Retain latex label when slicing DataFrame to Series

[AdamOrmondroyd]

When a dataframe of nested samples is sliced to a smaller dataframe, it retains the latex labels, e.g:

>>> ns[['x0']]
                          x0
labels                 $x_0$
     weights
0    0.000000e+00   0.839566
1    1.088407e-308  0.711607
2    4.830017e-301 -0.703718
3    6.951167e-262  0.980678
4    7.661544e-241  0.727711
...                      ...
>>> ns[['x0']].get_labels()
array(['$x_0$'], dtype=object)

However, if it is sliced to a series, the latex labels are lost:

>>> ns['x0']
      weights
0     0.000000e+00     0.839566
1     1.088407e-308    0.711607
2     4.830017e-301   -0.703718
3     6.951167e-262    0.980678
4     7.661544e-241    0.727711
                         ...
>>> ns['x0'].get_labels()
array([0.00000000e+000, 1.08840734e-308, 4.83001654e-301, ...,
       2.31405635e-001, 2.33317335e-001, 2.38103982e-001])

This is unfortunate, as all the plotting uses the series rather than dataframes, and it would be nice if the axes and legends could use the latex labels where they exist.

>>> ns.plot.kde_2d('x0', 'x1')

[AdamOrmondroyd]

I think the degenerate meaning of "labels" is quite confusing

[AdamOrmondroyd]

If one instead slices using the full multiindex, then this is used for the axis label:

>>> ns.plot.kde_2d(('x0', '$x_0$'), ('x1', '$x_1$'))

I believe the axis title matches the Series.name, so perhaps we could do something clever here?

[AdamOrmondroyd]

If NestedSamples are cast to DataFrame, then labels are retained after slicing, and shown in e.g.

ns['x0'].plot.kde()
df = pd.DataFrame(ns)
df['x0'].plot.kde()
plt.show()

NestedSamples

DataFrame