Merge pull request #6837 from anntzer/fix-integer-normalization

tacaswell · web-flow · commit 84d979b9b55d · 2016-08-01T15:29:42.000-04:00
FIX: Normalize(<signed integer array>). closes #6825
diff --git a/lib/matplotlib/colors.py b/lib/matplotlib/colors.py
@@ -884,8 +884,9 @@ def process_value(value):
         if is_scalar:
             value = [value]
         dtype = np.min_scalar_type(value)
-        dtype = (np.float32 if dtype.itemsize <= 2
-                 else np.promote_types(dtype, float))
+        if np.issubdtype(dtype, np.integer) or dtype.type is np.bool_:
+            # bool_/int8/int16 -> float32; int32/int64 -> float64
+            dtype = np.promote_types(dtype, np.float32)
         result = np.ma.array(value, dtype=dtype, copy=True)
         return result, is_scalar
 
@@ -903,7 +904,9 @@ def __call__(self, value, clip=None):
         result, is_scalar = self.process_value(value)
 
         self.autoscale_None(result)
-        vmin, vmax = self.vmin, self.vmax
+        # Convert at least to float, without losing precision.
+        (vmin,), _ = self.process_value(self.vmin)
+        (vmax,), _ = self.process_value(self.vmax)
         if vmin == vmax:
             result.fill(0)   # Or should it be all masked?  Or 0.5?
         elif vmin > vmax:
@@ -927,7 +930,8 @@ def __call__(self, value, clip=None):
     def inverse(self, value):
         if not self.scaled():
             raise ValueError("Not invertible until scaled")
-        vmin, vmax = self.vmin, self.vmax
+        (vmin,), _ = self.process_value(self.vmin)
+        (vmax,), _ = self.process_value(self.vmax)
 
         if cbook.iterable(value):
             val = np.ma.asarray(value)
diff --git a/lib/matplotlib/tests/test_colors.py b/lib/matplotlib/tests/test_colors.py
@@ -194,6 +194,12 @@ def test_Normalize():
     _scalar_tester(norm, vals)
     _mask_tester(norm, vals)
 
+    # Handle integer input correctly (don't overflow when computing max-min,
+    # i.e. 127-(-128) here).
+    vals = np.array([-128, 127], dtype=np.int8)
+    norm = mcolors.Normalize(vals.min(), vals.max())
+    assert_array_equal(np.asarray(norm(vals)), [0, 1])
+
     # Don't lose precision on longdoubles (float128 on Linux):
     # for array inputs...
     vals = np.array([1.2345678901, 9.8765432109], dtype=np.longdouble)
