facet_wrap broken in new release

[jbloom]

The facet_wrap command is broken in the latest release (0.5.0): it only plots the data for the first facet.

See this example:

import pandas
import plotnine
from plotnine import *

print(f"Using plotnine {plotnine.__version__}")

df = pandas.DataFrame({'facet_var':['a', 'a', 'a', 'a', 'b', 'b', 'b'],
                        'hist_var':[1, 2, 3, 3, 3, 4, 5]})

print(df)

ggplot(df, aes('hist_var')) + geom_histogram(binwidth=1, center=0) + facet_wrap('~ facet_var')

Using plotnine 0.5.0
  facet_var  hist_var
0         a         1
1         a         2
2         a         3
3         a         3
4         b         3
5         b         4
6         b         5

[TyberiusPrime]

It appears to not be a faceting bug per se (faceting geom_line works as expected),
but to be triggered by the combination of changes in
401741c

I've tracked it down to an indexing problem in transform_position/squish_infinite.
Line 97 in positions/position.py reads
data[xs] = data[xs].apply(trans_x)
but bounds.squish_infinite will return an index that does not match the one passed in.

(This is because squish_infinite casts the data to np.asarray, and then back to the original type - the question is whether
to special case dataframes in squish_infinite, or fix-the-index in transform_position)

[has2k1]

@TyberiusPrime, thanks for the debugging. The solution is to fix the functions in mizani so that the index of series objects is always preserved, i.e since np.asarray has a side-effect of stripping away the index information, need to put a guard around it.

[has2k1]

Fixed in mizani v0.5.2.

[jbloom]

Thanks, this indeed fixed my problems.