Correctly sort Categorical facets.

There's a bug where you get an error if you facet a geom_bar with a
single facet variable which is Categorical. This is an old bug, but
it's more visible now that there's a reason to use Categorical facet
variables. I think it's due to a bug in pandas:
pandas-dev/pandas#14011

Author: ChickenProp

ggplot/facets.py

@@ -4,6 +4,8 @@
 import pprint as pp
 from collections import OrderedDict
 
+from .utils import sorted_unique
+
 class Facet(object):
     def __init__(self, data, is_wrap, rowvar=None, colvar=None, nrow=None, ncol=None, scales=None):
         self.rowvar = rowvar
@@ -54,14 +56,14 @@ def __init__(self, data, is_wrap, rowvar=None, colvar=None, nrow=None, ncol=None
 
     def generate_subplot_index(self, data, rowvar, colvar):
         if rowvar and colvar:
-            for row_idx, row in enumerate(sorted(data[rowvar].unique())):
-                for col_idx, col in enumerate(sorted(data[colvar].unique())):
+            for row in sorted_unique(data[rowvar]):
+                for col in sorted_unique(data[colvar]):
                     yield (row, col)
         elif rowvar:
-            for row_idx, row in enumerate(sorted(data[rowvar].unique())):
+            for row in sorted_unique(data[rowvar]):
                 yield row
         elif colvar:
-            for col_idx, col in enumerate(sorted(data[colvar].unique())):
+            for col in sorted_unique(data[colvar]):
                 yield col
 
     def calculate_ndimensions(self, data, rowvar, colvar):

ggplot/ggplot.py

@@ -15,7 +15,7 @@
 from .themes import theme_gray
 from .themes import element_text
 from . import discretemappers
-from .utils import format_ticks
+from .utils import format_ticks, sorted_unique
 import urllib
 import base64
 import os
@@ -148,10 +148,8 @@ def add_labels(self):
         if self.facets.is_wrap:
             return
         if self.facets.rowvar:
-            for row, name in enumerate(sorted(self.data[self.facets.rowvar].unique())):
-                if self.facets.is_wrap==True:
-                    continue
-                elif self.facets.colvar:
+            for row, name in enumerate(sorted_unique(self.data[self.facets.rowvar])):
+                if self.facets.colvar:
                     ax = self.subplots[row][-1]
                 else:
                     ax = self.subplots[row]
@@ -160,7 +158,7 @@ def add_labels(self):
                 ax.set_ylabel(name, fontsize=10, rotation=-90)
 
         if self.facets.colvar:
-            for col, name in enumerate(sorted(self.data[self.facets.colvar].unique())):
+            for col, name in enumerate(sorted_unique(self.data[self.facets.colvar])):
                 if len(self.subplots.shape) > 1:
                     col = col % self.facets.ncol
                     ax = self.subplots[0][col]

ggplot/utils.py

@@ -96,3 +96,9 @@ def calc_n_bins(series):
     h = (2 * iqr) / (len(series)**(1/3.))
     k = (series.max() - series.min()) / h
     return k
+
+def sorted_unique(series):
+    """Return the unique values of *series*, correctly sorted."""
+    # This handles Categorical data types, which sorted(series.unique()) fails
+    # on. series.drop_duplicates() is slower than Series(series.unique()).
+    return list(pd.Series(series.unique()).sort_values())

tests/test_facets.py

@@ -1,5 +1,10 @@
+import pandas as pd
 from ggplot import *
 
+diamonds['cut'] = pd.Categorical(diamonds['cut'], ordered=True,
+                                 categories=['Fair', 'Good', 'Very Good', 'Premium', 'Ideal'])
+diamonds['clarity'] = pd.Categorical(diamonds['clarity'], ordered=True,
+                                     categories='I1 SI2 SI1 VS2 VS1 VVS2 VVS1 IF'.split())
 
 diaa = diamonds[['cut','color','table']]
 diab = diaa.groupby(['cut','color']).quantile([x/100.0 for x in range(0,100,5)])
@@ -9,3 +14,5 @@
 
 print ggplot(diamonds, aes(x='clarity', weight='price')) + geom_bar() + facet_grid('color', 'cut')
 print ggplot(diamonds, aes(x='clarity', weight='price', fill='color')) + geom_bar() + facet_grid('color', 'cut')
+
+print ggplot(diamonds, aes(x='clarity', y='price')) + geom_boxplot() + facet_wrap('cut')