Now handling all constraint notations in unpinned enviroment.yamls

[vringar]

@birdsarah mentioned in #858 (comment) that we don't pin dependencies specified by < or >.

This PR fixes that inconsistency and guarantees that the last not_pip dependency will get pinned if there are no pip dependencies

[birdsarah]

This looks fine from a cursory look. If repin works without changing the pinned env file with the new < then I'd say we're good to go.

What this is highlighting for me is how gross this is.

I built it to match the infrastructure we were replacing. I can't remember the logic for wanting a pruned environment file.

We could switch to this kind of thing which would let us have a similar setup but without this prune step which just feels like a maintenance headache.

[vringar]

I can't remember either why we didn't want full pinning
Paging @englehardt for context

[vringar]

This looks fine from a cursory look. If repin works without changing the pinned env file with the new < then I'd say we're good to go.

I ran ./repin.sh and nothing changed.

We could switch to this kind of thing which would let us have a similar setup but without this prune step which just feels like a maintenance headache.

These conda environments seem to be specific to the OS and Processor used. So I'm hesitant to use this.
I think we can just not do the dependency pruning after running conda env export

[birdsarah]

This looks fine from a cursory look. If repin works without changing the pinned env file with the new < then I'd say we're good to go.

I ran ./repin.sh and nothing changed.

We could switch to this kind of thing which would let us have a similar setup but without this prune step which just feels like a maintenance headache.

These conda environments seem to be specific to the OS and Processor used. So I'm hesitant to use this.
I think we can just not do the dependency pruning after running conda env export

YES! This is exactly what I ran into before. Thanks for jogging my memory.

[englehardt]

I can't remember either why we didn't want full pinning
Paging @englehardt for context

I prefer full pinning. IIRC not pruning would lead to issues. Maybe @birdsarah can speak to the motivation for implementing pruning?

[vringar]

Seeing as she said:

I built it to match the infrastructure we were replacing. I can't remember the logic for wanting a pruned environment file.

I don't think she can speak to why the pruning was implemented.
Do we just want to drop the pruning and see if we encounter any issues? Or should I try to find the PR and the associated issue for introducing conda and try to figure out why we were pruning?

[englehardt]

Seeing as she said:

I built it to match the infrastructure we were replacing. I can't remember the logic for wanting a pruned environment file.

I don't think she can speak to why the pruning was implemented.
Do we just want to drop the pruning and see if we encounter any issues? Or should I try to find the PR and the associated issue for introducing conda and try to figure out why we were pruning?

Check out the May 14th commits here: #648

It sounds like it might have been OS differences in dependencies? As in maybe plyvel requires one dependency on Linux and a different one on MacOS.

[vringar]

Also dependencies on XServer are probably not satisfiable under MacOS

[codecov]

Codecov Report

Merging #860 (1db7bc0) into master (9aaa3c0) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #860   +/-   ##
=======================================
  Coverage   49.66%   49.66%           
=======================================
  Files          34       34           
  Lines        3306     3306           
=======================================
  Hits         1642     1642           
  Misses       1664     1664           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9aaa3c0...1db7bc0. Read the comment docs.

[boolean5]

This is ready to merge, right?

[vringar]

Yes