Refactor groupby helper from tempita to fused types

Author: noamher

pandas/_libs/groupby.pyx

@@ -382,5 +382,61 @@ def group_any_all(uint8_t[:] out,
                 out[lab] = flag_val
 
 
+ctypedef fused floating:
+    float32_t
+    float64_t
+
+
+@cython.wraparound(False)
+@cython.boundscheck(False)
+def group_add_floating(floating[:, :] out,
+                       int64_t[:] counts,
+                       floating[:, :] values,
+                       const int64_t[:] labels,
+                       Py_ssize_t min_count=0):
+    """
+    Only aggregates on axis=0
+    """
+    cdef:
+        Py_ssize_t i, j, N, K, lab, ncounts = len(counts)
+        floating val, count
+        ndarray[floating, ndim=2] sumx, nobs
+
+    if not len(values) == len(labels):
+        raise AssertionError("len(index) != len(labels)")
+
+    nobs = np.zeros_like(out)
+    sumx = np.zeros_like(out)
+
+    N, K = (<object>values).shape
+
+    with nogil:
+
+        for i in range(N):
+            lab = labels[i]
+            if lab < 0:
+                continue
+
+            counts[lab] += 1
+            for j in range(K):
+                val = values[i, j]
+
+                # not nan
+                if val == val:
+                    nobs[lab, j] += 1
+                    sumx[lab, j] += val
+
+        for i in range(ncounts):
+            for j in range(K):
+                if nobs[i, j] < min_count:
+                    out[i, j] = NAN
+                else:
+                    out[i, j] = sumx[i, j]
+
+
+group_add_float32 = group_add_floating
+group_add_float64 = group_add_floating
+
+
 # generated from template
 include "groupby_helper.pxi"

pandas/_libs/groupby_helper.pxi.in

@@ -9,7 +9,7 @@ cdef extern from "numpy/npy_math.h":
 _int64_max = np.iinfo(np.int64).max
 
 # ----------------------------------------------------------------------
-# group_add, group_prod, group_var, group_mean, group_ohlc
+# group_prod, group_var, group_mean, group_ohlc
 # ----------------------------------------------------------------------
 
 {{py:
@@ -27,53 +27,6 @@ def get_dispatch(dtypes):
 {{for name, c_type in get_dispatch(dtypes)}}
 
 
-@cython.wraparound(False)
-@cython.boundscheck(False)
-def group_add_{{name}}({{c_type}}[:, :] out,
-                       int64_t[:] counts,
-                       {{c_type}}[:, :] values,
-                       const int64_t[:] labels,
-                       Py_ssize_t min_count=0):
-    """
-    Only aggregates on axis=0
-    """
-    cdef:
-        Py_ssize_t i, j, N, K, lab, ncounts = len(counts)
-        {{c_type}} val, count
-        ndarray[{{c_type}}, ndim=2] sumx, nobs
-
-    if not len(values) == len(labels):
-        raise AssertionError("len(index) != len(labels)")
-
-    nobs = np.zeros_like(out)
-    sumx = np.zeros_like(out)
-
-    N, K = (<object>values).shape
-
-    with nogil:
-
-        for i in range(N):
-            lab = labels[i]
-            if lab < 0:
-                continue
-
-            counts[lab] += 1
-            for j in range(K):
-                val = values[i, j]
-
-                # not nan
-                if val == val:
-                    nobs[lab, j] += 1
-                    sumx[lab, j] += val
-
-        for i in range(ncounts):
-            for j in range(K):
-                if nobs[i, j] < min_count:
-                    out[i, j] = NAN
-                else:
-                    out[i, j] = sumx[i, j]
-
-
 @cython.wraparound(False)
 @cython.boundscheck(False)
 def group_prod_{{name}}({{c_type}}[:, :] out,

pandas/core/groupby/ops.py

@@ -380,7 +380,7 @@ def get_func(fname):
             # otherwise find dtype-specific version, falling back to object
             for dt in [dtype_str, 'object']:
                 f = getattr(libgroupby, "{fname}_{dtype_str}".format(
-                    fname=fname, dtype_str=dtype_str), None)
+                    fname=fname, dtype_str=dt), None)
                 if f is not None:
                     return f