Re: Crash report for some ICU-52 (debian8) COLLATE and work_mem values
Peter Geoghegan <pg@bowt.ie>
From: Peter Geoghegan <pg@bowt.ie>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Peter Eisentraut <peter.eisentraut@2ndquadrant.com>,
Robert Haas <robertmhaas@gmail.com>, Daniel Verite <daniel@manitou-mail.org>, PostgreSQL mailing lists <pgsql-bugs@postgresql.org>
Date: 2017-08-14T18:14:59Z
Lists: pgsql-bugs, pgsql-hackers
On Mon, Aug 14, 2017 at 11:00 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote: >> I must admit that I missed that the description was cataloged as a >> comment. Is it out of the question to add another column to >> pg_collation, that just always has the ICU-provided description? These >> seem like separate concerns to me. > > That seems like a reasonable solution to me, if it's not too late > for another catversion bump. I like it better than the auto-comment > thing Peter E. suggests nearby. Another advantage of doing a catversion bump for v10 is that it lets us add a new SQL-callable function (or two). I would like to add a function to solve some of our discoverability problems around variant collations. We now agree that we should not add variants at initdb time, but Peter E's concern about where that leaves the discoverability of the variants (phonebook, emoji, pinyin, and so on) is a concern that I share. Similarly, it would also be helpful if users could inquire about both ICU version, and the corresponding CLDR version, using a new view. That way, they could easily find the right CLDR version/standard, which is where all the customization stuff is actually documented. Any thoughts on that? -- Peter Geoghegan
Commits
-
Reject use of ucol_strcollUTF8() before ICU 53
- d6391b03b302 10.0 landed
-
Rethink behavior of pg_import_system_collations().
- 0b13b2a7712b 10.0 cited
-
Faster expression evaluation and targetlist projection.
- b8d7f053c5c2 10.0 cited