[Neuroimaging] Should we prefer nii.gz or nii?

[Michael Waskom]

Hi folks,

I've typically used nii.gz format for all of my data. This is also the
prescription in the new BIDS <http://bids.neuroimaging.io/bids_spec0.4.pdf>
spec. Lately, though, I've been wondering whether this makes sense in all
cases. It seems like ZFS is becoming more common for storage in cluster
environments. As I understand it, this means that much of the compression
that you would get from using .nii.gz will also be applied to normal .nii
files. There are other benefits to .nii as well, for instance,
memory-mapping the files when you load them in nibabel. I'm not sure how
decompression affects load/save times when reading from ZFS using
compressed files, but there might be additional costs there.

Do others have thoughts about the relative merits of compressed vs.
uncompressed files?

Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/neuroimaging/attachments/20150829/c77e4464/attachment.html>