FIX: Permit relative path for concatenated_file input to Concatenate()

[effigies]

The example as given in the docs will fail to run, raising:

TraitError: The 'concatenated_file' trait of a ConcatenateOutputSpec instance must be an existing file name, but a value of 'bar.nii' <type 'str'> was specified.

While the ConcatenateInputSpec seems happy with a relative path, it seems ConcatenateOutputSpec needs an absolute path, so simply passing it through is insufficient. This change prepends the working directory. I've tested this on my own pipeline, and it resolves an error like the one above.

[coveralls]

Coverage decreased (-0.002%) to 72.981% when pulling 5c72074 on effigies:mri_concat_fix into c46b932 on nipy:master.

[coveralls]

Coverage decreased (-0.03%) to 72.951% when pulling 7052512 on effigies:mri_concat_fix into c46b932 on nipy:master.

[effigies]

Seeing FreeSurfer isn't installed in Circle, so tests can't run. Is this something that should change?

[satra]

freesurfer is a nasty download 4.6GB. perhaps we can create a separate minimal freesurfer just for recon-all. more recently i think i've seen an internal docker version. let me look into that.

in the meantime, let's not run this on circle. it would also take too long to run.

@dgellis90 - since you are running this on your cluster, perhaps you can keep us posted there.

[oesteban]

I'm currently migrating CircleCI tests to docker containers in some other projects. I think this could be the solution to the problem of downloading so large files.

[coveralls]

Coverage decreased (-0.6%) to 72.373% when pulling c9e8bdb on effigies:mri_concat_fix into c46b932 on nipy:master.

[chrisgorgo]

Freesurfer inside a docker image vs outside of it is still a 4GB download.
We can cache it (or the docker image with it) though.

On Wed, Mar 30, 2016 at 2:48 PM, Coveralls notifications@github.com wrote:

[image: Coverage Status] https://coveralls.io/builds/5605869

Coverage decreased (-3.3%) to 69.672% when pulling c9e8bdb
c9e8bdb
on effigies:mri_concat_fix into c46b932
c46b932
on nipy:master.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub
#1411 (comment)

[satra]

@chrisfilo is correct - that's why we need to restrict it to what recon-all needs - which is only about 350MB - but there is indeed an effort to again make freesurfer available through neurodebian - which may become the smallest distribution for the binaries.

[chrisgorgo]

I've tried to slim down FreeSurfer in this NeuroVault docker image
https://github.com/NeuroVault/NeuroVault/blob/master/fs_docker/Dockerfile,
but I never got even close to 350Mb...

On Wed, Mar 30, 2016 at 3:27 PM, Satrajit Ghosh notifications@github.com
wrote:

@chrisfilo https://github.com/chrisfilo is correct - that's why we need
to restrict it to what recon-all needs - which is only about 350MB - but
there is indeed an effort to again make freesurfer available through
neurodebian - which may become the smallest distribution for the binaries.

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#1411 (comment)

[effigies]

Okay. So, as far as this PR goes, I can report that nosetests nipype.interfaces.freesurfer.tests runs without error on my machine with FreeSurfer 5.3.0 installed.

[coveralls]

Coverage decreased (-0.5%) to 72.549% when pulling cd985f1 on effigies:mri_concat_fix into ee871ba on nipy:master.