Test 8-bit integers and raise error on 16-bit floats; add comments

Author: Christopher C. Aycock

pandas/src/joins_func_helper.pxi.in

@@ -169,6 +169,7 @@ def asof_join_{{on_dtype}}(ndarray[{{on_dtype}}] left_values,
 #----------------------------------------------------------------------
 
 def stringify(ndarray[object, ndim=2] xt):
+    """ Invokes .tostring() on each array entry within a 2D array. """
     cdef:
         Py_ssize_t n
         ndarray[object] result

pandas/tools/merge.py

@@ -949,7 +949,7 @@ def _asof_by_function(on_type, by_type):
     'int32': 'int32_t',
     'int16': 'int16_t',
     'int64': 'int64_t',
-    'float16': 'float',
+    'float16': 'error',
     'float32': 'float',
     'float64': 'double',
 }
@@ -959,6 +959,8 @@ def _get_cython_type(dtype):
     """ Given a dtype, return a C name like 'int64_t' or 'double' """
     type_name = _get_dtype(dtype).name
     ctype = _cyton_types.get(type_name, 'object')
+    if ctype == 'error':
+        raise MergeError('unsupported type: ' + type_name)
     return ctype
 
 
@@ -1063,6 +1065,7 @@ def _get_join_indexers(self):
         def flip_stringify(xs):
             """ flip an array of arrays and string-ify contents """
             xt = np.transpose(xs)
+            # numpy arrays aren't hashable, so we convert to a string
             return _join.stringify(_ensure_object(xt))
 
         # values to compare

pandas/tools/tests/test_merge_asof.py

@@ -565,17 +565,17 @@ def test_on_float(self):
 
     def test_on_specialized_type(self):
         # GH13936
-        for dtype in [np.uint16, np.uint32, np.uint64,
-                      np.int16, np.int32, np.int64,
+        for dtype in [np.uint8, np.uint16, np.uint32, np.uint64,
+                      np.int8, np.int16, np.int32, np.int64,
                       np.float32, np.float64]:
             df1 = pd.DataFrame({
-                'value': [5, 2, 25, 300, 78, 1040, 79],
+                'value': [5, 2, 25, 100, 78, 120, 79],
                 'symbol': list("ABCDEFG")},
                 columns=['symbol', 'value'])
             df1.value = dtype(df1.value)
 
             df2 = pd.DataFrame({
-                'value': [0, 100, 1000, 10000],
+                'value': [0, 80, 120, 125],
                 'result': list('xyzw')},
                 columns=['value', 'result'])
             df2.value = dtype(df2.value)
@@ -586,7 +586,7 @@ def test_on_specialized_type(self):
 
             expected = pd.DataFrame({
                 'symbol': list("BACEGDF"),
-                'value': [2, 5, 25, 78, 79, 300, 1040],
+                'value': [2, 5, 25, 78, 79, 100, 120],
                 'result': list('xxxxxyz')},
                 columns=['symbol', 'value', 'result'])
             expected.value = dtype(expected.value)
@@ -595,18 +595,18 @@ def test_on_specialized_type(self):
 
     def test_on_specialized_type_by_int(self):
         # GH13936
-        for dtype in [np.uint16, np.uint32, np.uint64,
-                      np.int16, np.int32, np.int64,
+        for dtype in [np.uint8, np.uint16, np.uint32, np.uint64,
+                      np.int8, np.int16, np.int32, np.int64,
                       np.float32, np.float64]:
             df1 = pd.DataFrame({
-                'value': [5, 2, 25, 300, 78, 1040, 79],
+                'value': [5, 2, 25, 100, 78, 120, 79],
                 'key': [1, 2, 3, 2, 3, 1, 2],
                 'symbol': list("ABCDEFG")},
                 columns=['symbol', 'key', 'value'])
             df1.value = dtype(df1.value)
 
             df2 = pd.DataFrame({
-                'value': [0, 100, 1000, 10000],
+                'value': [0, 80, 120, 125],
                 'key': [1, 2, 2, 3],
                 'result': list('xyzw')},
                 columns=['value', 'key', 'result'])
@@ -619,7 +619,7 @@ def test_on_specialized_type_by_int(self):
             expected = pd.DataFrame({
                 'symbol': list("BACEGDF"),
                 'key': [2, 1, 3, 3, 2, 2, 1],
-                'value': [2, 5, 25, 78, 79, 300, 1040],
+                'value': [2, 5, 25, 78, 79, 100, 120],
                 'result': [np.nan, 'x', np.nan, np.nan, np.nan, 'y', 'x']},
                 columns=['symbol', 'key', 'value', 'result'])
             expected.value = dtype(expected.value)