Fix wrong raise from dict nodes (v1.4.x)

[ramirezfranciscof]

Fixes #4454

When accessing an attribute as a dict (dictnode['key']), the error
raised was an AttributeError instead of a KeyError, which is rather
unusual. This has been fixed and a test that checks the correct error
raise has been added.

[codecov]

Codecov Report

Merging #4616 (d489fee) into support/1.4.x (c063609) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@                Coverage Diff                @@
##           support/1.4.x    #4616      +/-   ##
=================================================
+ Coverage          79.22%   79.22%   +0.01%     
=================================================
  Files                475      475              
  Lines              34828    34831       +3     
=================================================
+ Hits               27589    27593       +4     
+ Misses              7239     7238       -1     

Flag	Coverage Δ
django	73.07% <100.00%> (+0.01%)	⬆️
sqlalchemy	72.30% <100.00%> (+0.01%)	⬆️

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

Impacted Files	Coverage Δ
aiida/orm/nodes/data/dict.py	86.05% <100.00%> (+1.05%)	⬆️
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 c063609...d489fee. Read the comment docs.

[ramirezfranciscof]

@sphuber this can be merged in without the CI fix right?

[sphuber]

@sphuber this can be merged in without the CI fix right?

Prefer not to. I am about to merge the fix into release/1.5.1 so then you can rebase it and then I will merge it. That should then also be the last item for v1.5.1 so we could do the release. Could you prepare and do it?

[ramirezfranciscof]

So, I already rebased the one that goes into release/1.5.1 (here)but this is the one that goes into 1.4.4, I don't need to do any rebasing or anything for this right?

[sphuber]

So, I already rebased the one that goes into release/1.5.1 (here)but this is the one that goes into 1.4.4, I don't need to do any rebasing or anything for this right?

Ah damn, didn't realize this is another copy of the same PR. Well I think these tests would have failed as well and will start to fail once we merge it in support/1.4.x so we should actually put the same fix on that branch as well. I will do it now.

[ramirezfranciscof]

@sphuber Python 3.5 seems to have a problem with numpy

ERROR: No matching distribution found for numpy==1.19.4
Error: Process completed with exit code 1.

I'm not sure if maybe we need to copy this fix to python 3.5 as well??

    # Work-around issue caused by pymatgen's setup process, which will install the latest
    # numpy version (including release candidates) regardless of our actual specification:
    - name: Install numpy [py36]
      if: matrix.python-version == 3.6
      run: |
        pip install `grep 'numpy==' requirements/requirements-py-${{ matrix.python-version }}.txt`

[sphuber]

Good to go! Thanks