Introduce orientation to bar plot

[sbrugman]

The dendrogram has an option to change the orientation from top-down to left-right. This is convenient when having a large amount of columns. The bar plot misses this option. To make the bar plot effective for large numbers of columns, it should have a similar behaviour.

This pull requests adds the basic functionality. The figsize is not yet adjusted, this should be added.

[ResidentMario]

Thanks for contributing this PR! This looks like a good start.

There are still a couple of things missing from the re-oriented msno.bar view. To demonstrate, here's the standard orientation for msno.dendrogram:

import numpy as np
import pandas as pd

np.random.seed(42)
simple_df = pd.DataFrame((np.random.random((20, 10)) > 0.5), columns=range(0, 10)).replace(False, np.nan)
large_df = pd.DataFrame((np.random.random((250, 60)) > 0.5)).replace(False, np.nan)

import missingno as msno
import matplotlib.pyplot as plt

msno.dendrogram(simple_df)

And the reoriented view (msno.dendrogram(large_df)):

Here is the standard view for msno.bar (msno.bar(simple_df)):

And the reoriented view added in this PR (msno.bar(large_df)):

The reoriented msno.dendrogram preserves certain features of the plot that the reoriented msno.bar does not preserve. Namely:

• It chooses a "smart" figure size based on the number of columns in the input data that ensures there is adequate room for the labels; msno.bar does not do that in this PR (as you mentioned in your opening comment).
• It preserves the axis labels—specifically, the standard values formerly on the y-axis that are now on the x-axis. The reoriented msno.bar only preserves the column labels, all of the other labels are dropped.

If you fix these two issues, I'd be happy to merge this PR. This looks like a great feature!

[sbrugman]

Should be fixed.

[ResidentMario]

The sizes of the tick labels are inconsistent under rotation.

Default:

Rotated:

Compare this with what dendrogram does:

dendrogram doesn't use fontsize directly, and instead manipulates it using labelsize=int(fontsize / 16 * 20). This may be appropriate to do in this case as well.

[sbrugman]

Font size problem still requires some tinkering, but we are getting there.

[ResidentMario]

@sbrugman Hey, any update on this?

[sbrugman]

My focus is somewhere else right now, not sure if I have time soon to fix the font size.

[ResidentMario]

Ok, the amount of work remaining seems pretty small, maybe I'll take up finishing it sometime soon.

…

On Wed, Sep 11, 2019, 3:29 PM Simon Brugman ***@***.***> wrote: My focus is somewhere else right now, not sure if I have time soon to fix the font size. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#94?email_source=notifications&email_token=AA2OIZJVBVCMH77A3DE32NDQJFWLPA5CNFSM4IGZQHSKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD6QCZYY#issuecomment-530590947>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA2OIZN7MDRRA3MNOLNZ5YDQJFWLPANCNFSM4IGZQHSA> .

[ResidentMario]

For whatever reason I can't replicate the font size behavior now. Looks fine, so I'll just merge this as is.

I imagine you don't often submit PRs that take two years to merge. 😉