Make BlueCellulab NEURON 9.0+ compatible

[darshanmandge]

• Solves Make compatible with NEURON9.0 #41.
• Required for Update bluecellulab to v2.6.83 and libsonata v0.1.34 for SONATA parameter modification  Bluenaas#123

Main Changes

Based on NEURON 9.0 release notes: https://github.com/neuronsimulator/nrn/releases/tag/9.0.0, the following main changes were made:

• void** --> IvocVect** in DetAMPANMDA, DetGABAAB, ProbGABAAB_EMS, GluSynapse, and VecStim MOD files (both tests/mechanisms/ and examples/mechanisms/)
• RNG-related void** patterns kept as-is (explicit casts, valid C++)
• Replace deprecated neuron.run() with neuron.h.continuerun() in test_ballstick.py

Update test tolerances for NEURON 9.0 numerical differences

NEURON 9.0 release notes mention:

Usage of the Eigen library could introduce floating-point differences, but they are very small, validated and can be ignored. This required changes in relative tolerance for some of the tests.

• Eigen library replacing Meschach introduces small FP differences (tightened to rel=5e-4 to abs=5e-3 for affected assertions)
• RNG implementation changes produce different random sequences (relaxed tolerances for RNG-dependent test assertions)

Test reference files:

• The use of the Eigen library instead of Meschach (as in previous versions) also leads to floating-point differences in the h5 files generated by some tests.
• Hence, 10 SONATA H5 report output files in test needed an update based on simulation with NEURON 9.0.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
tests/test_analysis/test_analysis.py	98.65% <100.00%> (ø)
tests/test_ballstick.py	86.41% <100.00%> (ø)
tests/test_cell/test_injector.py	98.72% <ø> (ø)
tests/test_cell/test_random.py	100.00% <100.00%> (ø)
tests/test_tools.py	98.86% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[AurelienJaquier]

Very nice! Have you tried to run the examples with NEURON 9 ?

[darshanmandge]

I just ran the examples manually to see the figures generated in each notebook. They look fine at first glance.

The tests would have also run the examples.

[WeinaJi]

Are these mod files in example/mechanisms copied from neurodamus-models repo?
If yes, I am wondering if they can be copied on the fly during testing. That would be make them consistent

[darshanmandge]

Are these mod files in example/mechanisms copied from neurodamus-models repo?
I am wondering if they can be copied on the fly during testing. That would be make them consistent

Good idea. The README In the example/mechanisms folder does suggest that the mechanisms comes from neurodamus-models repository at specific commits. However, if I add the mod files from latest commit for the those folders neurodamus-models twe would probably need to update tests/examples. I can tackle this in a seperate issue if that is fine with you, @WeinaJi.

[WeinaJi]

I can tackle this in a seperate issue if that is fine with you

Yes, that plan sounds good. Probably you just do git clone neurodamus-models repo before executing this line in tox.
I also see that BlueCellLab has 2 places for mod files, examples/mechanisms and tests/mechanisms. It would be good to use one location with sub folders like neocortex, hippocampus, etc.

I don't have other comments for this PR. I will let @jamesgking comment.

[jamesgking]

Changes look straight forward to me. Yes, I agree with idea to have a future PR use directly the models from other repo to avoid duplication and out of sync files

[darshanmandge]

Thank you, @WeinaJi and @jamesgking ! I created the new issue for mechanisms.