Make common impl. with Index.searchsorted part II

Author: topper-123

pandas/core/common.py

@@ -16,6 +16,8 @@
 from pandas.compat import iteritems, PY2, PY36, OrderedDict
 from pandas.core.dtypes.generic import ABCSeries, ABCIndex, ABCIndexClass
 from pandas.core.dtypes.common import (is_integer, is_integer_dtype,
+ is_float_dtype, is_object_dtype,
+ is_categorical_dtype,
  is_numeric_dtype, is_number,
  is_scalar, ensure_platform_int)
 from pandas.core.dtypes.inference import _iterable_not_string
@@ -447,7 +449,7 @@ def ensure_within_integer_bounds(value, dtype):
  ValueError : if value is outside the bounds set in iinfo(dtype)
  """
  if PY2:
- # python 2 allows e.g. "a" < 1, avoid this
+ # python 2 allows "a" < 1, avoid such nonsense
  if not (is_number(value) or is_numeric_dtype(value)):
  msg = "value must be a number, was type {}"
  raise ValueError(msg.format(value))
@@ -481,24 +483,30 @@ def searchsorted_integer(arr, value, side="left", sorter=None):
  # but float 2.2 should *not* be converted to int 2
  value = np.asarray(value, dtype=dtype)
 
- return arr.searchsorted(value, side=side, sorter=sorter)
+ return np.searchsorted(arr, value, side=side, sorter=sorter)
 
 
 def searchsorted(arr, value, side="left", sorter=None):
  """
- Do a arr.searchsorted(value) with adjustments for dtypes.
+ Find indices where elements should be inserted to maintain order.
 
- :func:`numpy.searchsorted` is only fast if value is of same dtype
- as the searched array. Else numpy recasts arr to a higher dtype, which
- causes a slowdown. Below we ensure that value has the right dtype
- for giving fast results for arr.searchsorted, when possible.
+ Find the indices into a sorted array-like `arr` such that, if the
+ corresponding elements in `value` were inserted before the indices,
+ the order of `arr` would be preserved.
 
- See :meth:`Index.searchsorted` for details on parameters and return value.
+ See :class:`IndexOpsMixin.searchsorted` for more details and examples.
  """
  if sorter is not None:
  sorter = ensure_platform_int(sorter)
 
  if is_integer_dtype(arr):
  return searchsorted_integer(arr, value, side=side, sorter=sorter)
+ elif (is_object_dtype(arr) or is_float_dtype(arr) or
+ is_categorical_dtype(arr)):
+ return arr.searchsorted(value, side=side, sorter=sorter)
  else:
+ # fallback solution. E.g. arr is an array with dtype='datetime64[ns]'
+ # and value is a pd.Timestamp
+ from pandas.core.series import Series
+ value = Series(value)._values
  return arr.searchsorted(value, side=side, sorter=sorter)

pandas/core/series.py

@@ -2087,18 +2087,8 @@ def __rmatmul__(self, other):
  @Appender(base._shared_docs['searchsorted'])
  @deprecate_kwarg(old_arg_name='v', new_arg_name='value')
  def searchsorted(self, value, side='left', sorter=None):
- simple_types = (is_integer_dtype, is_float_dtype, is_object_dtype,
- is_categorical_dtype)
-
- if any(is_dtype(self) for is_dtype in simple_types):
- result = com.searchsorted(self._values, value,
- side=side, sorter=sorter)
- else:
- # e.g. self is datetimelike and value is a pd.Timestamp
- if sorter is not None:
- sorter = ensure_platform_int(sorter)
- value = Series(value)._values
- result = self._values.searchsorted(value, side=side, sorter=sorter)
+ result = com.searchsorted(self._values, value,
+ side=side, sorter=sorter)
 
  if is_scalar(result):
  # ensure that a 1-dim array is returned