Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • 14:07
    zzzeek edited #5364
  • 14:07
    zzzeek unlabeled #5364
  • 14:07
    zzzeek unlabeled #5364
  • 14:07
    zzzeek milestoned #5364
  • 14:07
    zzzeek demilestoned #5364
  • 13:54
    zzzeek closed #5418
  • 13:45
    zzzeek milestoned #6810
  • 13:45
    zzzeek demilestoned #6810
  • 13:45
    zzzeek milestoned #6700
  • 13:45
    zzzeek demilestoned #6700
  • 13:45
    zzzeek milestoned #6505
  • 13:45
    zzzeek demilestoned #6505
  • 13:45
    zzzeek milestoned #6468
  • 13:45
    zzzeek demilestoned #6468
  • 13:44
    zzzeek milestoned #7135
  • 13:44
    zzzeek demilestoned #7135
  • 13:44
    zzzeek demilestoned #7035
  • 13:44
    zzzeek milestoned #7035
  • 13:39
    zzzeek milestoned #9154
  • 13:39
    zzzeek demilestoned #9154
Federico Caselli
@CaselIT
btw I saw today that django has also started supporting async orm
oh on this one: sqlalchemy/sqlalchemy#8340
why is registry not a weak key dict?
mike bayer
@zzzeek
it could have an option for that
Federico Caselli
@CaselIT
currently it's a dict
but I get the idea that most people just leave the session there and don't close it
mike bayer
@zzzeek
the key there can be anything, like an integer
so it cant be genrically weak key
IMO using scoped session w/ async, or really at all, is a bad idea
globals == bad
also yes they should ahve teardown code
so you can do this, by making your own weakref and then link it to remove
i will suggest that
Federico Caselli
@CaselIT
we should probably add a dragon regarding removing it once the scope is done
mike bayer
@zzzeek
ok my weakref idea does not work, correct
i dont like the ida of using weakrefs, someone using this (which they shouldnt) should make sure their task does .remove()
Federico Caselli
@CaselIT
I agree. so maybe a dragon?
Jonathan Vanasco
@jvanasco
there be dragons.
Federico Caselli
@CaselIT
there be async dragons in this case
Federico Caselli
@CaselIT

" send in the dragons on async_scoped_session " sqlalchemy/sqlalchemy@ac46321

ahahah

Alex Grönholm
@agronholm
yup, scoped sessions are problematic exactly because people don't think to close them...just the other day I was advising somebody who thought scoped sessions were "automatically managed" and didn't need to be closed.
2 replies
Federico Caselli
@CaselIT
@zzzeek the dataclasses stuff seems pretty easy to extend
mike bayer
@zzzeek
@CaselIT im sure it's not too hard as long as you arent doing slots
Federico Caselli
@CaselIT
yes, I think slots would require support in the orm
since you would need to add the _sa_... slot too
Federico Caselli
@CaselIT
but since there is basically no change compared with a plain class, I think we can live without supporting them
mike bayer
@zzzeek
I got ranges and multiranges working on asyncpg etc
was MUCH easier than I thought, psycopg has lots of complex stuff that isn't needed
@CaselIT re: slots, yes we'd need a big attribute change, since the InstanceState needs to have a "dict", so it would need to know how to get a "dict" from a slots class
@CaselIT all very doable, but a lot of work and also InstanceState kind of defeats the "performance" goal
@CaselIT having a "dict" over a slots class has to be a cython thing otherwise it's much slower than not using slots
Federico Caselli
@CaselIT
I'm really not sure it's worth it, since the performance change is negligible compared with a normal class
mike bayer
@zzzeek
just tried to release and twine upload seems to be silently failing
Jonathan Vanasco
@jvanasco
is it maybe from that new "protected projects" thing?
mike bayer
@zzzeek
@jvanasco seems very likely
pypi/warehouse#12041
Federico Caselli
@CaselIT
Strange, we use a token to upload
the github action haa still not run. lets see if that fails, but in any case its strange that twine is happy but nothing happens
Federico Caselli
@CaselIT
I think there is something strange going on on pypi side. This errors https://test.pypi.org/project/SQLAlchemy/
Federico Caselli
@CaselIT
It seem to be working now
mike bayer
@zzzeek
they pushed a front-end thing that broke the view
so we were seeing cdn cached
and twine upload happily says nothing if file already exists
Federico Caselli
@CaselIT
Well at least it's not our fault
mike bayer
@zzzeek
nope
it's jsut twine's silent failure made me unsure if upload was just afiling
instead of dumb UX bug
Federico Caselli
@CaselIT
Still twine could say something when the file already exists even if it still exists with success
mike bayer
@zzzeek
yes
Federico Caselli
@CaselIT
we seem to have comments also on mssql