BUG: Choropleth mapping seems scrambled with empty classes

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of geopandas.

• (optional) I have confirmed this bug exists on the master branch of geopandas.

---

Code Sample, a copy-pastable example

from pysal.lib import examples
import mapclassify
import geopandas
import matplotlib.pyplot as plt

mx_ex = examples.load_example('mexico')
mx = geopandas.read_file(mx_ex.get_file_list()[0])

years = ['PCGDP1940', 'PCGDP1960', 'PCGDP1980', 'PCGDP2000']
pooled = mapclassify.Pooled(mx[years])

f, axs = plt.subplots(2, 2, figsize=(12, 12))
axs = axs.flatten()
for i, y in enumerate(years):
    mx.plot(
        y, 
        scheme='UserDefined', 
        classification_kwds={
            'bins': pooled.global_classifier.bins
        }, 
        legend=True, 
        ax=axs[i]
    )
    axs[i].set_axis_off()
    axs[i].set_title(y)
plt.tight_layout()
plt.show()

Problem description

The current behaviour generates a scrambled ledend when there aren’t values in any of the categories. Ideally, the same legend and classification should appear with the UserDefined classifier whether there are values or not.

Expected Output

Uniform legend across the four maps. Instead, this is the figure created:

Output of geopandas.show_versions()

This is run on the gds_env:6.1, versions info below:

SYSTEM INFO
-----------
python     : 3.8.8 | packaged by conda-forge | (default, Feb 20 2021, 16:22:27)  [GCC 9.3.0]
executable : /opt/conda/bin/python
machine    : Linux-5.4.0-77-generic-x86_64-with-glibc2.10

GEOS, GDAL, PROJ INFO
---------------------
GEOS       : 3.9.1
GEOS lib   : /opt/conda/lib/libgeos_c.so
GDAL       : 3.2.1
GDAL data dir: /opt/conda/share/gdal
PROJ       : 7.2.0
PROJ data dir: /opt/conda/share/proj

PYTHON DEPENDENCIES
-------------------
geopandas  : 0.9.0
pandas     : 1.2.3
fiona      : 1.8.18
numpy      : 1.20.1
shapely    : 1.7.1
rtree      : 0.9.7
pyproj     : 3.0.1
matplotlib : 3.3.4
mapclassify: 2.4.2
geopy      : 2.1.0
psycopg2   : 2.8.6 (dt dec pq3 ext lo64)
geoalchemy2: 0.8.4
pyarrow    : 3.0.0
pygeos     : 0.9

[martinfleis]

The colormap inconsistency should be fixed by #2019. The legend labels need to be handled separately.

[dfolch]

I think a min keyword solves the Mexico example, but I don’t think it solves the case where the lowest 2 (or more) bins are empty; or multiple upper end bins are empty.