytes, it doesn't really
complicate the format at all.
You absolutely do want the record length in the data, because deserialization
can be slow.
I absolutely don't think there should be any special code for fixed width
types. The cost of the variable width types is *really* small with the vint
encoding.
> New binary file format
> ----------------------
>
> Key: HADOOP-3315
> URL: https://issues.apache.org/jira/browse/HADOOP-3315
> Project: Hadoop Core
> Issue Type: New Feature
> Components: io
> Reporter: Owen O'Malley
>
> SequenceFile's block compression format is too complex and requires 4 codecs
> to compress or decompress. It would be good to have a file format that only
> needs
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
