Fix: handle edge case in synapse location calculation with syn_description

[ilkilic]

When retrieving the synapse location through syn_description, we did not properly handle cases where the location is 0. In NEURON, you cannot insert a synapse or any point process at the start (position 0) or end (position 1) of a segment if it needs an ion e.g. na, k, ca, cl etc. The section location must be strictly within the bounds (0, 1), excluding both endpoints.

[ilkilic]

also address #183

[AurelienJaquier]

Could you explain what you did with the synapses? I am not familiar with synapse stuff

[ilkilic]

sure, I added a description above, let me know if it is not clear

When retrieving the synapse location through syn_description, we did not properly handle cases where the location is 0. In NEURON, you cannot insert a synapse or any point process at the start (position 0) or end (position 1) of a segment. The section location must be strictly within the bounds (0, 1), excluding both endpoints.

[AurelienJaquier]

thanks for the explanations!

[AurelienJaquier]

looks good!

[darshanmandge]

Could you edit the description of the PR to say: "User cannot insert a point process located at positions 0 or 1 of a section if it needs an ion e.g. na, k, ca, cl etc. " ?

I checked with IClamp in NEURON. It works when inserted at the 0 end of a single compartment model as it uses a NONSPECIFIC_CURRENT i.

[ilkilic]

Thanks @darshanmandge for the precision, I have updated the text

[mgeplf]

@ilkilic > [...] insert a synapse or any point process at the start (position 0)

Do you know how neurodamus is handling this? I'm surprised that position 0 isn't allowed.

[darshanmandge]

@ilkilic is away. I will comment to keep a record here. Based on our discussion internally, neurodamus handles this similar to the code in this PR in location_to_point at this line. The syanpses at 0 are moved to 0.0000001 and those at 1 are moved to 0.9999999 .

[mgeplf]

Thanks! We need to make sure these stay in sync.

[darshanmandge]

Yes. true. I will also discuss with @anilbey to find if there is any other code in bluecellulab that we should check to keep in sync with neurodamus for maintaining reproducibility in results.

[anilbey]

We have been using these regression tests for keeping them sync. For the new scientific features adding new regression on local SONATA networks will help.