| From: | Jeff Davis <pgsql(at)j-davis(dot)com> |
|---|---|
| To: | Oleg Tselebrovskiy <o(dot)tselebrovskiy(at)postgrespro(dot)ru>, pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: Inconsistent string comparison using modified ICU collations |
| Date: | 2025-01-25 00:21:43 |
| Message-ID: | e55e987a4cf8fa36b5834951b10c43bc88e4b678.camel@j-davis.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Thu, 2025-01-23 at 13:29 -0800, Jeff Davis wrote:
> I had previously proposed[1] a GUC to control abbreviated keys, and
> it
> was rejected. Now that we've seen both a performance problem and a
> bug
> in the underlying dependency, perhaps we should reconsider.
Users may:
* trust libc strxfrm, or not
* trust ICU getSortKey & nextSortKeyPart, or not
* want to enable/disable abbreviated keys for testing (perf or
correctness)
And should we make the default the same as the current behavior --
trust ICU but not libc -- or should we trust neither by default?
We might need a few new GUCs here, unless someone has a better idea. I
suppose we could tie it to individual collations, but that seems worse.
Regards,
Jeff Davis
| From | Date | Subject | |
|---|---|---|---|
| Next Message | James Hunter | 2025-01-25 00:48:59 | Re: Proposal: "query_work_mem" GUC, to distribute working memory to the query's individual operators |
| Previous Message | Robins Tharakan | 2025-01-25 00:09:57 | Re: Convert sepgsql tests to TAP |