BF: fix import statement for get_cmap

[matthew-brett]

I think you can't do import module.attribute as attribute, and you have to
do something like from module import attribute.

[arokem]

Is there some way to create a test that fails without this fix? Any way to inject a value into the have_matplotlib variable?

[matthew-brett]

Actually the buildbots picked this one up here:

http://nipy.bic.berkeley.edu/builders/dipy-py2.6-osx-10.5-ppc/builds/312/steps/shell_8/logs/stdio

I guess we have to continue making sure some buildbots don't have matplotlib...

[arokem]

Oh - nice. Do you think there is any general way to simulate not having mpl
for these modules, for testing purposes while developing, and on Travis?

On Wed, Jul 30, 2014 at 12:24 AM, Matthew Brett notifications@github.com
wrote:

Actually the buildbots picked this one up here:

http://nipy.bic.berkeley.edu/builders/dipy-py2.6-osx-10.5-ppc/builds/312/steps/shell_8/logs/stdio

I guess we have to continue making sure some buildbots don't have
matplotlib...

—
Reply to this email directly or view it on GitHub
#397 (comment).

[matthew-brett]

I could imagine some sort of run-time patch to the optional package routines so that matplotlib appeared to be not installed, but that would have to be applied once before all relevant imports, so it would be fiddly to get right. We can certainly run an extra series of tests on travis, where we break matplotlib import somehow before testing.

[arokem]

OK - do you want to do this here, or would you prefer to put in an issue so
we keep track of this, and simply merge this one for now?

On Wed, Jul 30, 2014 at 10:24 AM, Matthew Brett notifications@github.com
wrote:

I could imagine some sort of run-time patch to the optional package
routines so that matplotlib appeared to be not installed, but that would
have to be applied once before all relevant imports, so it would be fiddly
to get right. We can certainly run an extra series of tests on travis,
where we break matplotlib import somehow before testing.

—
Reply to this email directly or view it on GitHub
#397 (comment).

[matthew-brett]

Added issue : #400 - OK to merge?

[arokem]

Thanks