[
https://issues.apache.org/jira/browse/CALCITE-5864?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17746979#comment-17746979
]
Jonathan A Sternberg commented on CALCITE-5864:
-----------------------------------------------
No, the number of milliseconds in a quarter (or month) are not well-defined. My
apologies for the confusion. My original intention was that week reports the
wrong number of milliseconds and quarter returns the wrong number of months and
I conflated the two in my title. I'll see if I can modify it.
> getValueAs reports the wrong number of milliseconds for QUARTER and WEEK
> ------------------------------------------------------------------------
>
> Key: CALCITE-5864
> URL: https://issues.apache.org/jira/browse/CALCITE-5864
> Project: Calcite
> Issue Type: Bug
> Components: core
> Reporter: Jonathan A Sternberg
> Priority: Major
> Labels: pull-request-available
>
> The change that introduced WEEK and QUARTER processing here did not correctly
> implement the `evaluateIntervalLiteralAsX` functions for these two time units.
> For both of them, it used the `fillIntervalValueArray` with the week
> parameter being used for months and the quarter parameter being used for
> months with no modifications to the underlying values.
> For weeks, this results in the field being misused as the hour parameter so 2
> weeks becomes 2 hours. For quarters, this results in the field being misused
> as months so 2 quarters becomes 2 months.
> I believe the proper way to implement these is to perform the modification so
> weeks gets translated to 7 days and quarters gets translated to 3 months.
> This seems to only affect the string versions such as `INTERVAL '2' WEEK`.
> The integer variants don't seem to use the literal pathway so they aren't
> affected.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
