You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A while back, we added conda packaging steps to zipline's CI for itself and its dependencies that were missing in the default anaconda channel. This work added time to CI, but enabled us to more easily release conda packages with each zipline release. The windows CI that was added ensured both that we could offer windows packages and that the tests passed on windows.
In more recent times, 1) conda-forge was built out and 2) we've released zipline less often. We've also found that our CI has rarely found packaging issues, except for missing dependencies.
We'd like to move our packaging automation to conda-forge to save our CI time and complexity. We'll keep our windows tests as part of zipline's and other repos' CI.
Some details:
PR and merge conda-forge feedstocks for zipline and its dependencies. A branch is in progress.
One dependency is an old version of bcolz. Either we should remove the upper bound that limits us to an old version and release a new version of zipline, or we should maintain a bcolz feedstock as well. I vote for new zipline release.
Another packaged-only-by-Quantopian dependency is lru-dict. We might remove this dependency instead of including it, since it requires a native build for each version of python.
Remove packaging from zipline's CI. This also requires changing the install steps not to rely on conda. Our GH actions migration would help with this.
The text was updated successfully, but these errors were encountered:
A while back, we added conda packaging steps to zipline's CI for itself and its dependencies that were missing in the default
anaconda
channel. This work added time to CI, but enabled us to more easily release conda packages with each zipline release. The windows CI that was added ensured both that we could offer windows packages and that the tests passed on windows.In more recent times, 1) conda-forge was built out and 2) we've released zipline less often. We've also found that our CI has rarely found packaging issues, except for missing dependencies.
We'd like to move our packaging automation to conda-forge to save our CI time and complexity. We'll keep our windows tests as part of zipline's and other repos' CI.
Some details:
One dependency is an old version of bcolz. Either we should remove the upper bound that limits us to an old version and release a new version of zipline, or we should maintain a bcolz feedstock as well. I vote for new zipline release.Another packaged-only-by-Quantopian dependency is lru-dict. We might remove this dependency instead of including it, since it requires a native build for each version of python.The text was updated successfully, but these errors were encountered: