ticks based on number of subplots

[ivanov]

here's an implementation of adjusting the tick locator depending on how many subplots there are. I implemented this at PyData a few weeks back, but forgot to submit the PR.

to see the goodies of this PR, use this in the ipython notebook

%pylab inline
#rcParams['figure.figsize']=2,2
plt.subplots()
plt.plot(24*np.random.rand(10), np.random.rand(10))
for x in range(7):
    fig, axes = plt.subplots(1,x+2, sharey=True)
    for ax in axes.flat: ax.plot(24*np.random.rand(10), np.random.rand(10))

plt.subplots()
plt.plot(np.random.rand(10), 24*np.random.rand(10))

for x in range(7):
    fig, axes = plt.subplots(x+2,1, sharex=True)
    for ax in axes.flat: ax.plot(np.random.rand(10), 24*np.random.rand(10))
#fig.tight_layout()

(I'll rebase this if this won't merge cleanly)

[jdh2358]

Looks like you left a print in here

[jdh2358]

Hey Paul,

This is a very useful PR, because the too many ticks problem is a common annoyance. I think "subplots" is the perfect place to add enhancements like these because a) it knows the whole geometry and b) it is still somewhat lightly used and so less likely to change/break people's existing code.

I would like to do more: in particular, I think we should expose a tight_layout=True kwarg to subplots so that tight layout is on by default (but configurable), Can you add this to the PR?

[ivanov]

Hi John,

I removed the print statements, and adjusted the logic for sharex and sharey, since locators get reused in that case.

as for tight_layout, I think with the current implementation, it only make sense to call tight_layout after whatever plotting will happen, since it's a one-time action which only takes into account the current state of the relevant artists. I hope @leejjoon or anyone else corrects me if I'm wrong.

[leejjoon]

Yes, tight_layout is a one-time action. Although we may implement some automatic use, see #774.

[jdh2358]

Hey Paul,

just wanted to point out that we are in the thick handling the v1.1.x release right now so are in a holding pattern on master PRs, but I am definitely interested in this so make sure to ping us if we fall quite on this PR after we get 1.1.1 out.

JDH

[pelson]

@ivanov : I think we can start looking at this again now that (for the most part) v1.2.0 is out of the way. Would you be willing to rebase this, and add appropriate tests and we can get this PR back on the agenda.

Cheers,

[pelson]

@ivanov is that a no, or a "I'm on it"? 😄

[ivanov]

both :)

I took a stab and there were too many conflicts immediately, and I didn't
have the mental bandwidth to juggle them at the time, let me take another
stab...