[
https://issues.apache.org/jira/browse/IMPALA-9531?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17066761#comment-17066761
]
Gabor Kaszab commented on IMPALA-9531:
--------------------------------------
I've collected some TODO's in bullet point (there might be more to cover but
this is good to start with):
* Before starting the implementation create a test file with some valid and
some dateless timestamps for later testing purposes
* If format is provided in a string to datetime conversion then query can be
rejected in format parsing time.
* Cover both SimpleDate format and IsoSql format parsers. See *FormatTokenizer
and *FormatParser classes.
* Cover the case when reading timestamps from disc.
* Might also be somewhere around SimpleDateFormatParser but has to double
check.
* If not in above class then check how the scanners read timestamps (ORC,
Parquet, etc.)
* Check if setting ABORT_ON_ERROR has an effect on your implementation
* Re-run tests incrementally and check which one fails.
* Try to understand what the intention of the failing test was
* If the test doesn't mean to cover dateless timestamps then fix it.
* If it adds test coverage to dateless timestamps then drop it.
> Drop support for "dateless timestamps"
> --------------------------------------
>
> Key: IMPALA-9531
> URL: https://issues.apache.org/jira/browse/IMPALA-9531
> Project: IMPALA
> Issue Type: Improvement
> Components: Backend
> Reporter: Csaba Ringhofer
> Assignee: Adam Tamas
> Priority: Minor
>
> Currently cast("15:42:00" as timestamp) doesn't return NULL, but it should,
> as the resulting timestamps are more or less useless:
> - other fileformats than text do not support them
> - most built-in unctions don't support them - even those where it could make
> sense
> IMPALA-5942 contains more details about these issues.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
