Errors/failures when running `nosetests --with-doctest` about scikit-image HOT 6 CLOSED

Opened duplicate issue #237, some discussion there.

from scikit-image.

Since you closed the other PR, I'll just copy your comment here so that it doesn't look like I'm talking to myself :)

In sklearn we put +SKIPTEST at the end of all lines of the non-doctests. That is clearly not optimal but we couldn't think of anything better. Using >> is an option but breaks the sphinx translation (so it will be ugly in the rendered html).

Good point about the Sphinx issue. I can't seem to find documentation for +SKIPTEST; do you mean +SKIP. It's pretty ugly, but if we limit the number of examples that aren't valid doctests (which is a good idea anyway), it shouldn't be too bad.

I also noticed that there where some issues with dtype. It was "expected" to be a string, but really was a numpy object. Do you think these where simply never tested?

I wouldn't be too surprised if they weren't ever tested, actually, but I can't really say for sure.

And do you have any idea waht the TypeError: 'module' object is not callable might be?

It looks like rank_order is both a module in the filter subpackage and a function imported into the namespace of the filter subpackage. I don't know why it would work in some cases but not in others. We should rename the rank_order module to _rank_order (this is the convention we've been using).

from scikit-image.

Oh actually, the dtype issue may have been a change in numpy. I vaguely remember some changes to dtype printing that were discussed on the numpy list. I see the same error as you though, so it's probably safe to change. (Unless someone else chimes in to the contrary).

from scikit-image.

Renaming some of the modules in the filter package seems to fix a lot of the errors (45 errors down to 30; 26 failures down to 20). See PR #240.

from scikit-image.

Unfortunately, some modules with similarly-named functions also have cython modules that already use the underscore-prefix convention (e.g. skeletonize and watershed).

from scikit-image.

Thanks for the feedback and for tackling this so quickly (at least partially).
I try to focus on #206 to get that finally done, then I'll try to give this one a look.

from scikit-image.