[WIP]v0 for spm parser

[remiadon]

This PR tries to introduce a parser for SPM, i.e a piece of program that takes .m files as those available to SPM users and produces the corresponding BIDSprov graph, in jsonld

• implement a v0
• CI script to check conversion is OK with a few different datasets
• bash example with visualization
• add more examples

[remiadon]

@cmaumet Here is what I get when I run

python bids_prov/spm_parser.py examples/spm_default/batch.m -o test.json && python bids_prov/visualize.py test.json && open test.png

[cmaumet]

Discussed w/ @remiadon on Feb 8, next steps include:

• Reverse dependencies into "used" instead of "wasGeneratedBy"
• Add wasAssociatedWith relation between all activities and the SPM software
• For each cfg_dep add a wasGeneratedBy relation between the entity and the corresponding activity
• Try and use prov library

[remiadon]

Here is what I get after f094136

[remiadon]

And here is what I get now, after 46436ff

@cmaumet do you see anything abnormal ?

[cmaumet]

Thanks @remiadon! I did not see anything specific but in general I think all activities should have a "used" relationship and this does not seem to be the case in the graph above. Can you have a look?

[remiadon]

@cmaumet what I get after eeb94c8

It just took a bit more time to debug than what I expected :)

Do you see any blocking item now ??

[remiadon]

Oh I see there is a buggy part : "file(1)_4" and "file(1)_3" should not be chained ...

[remiadon]

and here is what I get for the explicit_maskexample https://github.com/incf-nidash/nidmresults-examples/tree/master/spm_explicit_mask

[cmaumet]

Hey @remiadon, thanks for this.

One quick note on your latest graph:

• the links between the entities and the SPM software should be removed (to only retain the links between the activities and the software).

Once this is done, I am happy to merge this. For the next sprint, here are a few things we should look at to improve this example:

• file_move_2 has no wasGeneratedBy link. Do you think there could be a confusion with the output of file_move_1?
• in the name of the Activity could we keep everything after "spm.", for example "preproc._7" would become "spatial.preproc._7"
• To make the graph easier to read, is there an option to hide identifiers?

[remiadon]

@cmaumet Thanks for reviewing

As introduced in 5f14a6c

Here is the generated graph

[remiadon]

And in 434eeaf I fixed activity names :) A limit is left to 30 characters for activity names, otherwise I find it quite verbose

[cmaumet]

Ok let's merge and we'll keep the other items for the next iteration!

[remiadon]

@cmaumet I tried to fix what you mentionned about "file_move_2" by linking dependencies in another manner

Here is the graph I get

I do think it's way better, what do you think ?

[cmaumet]

Yes! There is one remaining missing link : "fmri_spec" used "Smoothed images"

[remiadon]

@cmaumet ready to merge