do not rely on equalwidth bin assumption for cumhist; trapz doesn't work correctly with new histogram semantic

mmetz-bn · mmetz-bn · commit 21e579d7b9d1 · 2008-05-16T17:06:53.000Z
svn path=/trunk/matplotlib/; revision=5148
diff --git a/lib/matplotlib/axes.py b/lib/matplotlib/axes.py
@@ -5420,9 +5420,9 @@ def hist(self, x, bins=10, normed=False, cumulative=False,
              bottom=None, histtype='bar', align='edge',
              orientation='vertical', width=None, log=False, **kwargs):
         """
-        HIST(x, bins=10, normed=False, bottom=None, histtype='bar',
-             align='edge', orientation='vertical', width=None,
-             log=False, **kwargs)
+        HIST(x, bins=10, normed=False, cumulative=False,
+             bottom=None, histtype='bar', align='edge',
+             orientation='vertical', width=None, log=False, **kwargs)
 
         Compute the histogram of x.  bins is either an integer number of
         bins or a sequence giving the bins.  x are the data to be binned.
@@ -5437,13 +5437,13 @@ def hist(self, x, bins=10, normed=False, cumulative=False,
 
           # trapezoidal integration of the probability density function
           pdf, bins, patches = ax.hist(...)
-          print np.trapz(pdf, bins)
+          print np.sum(pdf * np.diff(bins))
 
-        If cumulative is True then histogram is computed where each bin
+        If cumulative is True then a histogram is computed where each bin
         gives the counts in that bin plus all bins for smaller values.
         The last bins gives the total number of datapoints.  If normed is
         also True then the histogram is normalized such that the last bin
-        equals one (assuming equally spaced bins).
+        equals one.
 
         histtype = 'bar' | 'step'. The type of histogram to draw.
         'bar' is a traditional bar-type histogram, 'step' generates
@@ -5469,10 +5469,10 @@ def hist(self, x, bins=10, normed=False, cumulative=False,
             normed=bool(normed), new=True)
         
         if cumulative:
-            n = n.cumsum()
             if normed:
-                # normalize to 1
-                n *= (bins[1]-bins[0])
+                n = (n * np.diff(bins)).cumsum()
+            else:
+                n = n.cumsum()
 
         if histtype == 'bar':
             if width is None:
