Merge pull request #963 from serge-sans-paille/fix/pshape-equality

Fix pshape eq neq operator

pythran/pythonic/include/types/tuple.hpp

@@ -945,7 +945,23 @@ namespace sutils
   }
 
   template <class S, class B>
-  bool equals(S const &s, B const &other)
+  typename std::enable_if<
+      std::tuple_size<S>::value == std::tuple_size<B>::value, bool>::type
+  equals(S const &s, B const &other)
+  {
+    return equals(
+        s, other,
+        std::integral_constant<size_t, std::tuple_size<S>::value - 1>());
+  }
+  template <class S, class B>
+  typename std::enable_if<
+      std::tuple_size<S>::value != std::tuple_size<B>::value, bool>::type
+  equals(S const &s, B const &other)
+  {
+    return false;
+  }
+  template <class S, class B>
+  bool equals(S const &s, B *other)
   {
     return equals(
         s, other,

pythran/tests/cases/_histogram.py

@@ -0,0 +1,54 @@
+#pythran export histogram_neq_edges_weights(int32[][], int32[], int32[][])
+#runas import numpy as np; d = np.arange(10, dtype=np.int32).reshape(5,2); w = np.ones_like(d, dtype=np.int32); b = np.arange(4,  dtype=np.int32); histogram_neq_edges_weights(d, b, w)
+
+import numpy as np
+
+
+
+def _histogram_neq_edges_weights(data, bin_edges, weights):
+    _BLOCK = 65536
+
+    bin_edges_length = (len(bin_edges) - 1)
+
+    hist = np.zeros(len(bin_edges), weights.dtype)
+
+    # omp parallel for private(j, tmp_hist, i, value, bin_idx)
+    for j in range(0, len(data), _BLOCK):
+        tmp_hist = np.zeros(len(bin_edges), weights.dtype)
+        tmp_data = data[j:j + _BLOCK]
+        tmp_weights = weights[j:j + _BLOCK]
+
+        for i in range(0, len(tmp_data)):
+            value = tmp_data[i]
+
+            # Ignore data if NaN or not within range
+            if np.isnan(value) or not (bin_edges[0] <= value <= bin_edges[-1]):
+                continue
+
+            # Search for index
+            bin_idx = 0
+            while bin_idx < bin_edges_length and bin_edges[bin_idx + 1] <= value:
+                bin_idx += 1
+
+            tmp_hist[bin_idx] += tmp_weights[i]
+
+        # omp critical
+        hist += tmp_hist
+
+    # Last bin element is inclusive
+    hist[-2] += hist[-1]
+
+    # Remove the redundant extra bin
+    return hist[:-1], bin_edges
+
+
+
+def histogram_neq_edges_weights(data, bin_edges, weights):
+    if weights.shape != data.shape:
+        raise ValueError('weights should have the same shape as data.')
+    weights = weights.ravel()
+
+    data = data.ravel()
+    return _histogram_neq_edges_weights(data, bin_edges, weights)
+
+