[
https://issues.apache.org/jira/browse/BEAM-9179?focusedWorklogId=459468&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-459468
]
ASF GitHub Bot logged work on BEAM-9179:
----------------------------------------
Author: ASF GitHub Bot
Created on: 15/Jul/20 18:38
Start Date: 15/Jul/20 18:38
Worklog Time Spent: 10m
Work Description: robinyqiu commented on pull request #12247:
URL: https://github.com/apache/beam/pull/12247#issuecomment-658934753
> I know of at least one customer use case where they are disabling
BeamZetaSqlCalcRel so they can avoid JNI. We also have plans to turn
BeamCalcRel back on for simple operators to enable better performance.
That make sense. Thanks for sharing this information. But I do believe it
makes our code cleaner to move those files which are only used on the
BeamZetaSqlCalcRel disabled path together. So how about I move those code to a
new `/impl` subdirectory, instead of deleting them?
> I believe the preferred method is git pull --rebase or git rebase
upstream/master.
Sure, I can do that.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 459468)
Time Spent: 6.5h (was: 6h 20m)
> Refactor Beam ZetaSQL type translation code
> -------------------------------------------
>
> Key: BEAM-9179
> URL: https://issues.apache.org/jira/browse/BEAM-9179
> Project: Beam
> Issue Type: Improvement
> Components: dsl-sql-zetasql
> Reporter: Robin Qiu
> Assignee: Robin Qiu
> Priority: P2
> Time Spent: 6.5h
> Remaining Estimate: 0h
>
> Currently, how types work in Beam ZetaSQL is very confusing: code that does
> type conversion between Calcite/Beam/ZetaSQL type systems are everywhere and
> is likely to be not consistent. In addition, the recent migration to
> BeamZetaSqlCalcRel makes the situation even worse: it adds another layer for
> type conversion and makes a lot of code (for making ZetaSQL functions
> manually) obsolete. This makes debugging and supporting new features very
> hard. A few examples:
> * Support new ZetaSQL types (e.g. DATE/TIME/DATETIME)
> * Fix the long-standing precision & timezone issue of TIMESTMAP type
> * Support BeamSQL query parameters
> Therefore, I propose we work on the following items to make the story of Beam
> ZetaSQL type clearer:
> * Define a clear mapping between Calcite/Beam/ZetaSQL types and values
> * Refactor the codebase to move all type conversion code to a single source
> of truth
> * Add thorough testing for it
> * Remove obsolete code that is added to make functions work before
> BeamZetaSqlCalcRel
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
