Hey, Poetry 1.2 has made lots of improvements rece...
# development
h
Hey, Poetry 1.2 has made lots of improvements recently to issues that motivated us to use Pex lockfiles, including
--find-links
support, fixing
export
for "bifurcated requirements", and fixing the cache writing What do we think about permanently for at least now keeping
[python].lockfile_generator = {"pex", "poetry"}
? It's much more complex code for us, but it may ease incremental adoption because you can still use requirements.txt-style locks
👀 1
The main downside of Poetry lockfiles will always be performance -- we can't do subsetting as directly as Pex, and Pex will still end up doing a "resolve"
b
I can see leaving it in for a while, non-comittally, and seeing how things goes with the improvements. While reserving the right to pull it out if we feel that its slower performance is not in the best interests of Pants or its users. Maybe it becomes a permanent addition, maybe not, but it's nice that the improvement have made it possible for Pants to offer folks the flexibility of continuing to have that option a while longer at least.
👍 1
w
yea, fine with me.
relatedly: if we think that we can upgrade poetry this week, that would be ideal, because it would fix https://github.com/pantsbuild/pants/issues/15315, which would make
2.13.0a0
consumable by non-PEX users
b
I think long term we should "do one thing". If that is a barrier for entry we need to lower the barrier, not let people go around it 😛
b
Fair point. I just don't feel strongly about it right now. I think we've got time to let this play out.
b
oh yeah totally agree. Rug-pulling is a nasty business
h
if we think that we can upgrade poetry this week
They have not had their 1.2 stable release yet, so feels premature to bump Pants's default to 1.2.0b2