fix for SpecifySPMModel

[MartinPerez]

@satra you were right, I tested simply correcting the line you pointed out solved the problem. Fix for #1198.

[satra]

i think this can simply be: if len(val):

also could you please add your test to test_modelgen? it should fail without this change, and pass with it? there is already a section that checks concatenation.

[MartinPerez]

@satra I implemented the test and corrected based on your comments. In the test I added some comments to separate the intention of the different tests and added mine at the end. Also modified the last test (before mine) to contrast the fact that when there are three sessions and they are not concatenated the durations are detected just fine.

[satra]

thanks @MartinPerez

[MartinPerez]

thanks @satra

[chrisgorgo]

CircleCI tests are not passing on this PR. We should consider reverting the merge until this is resolved.

[satra]

@chrisfilo - is there an error message? this should only affect the spm test.

[chrisgorgo]

Both example SPM workflows fail to run. See https://circleci.com/gh/nipy/nipype/48 for details.

[MartinPerez]

Effectively this should be reverted. The key problem is the same that I had (non matching number of onsets and events / Number of event onsets (16) does not match the number of durations (4)) but i guess now its about specifying all durations with only one number?

When I ran the simple test, the example with one duration to describe all onsets seemed to be fine too.

[satra]

i think we should fix it to address both cases.

if len(val) == 1:
    # then replicate to number of onsets
else:
    if len(val) != len(number_onsets):
        raise ValueError('Mismatch in number of onsets and durations for session %d, condition %s')
    # append all durations

[MartinPerez]

I would say that we have to check the number of onsets, like:

if len(ons) > 1 and len(val) == 1:
# fill ons with val
elif len(ons) ==  len(val):
# add the durations
else:
    raise ValueError('Mismatch in number of onsets and durations for session %d, condition %s')

you agree?