Gzip file load error on latest master

[solitaryangler]

Hi,
I needed the TCK streamline format so I installed the latest master branch with:
pip install -e git+https://github.com/nipy/nibabel.git#egg=nibabel
I can now load TCK streamlines:

nib.streamlines.FORMATS
Out[2]: 
{'.tck': nibabel.streamlines.tck.TckFile,
 '.trk': nibabel.streamlines.trk.TrkFile}

But I cannot load gzipped nifti files any more. If I gunzip the nifti, then it works fine.
The error I get:

AttributeError                            Traceback (most recent call last)
<ipython-input-4-6e9570ba8f5d> in <module>()
----> 1 nib.load('Diffusion_brain_mask.nii.gz')

.../Softwares/nibabel/src/nibabel2/nibabel/loadsave.py in load(filename, **kwargs)
     41     sniff = None
     42     for image_klass in all_image_classes:
---> 43         is_valid, sniff = image_klass.path_maybe_image(filename, sniff)
     44         if is_valid:
     45             img = image_klass.from_filename(filename, **kwargs)

.../Softwares/nibabel/src/nibabel2/nibabel/filebasedimages.py in path_maybe_image(klass, filename, sniff, sniff_max)
    508         sniff = klass._sniff_meta_for(filename,
    509                                       max(klass._meta_sniff_len, sniff_max),
--> 510                                       sniff)
    511         if sniff is None or len(sniff[0]) < klass._meta_sniff_len:
    512             return False, sniff

.../Softwares/nibabel/src/nibabel2/nibabel/filebasedimages.py in _sniff_meta_for(klass, filename, sniff_nbytes, sniff)
    455         # Attempt to sniff from metadata location
    456         try:
--> 457             with ImageOpener(meta_fname, 'rb') as fobj:
    458                 binaryblock = fobj.read(sniff_nbytes)
    459         except IOError:

.../Softwares/nibabel/src/nibabel2/nibabel/openers.py in __init__(self, fileish, *args, **kwargs)
    135         if 'compresslevel' in arg_names and 'compresslevel' not in kwargs:
    136             kwargs['compresslevel'] = self.default_compresslevel
--> 137         self.fobj = opener(fileish, *args, **kwargs)
    138         self._name = fileish
    139         self.me_opened = True
.../Softwares/nibabel/src/nibabel2/nibabel/openers.py in _gzip_open(filename, mode, compresslevel)
     72     # use indexed_gzip if possible for faster read access
     73     if mode == 'rb' and HAVE_INDEXED_GZIP:
---> 74         gzip_file = SafeIndexedGzipFile(filename)
     75 
     76     # Fall-back to built-in GzipFile (wrapped with the BufferedGzipFile class

indexed_gzip.pyx in indexed_gzip.IndexedGzipFile.__cinit__ (indexed_gzip.c:1404)()

AttributeError: 'str' object has no attribute 'mode'

I'm using python 2.7.12 with conda on a mac.
Many thanks for this excellent work.
Best,

-Auro

[effigies]

Hi, thanks for the report. We'll look into this ASAP.

[solitaryangler]

Thanks for the quick response. I'll keep my eyes open for an update/solution.

[pauldmccarthy]

Hi @solitaryangler, could you update your version of indexed_gzip to the latest available? Versions of indexed_gzip prior to 0.4.0 (released about a month ago) expect the filename to be passed as a keyword argument, whereas the code in nibabel has been written against indexed_gzip>=0.4.0

[solitaryangler]

Hi @pauldmccarthy many thanks. This is to confirm that upgrading indexed_gzip:
pip install indexed_gzip -U
resolved my problem.

[effigies]

@pauldmccarthy We shouldn't try to use indexed_gzip older than 0.4.0, then. Maybe print a warning and set HAVE_INDEXED_GZIP=False if the version is too old?

[pauldmccarthy]

@effigies sure, I can add this to PR #562