Topup output not found

[arnaudbore]

Dear nipype experts,

When I run topup it looks for a ".nii.gz" file as output 'out_fieldcoef'.
Because all my FSL settings are in ".nii" file it doesn't find anything and then crashes.

Thank you for your help.
Arnaud

[satra]

this is being fixed here:

#794

[arnaudbore]

Thank you Satra. Can you tell me when is going to be in the new version of nipype ? Maybe 0.9.3 ?

[satra]

definitely in 0.9.3 - it should be in master soon. we'll see if we can release 0.9.3 by end of february. there are a few other things that need to happen as well.

[oesteban]

The output 'out_fieldcoef' is assigned here: https://github.com/nipy/nipype/blob/master/nipype/interfaces/fsl/epi.py#L222 using the fsl.base method to generate filenames, as the argument change_ext is not overridden, it should observe the environment settings.

Is this happening to you with other interfaces of FSL?

[oesteban]

@arnaudbore Did you find out the problem?

[satra]

@oesteban - i think this was fixed in #794

@arnaudbore - did you check?

[rfdougherty]

I don't think it's been fixed. I just pulled master and still get a FileNotFoundError because it's looking for the fieldcoef file in the current directory rather than the directory specified in out_base (which is where the fieldcoef file actually is).

[oesteban]

That's a bright hypothesis, I did not think of out_base. I'm currently looking at this. Thanks @rfdougherty .

EDIT: ok, yes, a fast doctest writing out_base shows that the outputs are not updated.

[oesteban]

@satra this is the diagnostic:

we have two input traits that should name_source from the same trait (out_base) that should be sourced itself from in_file. This dependency is broken by the sorted return of traits within _parse_inputs.

I've thought of two solutions:

1. Review this sorted iteration of traits, so that dependencies are identified and the sources set in front of the list.

2. Modify _filename_from_souce to be recursive. This is probably a faster fix, but also produces risky redundancies in parsing.

[oesteban]

Submitted a fix, using option 2). If inputs are not changed during parsing, it should work out.

[satra]

@rfdougherty and @oesteban - could you take a look at this notebook before we make any changes?

http://nbviewer.ipython.org/urls/dl.dropbox.com/s/l84fgl333fperqb/Eddy.ipynb?dl=0

this workflow ran fine for me. i'll run it again to make sure things haven't changed.

[rfdougherty]

Hi Satra,

Everything does work fine if cwd == out_base. Reading through your workflow, it seems that's how you run it. E.g., " merged_file = os . path . join ( os . getcwd (), 'merged.nii.gz' )". But it will fail if out_base != cwd.

cheers,
bob

Robert F. Dougherty, PhD
Research Director
Stanford Center for Cognitive and Neurobiological Imaging
70 Jordan Hall * Stanford CA 94305 * 650-725-0051
http://www.stanford.edu/~bobd

----- Original Message -----

From: "Satrajit Ghosh" [email protected]
To: "nipy/nipype" [email protected]
Cc: "Bob Dougherty" [email protected]
Sent: Sunday, September 21, 2014 6:20:24 AM
Subject: Re: [nipype] Topup output not found (#796)

@rfdougherty and @oesteban - could you take a look at this notebook before we make any changes?

http://nbviewer.ipython.org/urls/dl.dropbox.com/s/l84fgl333fperqb/Eddy.ipynb?dl=0

this workflow ran fine for me. i'll run it again to make sure things haven't changed.

—
Reply to this email directly or view it on GitHub .

[satra]

thanks @rfdougherty:

narrowed it to this - the only thing that's different here is that out_base is specified as a location somewhere:

In [1]: from nipype.interfaces.fsl import TOPUP
In [2]: topup = TOPUP()
In [3]: topup.inputs.in_file = "b0_b0rev.nii"
In [4]: topup.inputs.encoding_file = "topup_encoding.txt"
In [5]: topup.cmdline
Out[5]: 'topup --config=b02b0.cnf --datain=topup_encoding.txt --imain=b0_b0rev.nii --out=b0_b0rev_base --iout=b0_b0rev_corrected.nii.gz --fout=b0_b0rev_field.nii.gz --logout=b0_b0rev_topup.log'

In [6]: topup.inputs.out_base = '/tmp/foo'
In [7]: topup.cmdline
Out[7]: 'topup --config=b02b0.cnf --datain=topup_encoding.txt --imain=b0_b0rev.nii --out=/tmp/foo --iout=b0_b0rev_corrected.nii.gz --fout=b0_b0rev_field.nii.gz --logout=b0_b0rev_topup.log'
In [8]: topup._list_outputs()
Out[8]: 
{'out_corrected': '/software/nipy-repo/nipype/nipype/testing/data/b0_b0rev_corrected.nii.gz',
 'out_enc_file': <undefined>,
 'out_field': '/software/nipy-repo/nipype/nipype/testing/data/b0_b0rev_field.nii.gz',
 'out_fieldcoef': '/software/nipy-repo/nipype/nipype/testing/data/foo_fieldcoef.nii.gz',
 'out_logfile': '/software/nipy-repo/nipype/nipype/testing/data/b0_b0rev_topup.log',
 'out_movpar': '/software/nipy-repo/nipype/nipype/testing/data/foo_movpar.txt'}

we need to do two things:

1. check to ensure that fsl respects such a location specification. (i'll test this)
2. depending on the answer modify topup. i don't think a base class modification is required here.

[satra]

results from 1. topup doesn't appear to like it when out contains a location that doesn't exist. so we have to be careful how this interface gets redesigned. i'm now running it with the same command as below, but having created /tmp/foo

$ topup --config=b02b0.cnf --datain=/software/data/isn/eddy/preproc/b0_acq.txt --imain=/software/data/isn/eddy/preproc/b0_merged.nii.gz --numprec=float --out=/tmp/foo/b0_merged_base --iout=b0correct.nii.gz --fout=b0_merged_field.nii.gz --logout=b0_merged_topup.log
Error: failed to open file /tmp/foo/b0_merged_base_fieldcoef.nii.gz
Image Exception : #22 :: ERROR: Could not open image /tmp/foo/b0_merged_base_fieldcoef
libc++abi.dylib: terminating with uncaught exception of type RBD_COMMON::BaseException
Abort trap: 6

[satra]

so this works:

$ topup --config=b02b0.cnf --datain=/software/data/isn/eddy/preproc/b0_acq.txt --imain=/software/data/isn/eddy/preproc/b0_merged.nii.gz --numprec=float --out=/tmp/foo/b0_merged_base --iout=b0correct.nii.gz --fout=b0_merged_field.nii.gz --logout=b0_merged_topup.log

but the root directory (/tmp/foo in this case) has to exist.

i'll move over to the PR for comments now, but i think this can be tackled with a simpler change.

[oesteban]

I think this is fixed