[
https://issues.apache.org/jira/browse/CASSANDRA-7498?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexey Zotov updated CASSANDRA-7498:
------------------------------------
Attachment: CASSANDRA-2.1-7498.txt
I have already created a patch before your comment.
The patch contains required changes for processing keys in sstable
export/import utilities. Regarding backward compatibility, I added an ability
to ignore the current key validator type to allow users import
previously-generated json files for "non-bytes" keys. For that users need to
add -Dskip.key.validator=true parameter during import.
> sstable2json exports keys in a wrong format
> -------------------------------------------
>
> Key: CASSANDRA-7498
> URL: https://issues.apache.org/jira/browse/CASSANDRA-7498
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Alexey Zotov
> Assignee: Alexey Zotov
> Labels: tools
> Attachments: CASSANDRA-2.1-7498.txt
>
>
> I examined code of _SSTableImport_ and _SSTableExport_ classes and found one
> thing that looks strange to me. Currently processing (encoding/decoding) of
> columns' names and values is performed based on CF Comparator and Value
> Validator correspondingly. But row keys are processed using _#bytesToHex_ and
> _#hexToBytes_ methods despite of there is Key Validator. Such approach works
> properly for "bytes" keys only. For other types it exports keys in a wrong
> format (i.e. they are exported in unreadable form).
> In the scope of this ticket we need to consider backward compatibility
> (import of previously exported sstables). For "bytes" keys it will work as
> previously since _#bytesToHex_ and _#hexToBytes_ methods are used in
> internals of _BytesType_ class. For other types we need to allow an ability
> to import data using "bytes" despite of configured validator.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
