Unpin numpy version from 1.17.X to allow the newest version (1.19)

[CasperWA]

This is an attempt to try and unpin the hard constraint on numpy.

This was introduced in f5d6cba, the reason for it can be seen in the commit message.

I have tried following the wiki page on how to update dependencies, and am currently trying the semi-automatic update of the requirements files by creating this PR.

[csadorf]

According to our current compatibility timeline we must support numpy version 1.17 until 26-Jul-2021. The compatiblity timeline is derived from AEP 003.

Please note that the 26-Jul-2021 date is not explicitly mentioned in AEP 003, but required to be compatible with NEP 29.

Edit: This means the requirement should be specified as: numpy~=1.17.

[codecov]

Codecov Report

Merging #4378 into develop will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop    #4378      +/-   ##
===========================================
+ Coverage    79.33%   79.33%   +0.01%     
===========================================
  Files          468      468              
  Lines        34755    34757       +2     
===========================================
+ Hits         27568    27571       +3     
+ Misses        7187     7186       -1     

Flag	Coverage Δ
#django	72.96% <100.00%> (+0.01%)	⬆️
#sqlalchemy	72.16% <100.00%> (-<0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
aiida/tools/data/array/kpoints/legacy.py	31.40% <100.00%> (+0.20%)	⬆️
aiida/engine/daemon/runner.py	82.76% <0.00%> (+3.45%)	⬆️

---

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 c6bca06...0da7e1b. Read the comment docs.

[csadorf]

Please update the requirements/*.txt files according to the following table:

• 3.5: numpy==1.18.5
• 3.6: numpy==1.19.2
• 3.7: numpy==1.19.2
• 3.8: numpy==1.19.2

[csadorf]

It appears that there is an actual incompatibility:

tests/test_dataclasses.py:3203: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
aiida/tools/data/array/kpoints/__init__.py:87: in get_explicit_kpoints_path
    return method(structure, **kwargs)
aiida/tools/data/array/kpoints/__init__.py:216: in _legacy_get_explicit_kpoints_path
    cell=structure.cell, pbc=structure.pbc, **kwargs
aiida/tools/data/array/kpoints/legacy.py:353: in get_explicit_kpoints_path
    numpy.linspace(ini_coord[0], end_coord[0], num_points[count_piece]),
<__array_function__ internals>:6: in linspace
    ???
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

My assumption is that the third argument in the call to numpy.linspace is a float instead of an integer here:

aiida-core/aiida/tools/data/array/kpoints/legacy.py

Line 353 in b002a9a

numpy.linspace(ini_coord[0], end_coord[0], num_points[count_piece]),

[CasperWA]

It appears that there is an actual incompatibility:

tests/test_dataclasses.py:3203: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
aiida/tools/data/array/kpoints/__init__.py:87: in get_explicit_kpoints_path
    return method(structure, **kwargs)
aiida/tools/data/array/kpoints/__init__.py:216: in _legacy_get_explicit_kpoints_path
    cell=structure.cell, pbc=structure.pbc, **kwargs
aiida/tools/data/array/kpoints/legacy.py:353: in get_explicit_kpoints_path
    numpy.linspace(ini_coord[0], end_coord[0], num_points[count_piece]),
<__array_function__ internals>:6: in linspace
    ???
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

My assumption is that the third argument in the call to numpy.linspace is a float instead of an integer here:

aiida-core/aiida/tools/data/array/kpoints/legacy.py

Line 353 in b002a9a

numpy.linspace(ini_coord[0], end_coord[0], num_points[count_piece]),

Yes, already looked into this and just pushed a possible fix.

[CasperWA]

Please update the requirements/*.txt files according to the following table:

• 3.5: numpy==1.18.5
• 3.6: numpy==1.19.2
• 3.7: numpy==1.19.2
• 3.8: numpy==1.19.2

Just to understand. Did you write this comment manually or is this part of the semi-automatic workflow?

[csadorf]

Please update the requirements/*.txt files according to the following table:

• 3.5: numpy==1.18.5
• 3.6: numpy==1.19.2
• 3.7: numpy==1.19.2
• 3.8: numpy==1.19.2

Just to understand. Did you write this comment manually or is this part of the semi-automatic workflow?

I checked those manually. The automated workflow creates a PR if and only if the tests succeed and also only if the requirements files are outdated, which with the change to numpy~=1.17 they actually are not. So we actually don't have to change them and now that I think about it, maybe we shouldn't.

[csadorf]

Looks good to me. Codecov is complaining about the coverage, I'm ok to just ignore that, although we might need to check with someone else whether we can override that check.

But before we do that, maybe you can give it a 5min try of providing the faulty input to cover the section?

[CasperWA]

@csadorf and @sphuber This is ready for a re-review + merge. Your old review comments have all been addressed.

Implemented.

[sphuber]

All good, thanks @CasperWA