not sure I care this much to dig further. it is clear that this is on the edge of undefined behaviour and changing the script to explicitly define the behaviour seems like the way to go.
2020-11-04 30921, 2020
alastairp
> In the past, Postgres was inconsistent in this area: you would get the desirable behavior if the ordering were performed via an indexscan, but not if it had to be done by an explicit sort step.
2020-11-04 30924, 2020
alastairp
yeah, that's interesting
2020-11-04 30905, 2020
alastairp
I think that you're right. either do it explicitly in python (if it can handle that much data), or your subquery of subquery idea
2020-11-04 30928, 2020
ruaok
I doubt the subquery of subquery will be stable either.
2020-11-04 30949, 2020
ruaok
and I am literally selecting 1 rowid for insertion into my table. zero problems for python.
2020-11-04 30903, 2020
alastairp
oh, I thought that this was acting on millions of rows. fine then
2020-11-04 30937, 2020
ruaok
yeah, its a few million, but the result is very skinny, so no biggie
2020-11-04 30901, 2020
alastairp
mmm
2020-11-04 30905, 2020
alastairp
I'm just trying to understand the query. It's returning all release ids, but in an order where the first occurrence of a release is probably the one that you want (because of media, date, etc)?
2020-11-04 30924, 2020
alastairp
and the table has columns (id, release), and you're inserting into release?
2020-11-04 30949, 2020
ruaok
yes to all.
2020-11-04 30900, 2020
ruaok
id is SERIAL
2020-11-04 30904, 2020
alastairp
and you tried explicitly inserting nextval of the id sequence as the id?
2020-11-04 30912, 2020
alastairp
what about using a window function as id?
2020-11-04 30920, 2020
ruaok
yes, it retains the incorrect sequence as it does without.
2020-11-04 30944, 2020
alastairp
so you no longer use the sequence, you just explicitly number all of the rows, and insert _that_ number as id
2020-11-04 30947, 2020
ruaok
window functions need a sort order to go by, but my sort order is quite nuts and I've not gotten that to work either.
2020-11-04 30955, 2020
alastairp
got it
2020-11-04 30905, 2020
ruaok
alastairp: that was my proposed fix, but no luck so far.
2020-11-04 30914, 2020
alastairp
so. use python then :)
2020-11-04 30921, 2020
ruaok
ding.
2020-11-04 30906, 2020
v6lur has quit
2020-11-04 30932, 2020
v6lur joined the channel
2020-11-04 30951, 2020
alastairp
ok. radiator is no longer dripping water, I think I'd call that a successful day
2020-11-04 30920, 2020
v6lur has quit
2020-11-04 30946, 2020
BrainzGit
[musicbrainz-server] reosarevok merged pull request #1773 (master…update-AllowAmending-grace-period): Align comments with new code (very minor low priority PR) https://github.com/metabrainz/musicbrainz-server/…
2020-11-04 30904, 2020
reosarevok
bitmap, yvanzo: hi!
2020-11-04 30911, 2020
BrainzGit
[musicbrainz-server] mwiencek opened pull request #1776 (master…mbs-11092): Switch to SameSite=Strict for login cookies and preserve POST parameters on login (MBS-4555, MBS-11092) https://github.com/metabrainz/musicbrainz-server/…
I would also prefer to distinguish actually submitted fields from others in edit display in general, when possible.
2020-11-04 30902, 2020
reosarevok
Ok, I can submit a change for that into master :)
2020-11-04 30946, 2020
reosarevok
yvanzo: re that last linked PR: is it ok to remove /gakki in that situation? I wouldn't mind adding the three options to the regex in URL.pm either, but IIRC we'd need three separate lines for constants.js?
2020-11-04 30902, 2020
reosarevok
(you suggested adding it at first, so thought I'd explicitly ask :) )
2020-11-04 30925, 2020
niceplace has quit
2020-11-04 30932, 2020
yvanzo
reosarevok: why was the path specified in the first place?
2020-11-04 30948, 2020
reosarevok
The site has stuff that seems unrelated to this
2020-11-04 30906, 2020
reosarevok
"Saisai Batake is a graphic tool dedicated to creating small images that uniquely color your homepage ."
2020-11-04 30910, 2020
reosarevok
is the main page :)
2020-11-04 30918, 2020
yvanzo
oh ok :)
2020-11-04 30919, 2020
reosarevok
It's just that the person also runs an instrument DB in the same domain
