OK...
As you've seen the code is also in pretty bad state and untested, but it's not mine, so I'll revert the latest commit.

Thanks - I agree the SPM code does not some cleaning up and some tests. I'll try and have a look soon to see what is of use in there - the ReML code in particular looked substantial. It may be we should move this stuff around somewhere, or even remove it, but I'd like to give it some thought, so thanks for reverting for now, and I'll put an issue up for it.

Ok to rebase this one and merge?

#166

if code is lying around and not being used anywhere else in the library and is not part of the api, remove it (this is git - you can always get back to it if needed). also given that the most useful aspect here is reml, that should be moved to a more aptly named place - reml is an algorithm and has nothing to do with spm per se except they use it. it should really be part of scipy or statsmodels.

Satra - please consider moving your comment to the issue.

We carry lots of general code, such as pca, model fitting and so on. Once the work has been done to push generic code upstream, we can remove it.

For now, we need someone who has the time to work out which pieces of this code are functional. My own approach to this is to first write tests, to work out what the code is doing. Once I have done that, I work out whether the code is useful and if so, where it should go.

picking up the maintenance of something that might be useful, but is currently not used anywhere, may not be the most useful contribution to the project at this stage. there are plenty of other things that need to get cleaned up. and git will allow revisiting this when there is interest. that said, i did look at the code and it implements spm's reml with similar variable naming convention (which i would recommend altering). it even has a little section under __name__ == '__main__' that could be turned into a test. but it seems too isolated a piece of code to invest the time into at this point.

If you see somewhere that you can make a useful contribution, then please do, all hands help.

OK for merging and discussing later the spm module.

Rebased and merged as commit e6dfbcf

Bertrand - the cleanup caused some doctest errors:

http://nipy.bic.berkeley.edu/builders/nipy-py2.6/builds/125/steps/shell_1/logs/stdio

I think you and I both missed the newline in the doctests.

Fixed in commit 000a331

I've got travis-ci continuous tests activated - it's just a few clicks to activate - I find it very useful because it runs all the tests on every commit.

On 05/19/2012 03:24 PM, Matthew Brett wrote:

Bertrand - the cleanup caused some doctest errors:

http://nipy.bic.berkeley.edu/builders/nipy-py2.6/builds/125/steps/shell_1/logs/stdio

I think you and I both missed the newline in the doctests.

Fixed in commit 000a331

I've got travis-ci continuous tests activated - it's just a few clicks to activate - I find it very useful because it runs all the tests on every commit.

---

Reply to this email directly or view it on GitHub:
#157 (comment)
Oops. Thanks for the fix and for the hint.

Bertrand