Str Categorical Axis Support

[story645]

Cleaned up and updated version of #6612, this adds support for directly plotting lists of strings, so that the following is now possible:

data = {'apples':10, 'oranges':15, 'lemons':5, 'limes':20}
fig = plt.figure()
ax = fig.add_subplot(1,1,1)
ax.bar(list(data.keys()), list(data.values()), align='center', color='lightgray')
ax.xaxis.set_tick_params(size=0)

cat = ["bored", "happy", "bored", "bored", "happy", "bored"]
dog = ["happy", "happy", "happy", "happy", "bored", "happy"]
time = ["combing", "drinking","feeding", "napping", "playing", "washing"]

fig = plt.figure()
ax = fig.add_subplot(1,1,1)
ax.plot(time, dog, label="dog")
ax.plot(time, cat, label="cat")
ax.legend()

Now fighting with the update ticks (@tacaswell) use case:

fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.plot(['a', 'b'])
ax.plot(['a', 'b', 'd'])
ax.plot(['b', 'c', 'd'])

'cause it breaks in randomly different ways without the other tests failing.

And beats my why, but Python2.7 triggers weirdness with the bounds & view_limits that Python3.5 doesn't:

_axes.py", line 4023, in scatter: self.autoscale_view()
/axes/_base.py", line 2283, in autoscale_view: margins['left'], margins['right'], self.set_xbound)
axes/_base.py", line 2278, in handle_single_axis: set_bound(x0, x1)
axes/_base.py", line 2763, in set_xbound: self.set_xlim(lower, upper, auto=None)
axes/_base.py", line 2830, in set_xlim: left = self.convert_xunits(left)
artist.py", line 197, in convert_xunits: return ax.xaxis.convert_units(x)
axis.py", line 1500, in convert_units: ret = self.converter.convert(x, self.units, self)
category.py", line 24, in convert: vals = vmap[value]
KeyError: -1.1049590217242939

this was 'caused by plt.plot('a')

[tacaswell]

typo in comment.

[tacaswell]

This file looks like it get renamed and then rebase fought back?

[story645]

Ugh, probably. Rebase and I are not friends.

[tacaswell]

merge detris

[tacaswell]

Don't put mutables in the signatures as defaults, it can lead to some fun closures.

def foo(d=[]):
    d.append(len(d))
    print(d)

gives

In [24]: foo()
[0]

In [25]: foo()
[0, 1]

In [26]: foo()
[0, 1, 2]

In [27]: foo()
[0, 1, 2, 3]

Either make this an empty tuple or make the default None and put logic in the function like

if old_map is None:
    old_map = []

[tacaswell]

Can you revert the merge of master into this branch?

[tacaswell]

Ah, I see there are a bunch of them, I will see if I can de-tangle this.

[tacaswell]

If you look at the category branch on my gh I have cherry-picked off all of your commits and rebased them onto current master.

The changes to the code are exactly the same, but the history is cleaner

22:15 $ git diff story645/category tacaswell/category

shows that there are no code changes.

I suggest you do the following

• make a new branch on your current branch (just to be safe)
• reset your local category branch to be my cleaned one git reset --hard tacaswell/category
• push that branch to your gh git push --force

If you have not discovered magit (https://github.com/magit/magit) yet, I can not suggest it highly enough.

[tacaswell]

appveyor failure is spurious (failed upload, tests passed).

[tacaswell]

appveyor failure is spurious (failed upload, tests passed).

The plan is to collect a single what_new entry + examples towards the end.

[tacaswell]

🎆 🎉 🎆