[MRG] Fix to surface reading with new quad.

[jaeilepp]

Follow up to #460.
I didn't notice that the data format is actually different when reading new quad files. I could not find a file in freesurfer that would use this format so I couldn't add a test for it.

[coveralls]

Coverage decreased (-0.004%) to 96.107% when pulling d24cb38 on jaeilepp:fix-surf into f4a3e0a on nipy:master.

[effigies]

This should be an old-style quad file (mentioned here). Not sure if anybody's seen a new-style quad. @Eric89GXL?

[larsoner]

No idea, it took me quite a while to find the old one :) The best bet is probably to ping the FreeSurfer listserv.

[jaeilepp]

This should be an old-style quad file (mentioned here). Not sure if anybody's seen a new-style quad.

I think that is a new quad file. But it indeed seems like quad files are not tested at all. Someone needs to update the testing dataset to make it possible to test them.

[agramfort]

just make available a file and @matthew-brett will add it to the freesurfer data package

[jaeilepp]

This will be fine https://github.com/mne-tools/mne-testing-data/tree/master/subjects/bert/surf

[matthew-brett]

@jaeilepp - what's the license for that file?

[jaeilepp]

It's part of the FS tutorial dataset, I suppose it's using the freesurfer license.

[matthew-brett]

Any way of being more certain about the license? Where does the file live normally?​

[jaeilepp]

It lives in the 'lightweight' tutorial dataset http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/Data

[matthew-brett]

From tutorial_data/buckner_public_distribution.README.txt - appears to be derived from https://web.archive.org/web/20070711224104/http://www.fmridc.org/f/fmridc/78.html

But - the copy I see in tutorial_data/tutorial_subjs/bert/surf/lh.inflated.nofix from ftp://surfer.nmr.mgh.harvard.edu/pub/data/tutorial_data.tar.gz has a different hash from the one at https://github.com/mne-tools/mne-testing-data/tree/master/subjects/bert/surf - any ideas why? Are they meant to be the same file?

[matthew-brett]

If the source is the FMRIDC, the sort-of license is:

For data housed at the Center, anyone has the right to publish findings 
based on these datasets. Papers whose results are based on datasets 
obtained from the Center should credit the authors of the original study
and acknowledge the Center and accession number of the dataset.

https://web.archive.org/web/20070630230750/http://www.fmridc.org/f/fmridc/help/faq.html#DataSharing

[jaeilepp]

Oh, I just assumed it's the same file with the same name and all. Any idea @Eric89GXL ?

[larsoner]

From the note it comes from FreeSurfer 3.0. We should probably ping the FreeSurfer folks to see if it's okay to relicense it to whatever form is necessary for nibabel.

[matthew-brett]

Is the file version in the tutorial dataset OK for the test? In that case we could use that, because the license is reasonably clear.

[matthew-brett]

The file at https://github.com/mne-tools/mne-testing-data/tree/master/subjects/bert/surf does have the same hash as freesurfer/subjects/bert/surf/lh.inflated.nofix from ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/3.0.3/freesurfer-Linux-rh7.3-stable-pub-v3.0.3-full.tar.gz

[jaeilepp]

Is the file version in the tutorial dataset OK for the test?

Actually no. Seems that it is the same triangle type as the other files.
The file from ftp://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/3.0.3/freesurfer-Linux-rh7.3-stable-pub-v3.0.3-full.tar.gz would work.

[matthew-brett]

Sadly the freesurfer 3 archive is silent on the license for the file, as far as I can see.

[matthew-brett]

How about the file subjects/bert/surf/lh.inflated.nofix in the current 5.3.0 freesurfer archive? I'm asking, because I'm preparing a question to the freesurfer mailing list, and I guess it's easier for them if the question refers to their latest release.

[larsoner]

At least the MD5's don't match

33dd533812f168cddc464c7d64d69166  /home/larsoner/custombuilds/freesurfer/current/subjects/bert/surf/lh.inflated.nofix
82389cce73cc31ea59ab81bc24ff3160  /home/larsoner/custombuilds/mne-testing-data/subjects/bert/surf/lh.inflated.nofix

And the file sizes are 4863500 and 3543685, respectively.

[matthew-brett]

Right - I knew they were different files, but I was wondering if the newest version from freesurfer 5.3.0 can still be used for the tests.

[larsoner]

I was pretty sure I had to go back that far to find a quad file for testing MNE-Python. I just checked and can confirm that nibabel master cannot read the old file but can read the new file.

[matthew-brett]

Great - thanks - I emailed the freesurfer list.

[matthew-brett]

OK - all agreed : http://www.mail-archive.com/[email protected]/msg47878.html

[matthew-brett]

Latest nibabel master now has the updated submodule with the new data file at nibabel-data/nitest-freesurfer/subjects/bert/surf/lh.inflated.nofix.

[codecov-io]

Current coverage is 94.21%

Merging #461 into master will increase coverage by 0.10%

@@             master       #461   diff @@
==========================================
  Files           160        160          
  Lines         21164      21173     +9   
  Methods           0          0          
  Messages          0          0          
  Branches       2265       2265          
==========================================
+ Hits          19919      19949    +30   
+ Misses          823        803    -20   
+ Partials        422        421     -1   

Powered by Codecov. Last updated by d6945c4...24a05ec

[coveralls]

Coverage increased (+0.1%) to 96.207% when pulling 24a05ec on jaeilepp:fix-surf into d6945c4 on nipy:master.

[jaeilepp]

I added the test, but now that I think about it, maybe I should have added a separate function for testing the quad file to avoid the needs_nibabel_data decorator. Wdyt?

[coveralls]

Coverage increased (+0.1%) to 96.209% when pulling 9b66562 on jaeilepp:fix-surf into d6945c4 on nipy:master.

[jaeilepp]

Okay, I made it a separate function. Ready for review.

[larsoner]

LGTM

[agramfort]

LGTM 2

@matthew-brett merge if you're happy

[matthew-brett]

Great - thanks for slogging through.