[
https://issues.apache.org/jira/browse/HIVE-6316?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14010341#comment-14010341
]
Eugene Koifman commented on HIVE-6316:
--------------------------------------
[~leftylev],
Null and Throw are the only possible values. The description of HIVE-5814 has
a usage example:
{noformat}
HCatStorer('','', '-onOutOfRangeValue Throw')
{noformat}
hcat.pig.store.onoutofrangevalue does NOT need to be documented, it's internal.
This only applies when using HCat fro Pig, where the user is expected to use
'onOutOfRangeValue in HCatStorer. Is not really related to Data Promotion
Behavior.
The HCatInputFormat and HCatOutputFormat section need the same update to the
"type mapping" tables as HCatLoader/HCatStorer. I think it would be easier to
just create link from all 4 current tables to a single page that has the whole
table in
https://issues.apache.org/jira/secure/attachment/12626251/HCat-Pig%20Type%20Mapping%20Hive%200.13.pdf
exactly. The headers in the table actually indicate a mapping of Hive
Type/Value system to Pig Type/Value system.
Logically speaking there is no such thing as HCatalog type/value system.
HCatalog connects Hive tables to Pig/Map Reduce. Pig has it's own type/value
system; MR does not as such and is expected to use (in HCatRecord) the same
classes as used in Hive internally.
so the data type mapping is really Hive<->Pig (HCatLoader/Storer) and Hive<->MR
(HCatInput/OutputFormat) which is why it's all summarized in a single table in
my document.
> Document support for new types in HCat
> --------------------------------------
>
> Key: HIVE-6316
> URL: https://issues.apache.org/jira/browse/HIVE-6316
> Project: Hive
> Issue Type: Sub-task
> Components: Documentation, HCatalog
> Affects Versions: 0.13.0
> Reporter: Eugene Koifman
> Assignee: Lefty Leverenz
>
> HIVE-5814 added support for new types in HCat. The PDF file in that bug
> explains exactly how these map to Pig types. This should be added to the
> Wiki somewhere (probably here
> https://cwiki.apache.org/confluence/display/Hive/HCatalog+LoadStore).
> In particular it should be highlighted that copying data from Hive TIMESTAMP
> to Pig DATETIME, any 'nanos' in the timestamp will be lost. Also, HCatStorer
> now takes new parameter which is described in the PDF doc.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
