You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@parquet.apache.org by "Wes McKinney (JIRA)" <ji...@apache.org> on 2018/11/12 21:51:00 UTC
[jira] [Updated] (PARQUET-1056) [C++] Add function to snapshot
FileMetaData to a separate file, in case writing a row group fails, then
prior row groups can be recovered
[ https://issues.apache.org/jira/browse/PARQUET-1056?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Wes McKinney updated PARQUET-1056:
----------------------------------
Summary: [C++] Add function to snapshot FileMetaData to a separate file, in case writing a row group fails, then prior row groups can be recovered (was: Partial Reading when process crash)
> [C++] Add function to snapshot FileMetaData to a separate file, in case writing a row group fails, then prior row groups can be recovered
> ------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: PARQUET-1056
> URL: https://issues.apache.org/jira/browse/PARQUET-1056
> Project: Parquet
> Issue Type: Wish
> Components: parquet-cpp
> Affects Versions: cpp-1.1.0
> Reporter: yugu
> Priority: Major
>
> questions are actually twofold:
> 1. is there a way to write data column by column ?
> I see a columnwriter but seems that's for arrow..
> 2. was wondering if there is a way to preserve the footer information once in a while, in case the process crashes. for now if the proc crash the parquet file become unreadable...
> In the code the close function (for parquetfilewriter) basically just flushes and close the stream. maybe a separate api for flush footer to target file?
> (or maybe there is an existing way for making things robust I missed)
> thanks!
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)