Fixed build errors due to recent changes (homebrew/pip)

[starrify]

No description provided.

[codecov]

Codecov Report

Merging #376 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #376      +/-   ##
==========================================
+ Coverage   91.49%   91.49%   +<.01%     
==========================================
  Files          34       34              
  Lines        2362     2364       +2     
==========================================
+ Hits         2161     2163       +2     
  Misses        201      201

Impacted Files	Coverage Δ
shub/utils.py	90.47% <100%> (+0.03%)	⬆️

---

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 2871384...97acf9f. Read the comment docs.

[starrify]

It's fixed finally. I initially assumed there was only one error. 🤦‍♀️

[vshlapakov]

@starrify Thanks for looking into this 🙌

I'm not surprised by the homebrew issue, it adds some complexities from time to time. However, in a case of changes in the recent pip, I have a feeling that it'd be easier to update our imports instead of restricting pip version - many people update it regularly and I'd like to avoid the version limitation if possible. Do you agree or there's something I'm missing?

[vshlapakov]

Good catch!

[vshlapakov]

It seems that in the recent pip, the main function was moved from pip._internal.init to pip._internal.main. It's not the best idea to rely on such internal functions, but I'd say there's a simpler way to fix this without restrictions on pip version

try:
    from pip import main as pip_main
except:
    try:
        from pip._internal.main import main as pip_main
    except:
        from pip._internal import main as pip_main

[starrify]

Yeah this is good observation! I should have looked into this. Though I'm not sure for now how to write tests for this. 😅

[vshlapakov]

I guess these should be uncommented?

[starrify]

The same file has requests>=2.20.0 and pyyaml>=4.2b1 below which may lead to double requirements issues. 🤔

[starrify]

Hi @vshlapakov , thanks for the review! I was thinking of some quick (and maybe temporary) fixes for the broken CI when creating this pull request, since it affects also pull requests. I didn't look much into pip's changes, assuming someone (maybe I myself) would soon provide a better approach than restricting the version -- but back then I intended to have the failed CIs resolved ASAP.