Implements the first necessary step to support patchclamp data...

[SylvainTakerkart]

Temporary solution, as discussed before the summer: we add an if statement to differentiate extra- and intra-cellular ephys. This decides whether there's a 'session' level in the BIDS hierarchy or not...

[SylvainTakerkart]

this needs a bit more of testing... indeed, my [full-scale] test [on real data] for which the code is not included in this PR shows unexpected behavior of the little changes introduced in this PR!

therefore, this PR is not ready for reviewing!

[SylvainTakerkart]

todo...

• [x] rename extra into ece and intra into ice
• [x] make ephys_type a class attribute
• [x] make filename_stem a class attribute
• [x] refactor filname_stem usage consequently everywhere
• [x] add unit test for the use of ice and ece ephys types

[SylvainTakerkart]

Info on the unit tests of BEP032Generator and what I've done so far:

• from what I understood, in the existing class, nothing was actually testing for the existence of the session-level directory (the variable which is called session_folder actually designates the ephys folder where the data files live) nor the session.tsv metadata files;
• in consequence, with the new ice or ece values (which decide whether there is a session-level directory or not), all the existing tests should pass whether we use ephys_type=ece (old behavior, with session-level directory) or ephys_type=ice (new behavior for patchclamp data, without session-level directory); in practice, I tried it, and it's indeed the case;
• for this PR, I therefore decided to propose a symbolic evolution of these unit tests: I duplicated the class Test_BEP032Data into two classes Test_BEP032Data_ice and Test_BEP032Data_ece; this offers the possibility in the future to differentiate the tests between ice and ece data; but I havn't implemented anything for now (the two classes are identical, except for the value of the ephys_type attribute, which is either ice or ece); I think we need to discuss further what we want to do (larger refactoring, etc.) before it's necessary to implement something in this direction...

@JuliaSprenger , if you're ok with this symbolic change in the unit tests, this PR is now ready for review!

[JuliaSprenger]

Hi @SylvainTakerkart It looks like the existence or absence of the session folder level is the only difference between the ece and the ice case so far. Are you sure this is the main conceptual difference between these two data types? In the way the current specifications are written also ece datasets can omit the session level folder and it might as well be possible for ice datasets to require a session level. Wouldn't it be better to have the optional session level generation independent of the ece/ice mode? I.e. make the session name argument optional and omit the session level folder if no value is provided?

[SylvainTakerkart]

Hi @JuliaSprenger, I see what you mean...

The question is whether we want this code to be i) an implementation of the specs with bijective correspondances between the specs and the code, or ii) something more practical, with "recommended implementations" (for which we could e.g recommend to have the session level for ece and not for ice). For the base class BEP032Data, we clearly have to go with i), so I agree with you!

A consequence of this is that almost everything becomes "optional": the session directory, as well as everything in the data filenames exept 'sub-' ('ses-', 'run-', 'task-' are optional)... So we might have to remove a good portion of what's coded in the base class' methods... I'm going to look at this! (all this is not clear in my mind anyhow ;) )... If you have hints, I welcome them ;)

[SylvainTakerkart]

With this latest commit, in fact, ephys_type is not used anywhere... That means in fact that there might not be any "conceptual differences" between ice and ece, but only different usages of the same concepts... WDYT? If you're OK, I'll make ephys_type disappear totally...

[JuliaSprenger]

Yes, that looks like the better approach to me :+1:

[SylvainTakerkart]

I removed the ephys_type class attribute... I left the two test classes for ece and ice, this might be useful for the future, we'll see... Ready for merging?