1.2.0 regression: custom scale not working

[torfbolt]

Custom scales are not respected by matplotlib 1.2.0, as can be seen comparing the two gallery examples
http://matplotlib.org/1.1.1/examples/api/custom_scale_example.html
and
http://matplotlib.org/examples/api/custom_scale_example.html

A quick print of matplotlib.scale.get_scale_names() shows that it is in the list. However, it does not seem to be applied when calling gca().set_yscale('mercator'). Calling gca().set_yscale('log') works however.

[torfbolt]

Ok, it seems I found the problem:
The actual transformation function in the example is called transform(self, a), but with matplotlib 1.2 it has to be transform_non_affine(self, a).
Maybe someone can fix this in the example.

[pelson]

Thanks @torfbolt .

We should try to fix this before we tag v1.2.1.

@torfbolt - are you interested in submitting a PR?

[torfbolt]

Sorry, don't have a branch of matplotlib at the moment, I just fixed it in my private class. Btw, are there plans to include further scales into matplotlib? I could supply an implementation of the sigmoid scale:
http://en.wikipedia.org/wiki/Sigmoid_function

Are there also test cases for the examples?

[pelson]

Are there also test cases for the examples?

Sadly not. Its something I'd like to see, but there are some hurdles we would have to overcome first. Under https://github.com/matplotlib/matplotlib/wiki/MEP12 we are also planning to tidy up most of the examples, but that really is a massive job (note to self: might be a good task for a sprint at SciPy'2013...)

[dmcdougall]

...don't have a branch of matplotlib at the moment, I just fixed it in my private class...

If you do a diff -u old_class.py new_class.py and paste the output here, we'll be able to see what you changed to fix it. Also, I think Github has an inline web editor that turns your edits into a pull request.

[mdboom]

Closing, since #1851 is closed.