cythonize boundscheck

Author: chris-b1

pandas/algos.pyx

@@ -1302,7 +1302,7 @@ def group_last_bin_object(ndarray[object, ndim=2] out,
             else:
                 out[i, j] = resx[i, j]
 
-cdef inline float64_t _median_linear(float64_t* a, int n): 
+cdef inline float64_t _median_linear(float64_t* a, int n):
     cdef int i, j, na_count = 0
     cdef float64_t result
     cdef float64_t* tmp

pandas/core/indexing.py

@@ -9,7 +9,8 @@
                                  is_categorical_dtype,
                                  is_list_like,
                                  is_sequence,
-                                 is_scalar)
+                                 is_scalar,
+                                 _ensure_int64)
 from pandas.types.missing import isnull, _infer_fill_value
 
 from pandas.core.index import Index, MultiIndex
@@ -1851,23 +1852,11 @@ def maybe_convert_indices(indices, n):
     """ if we have negative indicies, translate to postive here
     if have indicies that are out-of-bounds, raise an IndexError
     """
-    if isinstance(indices, list):
-        indices = np.array(indices)
-        if len(indices) == 0:
-            # If list is empty, np.array will return float and cause indexing
-            # errors.
-            return np.empty(0, dtype=np.int_)
-
-    # cythonize this to one pass?
-    mask = indices < 0
-    if mask.any():
-        # don't mutate original array
-        indices = indices.copy()
-        indices[mask] += n
-    mask = (indices >= n) | (indices < 0)
-    if mask.any():
-        raise IndexError("indices are out-of-bounds")
-    return indices
+    from pandas.algos import take_bounds_check
+    indices = _ensure_int64(indices)
+    out = np.empty(len(indices), dtype='int64')
+    take_bounds_check(indices, out, n)
+    return out
 
 
 def maybe_convert_ix(*args):

pandas/indexes/base.py

@@ -1496,8 +1496,6 @@ def _assert_take_fillable(self, values, indices, allow_fill=True,
             taken = algos.take_nd(values, indices, allow_fill=False)
         return taken
 
-
-
     @cache_readonly
     def _isnan(self):
         """ return if each value is nan"""

pandas/indexes/category.py

@@ -14,7 +14,6 @@
                                     deprecate_kwarg)
 from pandas.core.config import get_option
 from pandas.indexes.base import Index, _index_shared_docs
-import pandas.core.algorithms as algos
 import pandas.core.base as base
 import pandas.core.missing as missing
 import pandas.indexes.base as ibase

pandas/src/algos_take_helper.pxi

@@ -4947,3 +4947,22 @@ def take_2d_multi_object_object(ndarray[object, ndim=2] values,
                     out[i, j] = fv
                 else:
                     out[i, j] = values[idx, idx1[j]]
+
+
+@cython.wraparound(False)
+@cython.boundscheck(False)
+def take_bounds_check(int64_t[:] indexer, int64_t[:] out, int64_t n):
+    cdef:
+        Py_ssize_t i
+        int64_t label
+
+    with nogil:
+        for i in range(indexer.shape[0]):
+            label = indexer[i]
+            if label < 0:
+                label += n
+
+            if label >= n or label < 0:
+                with gil:
+                    raise IndexError("indicies are out of bounds")
+            out[i] = label

pandas/src/algos_take_helper.pxi.in

@@ -258,4 +258,23 @@ def take_2d_multi_{{name}}_{{dest}}(ndarray[{{c_type_in}}, ndim=2] values,
                 else:
                     out[i, j] = {{preval}}values[idx, idx1[j]]{{postval}}
 
-{{endfor}}
+{{endfor}}
+
+
+@cython.wraparound(False)
+@cython.boundscheck(False)
+def take_bounds_check(int64_t[:] indexer, int64_t[:] out, int64_t n):
+    cdef:
+        Py_ssize_t i
+        int64_t label
+
+    with nogil:
+        for i in range(indexer.shape[0]):
+            label = indexer[i]
+            if label < 0:
+                label += n
+
+            if label >= n or label < 0:
+                with gil:
+                    raise IndexError("indicies are out of bounds")
+            out[i] = label