but the note here "The “threadlocal” feature is generally discouraged. It’s designed for a particular pattern of usage which is generally considered as a legacy pattern." makes me think twice
2015-10-15 28835, 2015
alastairp
the other thing is that there is only one case where this is a problem currently, and they're both SELECTs, so there's no issues with transactions
2015-10-15 28839, 2015
ruaok
it is exactly this type of stuff that makes me dislike ORMs and SQLalchemy in particular.
2015-10-15 28847, 2015
alastairp
ruaok: we're not using ORMs
2015-10-15 28809, 2015
ruaok
s/ORM/overly clever db abstractions/g
2015-10-15 28812, 2015
alastairp
theis issue was also present when we were using psycopg2
2015-10-15 28817, 2015
Gentlecat
and it wasn't better before
2015-10-15 28820, 2015
Gentlecat
right
2015-10-15 28827, 2015
alastairp
we just put our heads in the sand and hoped it wouldn't be a problem
2015-10-15 28830, 2015
alastairp
I'm trying to fix that
2015-10-15 28853, 2015
alastairp
s/ORMs/webapps/
2015-10-15 28806, 2015
ruaok
this feels like shoveling it to the a different part of the sandbox and sticking your head into a different part of the sandbox.
2015-10-15 28823, 2015
RJ2 joined the channel
2015-10-15 28837, 2015
ruaok
and yes, in this case transactions don't come into the game, unless you need 100% garanteed data consistency between these two calls.
2015-10-15 28854, 2015
ruaok
in which case you would use a transaction in order to use the isolation feature.
2015-10-15 28851, 2015
alastairp
right. I think this is what I'm getting to
2015-10-15 28806, 2015
D4RK-PH0ENiX has quit
2015-10-15 28848, 2015
alastairp
where we don't need a transaction for this specific case, so we shouldn't try and make a generic solution for a problem we don't have
2015-10-15 28817, 2015
ruaok
yes, but you will likely run into a problem in the future, so having a plan for it is good.
2015-10-15 28844, 2015
Gentlecat
alastairp: maybe another way to think about that is that each interface/function in the db package should be isolated
2015-10-15 28813, 2015
alastairp
that doesn't make sense for duplication of code, though
2015-10-15 28843, 2015
alastairp
we will end up having "get lowlevel data" written 4 times in the module so that we can keep code within a function
2015-10-15 28855, 2015
alastairp
(this is the problem that I've been trying to fix, it was already duplicated twice)
2015-10-15 28840, 2015
Gentlecat
why would we have it 4 times?
2015-10-15 28852, 2015
alastairp
well, we already had it twice
2015-10-15 28810, 2015
alastairp
get_low_level, and get_summary - which loaded lowlevel and highlevel
2015-10-15 28827, 2015
alastairp
I could see there being another case where we had to get lowlevel data to compare it to something else
2015-10-15 28851, 2015
alastairp
hmm
2015-10-15 28802, 2015
alastairp
I'm looking at what I did with messybrainz - I pass a connection to all database methods
2015-10-15 28830, 2015
alastairp
which helps make it explicit. I think I just did it at that time to make it work as quickly as possible
ah, ruaok was smart in the HL extractor, all the database work is done on the mainthread
2015-10-15 28824, 2015
ruaok
smart? don't confuse laziness for intelligence.
2015-10-15 28842, 2015
alastairp
perfect. I'll keep believing you're smart
2015-10-15 28854, 2015
ruaok
keep?
2015-10-15 28805, 2015
alastairp
<- optimist
2015-10-15 28829, 2015
larsen has quit
2015-10-15 28804, 2015
larsen joined the channel
2015-10-15 28810, 2015
LordSputnik joined the channel
2015-10-15 28826, 2015
samphippen joined the channel
2015-10-15 28854, 2015
LordSputnik has quit
2015-10-15 28857, 2015
legoktm
good morning o/
2015-10-15 28859, 2015
LordSputnik joined the channel
2015-10-15 28855, 2015
legoktm
ruaok: I did most of the easy mysql schema changes last night, when I tried to do the larger ones with pt-online-schema-change, it complained the mediawiki mysql user didn't have the PROCESS right. Is it alright if I create a new mysql user specifically for pt-osc with that right?
2015-10-15 28810, 2015
legoktm
(and whatever other rights are needed)
2015-10-15 28828, 2015
ruaok
sure.
2015-10-15 28843, 2015
ruaok
that VM is your domain, feel free to do things as you need to.
2015-10-15 28814, 2015
alastairp
OK. We have a mid-term solution that we'll go for
2015-10-15 28833, 2015
alastairp
but I'd still like to talk to someone who actually knows what they're doing
2015-10-15 28836, 2015
ruaok bets luks does
2015-10-15 28833, 2015
alastairp
for now we're going to open explicit transactions in the methods we know require them (write lowleve, write highlevel, some dataset methods)
2015-10-15 28800, 2015
D4RK-PH0ENiX joined the channel
2015-10-15 28829, 2015
D4RK-PH0ENiX has quit
2015-10-15 28835, 2015
diana_olhovik_ has quit
2015-10-15 28858, 2015
D4RK-PH0ENiX joined the channel
2015-10-15 28809, 2015
D4RK-PH0ENiX has quit
2015-10-15 28817, 2015
D4RK-PH0ENiX joined the channel
2015-10-15 28841, 2015
ruaok
are you a fan of Kraftwerk? Love "Autobahn"? If so, this is a must for you:
07:23 < ruaok> it is exactly this type of stuff that makes me dislike ORMs and SQLalchemy in particular.
2015-10-15 28818, 2015
darwin
if you are a small team, the overhead of ORM is sometimes worse than what it cures
2015-10-15 28828, 2015
darwin
if you are a larger team, you really don't want devs wrting SQL
2015-10-15 28836, 2015
ruaok
darwin: makes sense to me.
2015-10-15 28846, 2015
ruaok
Leftmost: ping
2015-10-15 28824, 2015
ruaok
LordSputnik: ping
2015-10-15 28857, 2015
LordSputnik
ruaok: pong
2015-10-15 28818, 2015
ruaok
I'm trying to order stickers for all of the projects.
2015-10-15 28823, 2015
ruaok
what do I do about the BB stickers?
2015-10-15 28828, 2015
LordSputnik
Keep brown
2015-10-15 28836, 2015
ruaok
ok
2015-10-15 28807, 2015
ruaok
3"x3" stickers. Die cut (in the hexagon shape). Of MB, AB, LB, picard, CB, MeB and BB. yes, no?
2015-10-15 28801, 2015
LordSputnik
Do they do custom T-shirts as well as stickers?
2015-10-15 28832, 2015
ruaok
don't think so.
2015-10-15 28836, 2015
ruaok
bitmap: are you around?
2015-10-15 28822, 2015
ruaok has quit
2015-10-15 28840, 2015
yeeeargh joined the channel
2015-10-15 28822, 2015
JesseW joined the channel
2015-10-15 28805, 2015
ruaok joined the channel
2015-10-15 28847, 2015
JesseW has quit
2015-10-15 28847, 2015
alastairp
ruaok: we can move lowlevel to metadata/json tables too!
2015-10-15 28813, 2015
ruaok
clean up lots of sins. :)
2015-10-15 28829, 2015
alastairp
darwin: I'm a co-maintainer on an ORM-based app written by people who don't understand relational databases
2015-10-15 28833, 2015
alastairp
it shows
2015-10-15 28858, 2015
alastairp
I'm really glad that I started development with sql, then started with ORMs
2015-10-15 28807, 2015
alastairp
I'm finally starting to learn the middleground
2015-10-15 28815, 2015
darwin
alastairp: haha, that sounds... fun.
2015-10-15 28853, 2015
alastairp
darwin: newrelic free trial + some indexes -> postgres cpu usage from 90% to 15%
2015-10-15 28814, 2015
reosarevok joined the channel
2015-10-15 28814, 2015
reosarevok has quit
2015-10-15 28814, 2015
reosarevok joined the channel
2015-10-15 28849, 2015
drsaunders
fyi i think there may be a server issue atm
2015-10-15 28815, 2015
ruaok
thanks will look.
2015-10-15 28823, 2015
ruaok
zas: you around?
2015-10-15 28832, 2015
ruaok
here is your first chance to save the day! :)
2015-10-15 28853, 2015
ruaok
uh oh.
2015-10-15 28803, 2015
ruaok
corosync on ernie ia 100% cpu load.
2015-10-15 28847, 2015
bitmap
ruaok: I'm here now
2015-10-15 28834, 2015
ruaok
yeah, I was wondering about sending stickers to you and have you bring them, but you leave too soon.
2015-10-15 28838, 2015
ruaok
I'll have to pester Leftmost
2015-10-15 28853, 2015
bitmap
ah, right
2015-10-15 28834, 2015
bitmap
zas: it looks like you're already an owner for the metbrainz org, and the 'musicbrainz' account password is in syswiki, not sure there's anything else I can do
2015-10-15 28859, 2015
bitmap
I don't see a way to make someone admin of a repo, if that exists