CLN: assorted cleanups (#27301)

Author: jbrockmendel

pandas/core/computation/expressions.py

@@ -12,9 +12,10 @@
 
 from pandas._config import get_option
 
+from pandas._libs.lib import values_from_object
+
 from pandas.core.dtypes.generic import ABCDataFrame
 
-import pandas.core.common as com
 from pandas.core.computation.check import _NUMEXPR_INSTALLED
 
 if _NUMEXPR_INSTALLED:
@@ -129,9 +130,7 @@ def _evaluate_numexpr(op, op_str, a, b, truediv=True, reversed=False, **eval_kwa
 
 def _where_standard(cond, a, b):
     return np.where(
-        com.values_from_object(cond),
-        com.values_from_object(a),
-        com.values_from_object(b),
+        values_from_object(cond), values_from_object(a), values_from_object(b)
     )
 
 

pandas/core/internals/blocks.py

@@ -397,10 +397,6 @@ def fillna(self, value, limit=None, inplace=False, downcast=None):
                 raise ValueError("Limit must be an integer")
             if limit < 1:
                 raise ValueError("Limit must be greater than 0")
-            if self.ndim > 2:
-                raise NotImplementedError(
-                    "number of dimensions for 'fillna' is currently limited to 2"
-                )
             mask[mask.cumsum(self.ndim - 1) > limit] = False
 
         if not self._can_hold_na:
@@ -853,6 +849,8 @@ def setitem(self, indexer, value):
         `indexer` is a direct slice/positional indexer. `value` must
         be a compatible shape.
         """
+        transpose = self.ndim == 2
+
         # coerce None values, if appropriate
         if value is None:
             if self.is_numeric:
@@ -901,8 +899,8 @@ def setitem(self, indexer, value):
             dtype, _ = maybe_promote(arr_value.dtype)
             values = values.astype(dtype)
 
-        transf = (lambda x: x.T) if self.ndim == 2 else (lambda x: x)
-        values = transf(values)
+        if transpose:
+            values = values.T
 
         # length checking
         check_setitem_lengths(indexer, value, values)
@@ -961,7 +959,9 @@ def _is_empty_indexer(indexer):
 
         # coerce and try to infer the dtypes of the result
         values = self._try_coerce_and_cast_result(values, dtype)
-        block = self.make_block(transf(values))
+        if transpose:
+            values = values.T
+        block = self.make_block(values)
         return block
 
     def putmask(self, mask, new, align=True, inplace=False, axis=0, transpose=False):

pandas/core/internals/managers.py

@@ -583,8 +583,9 @@ def astype(self, dtype, **kwargs):
     def convert(self, **kwargs):
         return self.apply("convert", **kwargs)
 
-    def replace(self, **kwargs):
-        return self.apply("replace", **kwargs)
+    def replace(self, value, **kwargs):
+        assert np.ndim(value) == 0, value
+        return self.apply("replace", value=value, **kwargs)
 
     def replace_list(self, src_list, dest_list, inplace=False, regex=False):
         """ do a list replace """
@@ -617,6 +618,7 @@ def comp(s, regex=False):
             # replace ALWAYS will return a list
             rb = [blk if inplace else blk.copy()]
             for i, (s, d) in enumerate(zip(src_list, dest_list)):
+                # TODO: assert/validate that `d` is always a scalar?
                 new_rb = []
                 for b in rb:
                     m = masks[i][b.mgr_locs.indexer]

pandas/core/ops/__init__.py

@@ -11,7 +11,7 @@
 
 import numpy as np
 
-from pandas._libs import lib, ops as libops
+from pandas._libs import Timedelta, Timestamp, lib, ops as libops
 from pandas.errors import NullFrequencyError
 from pandas.util._decorators import Appender
 
@@ -87,7 +87,7 @@ def get_op_result_name(left, right):
         Usually a string
     """
     # `left` is always a pd.Series when called from within ops
-    if isinstance(right, (ABCSeries, pd.Index)):
+    if isinstance(right, (ABCSeries, ABCIndexClass)):
         name = _maybe_match_name(left, right)
     else:
         name = left.name
@@ -151,14 +151,14 @@ def maybe_upcast_for_op(obj):
         # GH#22390  cast up to Timedelta to rely on Timedelta
         # implementation; otherwise operation against numeric-dtype
         # raises TypeError
-        return pd.Timedelta(obj)
+        return Timedelta(obj)
     elif isinstance(obj, np.timedelta64) and not isna(obj):
         # In particular non-nanosecond timedelta64 needs to be cast to
         #  nanoseconds, or else we get undesired behavior like
         #  np.timedelta64(3, 'D') / 2 == np.timedelta64(1, 'D')
         # The isna check is to avoid casting timedelta64("NaT"), which would
         #  return NaT and incorrectly be treated as a datetime-NaT.
-        return pd.Timedelta(obj)
+        return Timedelta(obj)
     elif isinstance(obj, np.ndarray) and is_timedelta64_dtype(obj):
         # GH#22390 Unfortunately we need to special-case right-hand
         # timedelta64 dtypes because numpy casts integer dtypes to
@@ -1864,7 +1864,7 @@ def wrapper(self, other, axis=None):
                     )
                 msg = "\n".join(textwrap.wrap(msg.format(future=future)))
                 warnings.warn(msg, FutureWarning, stacklevel=2)
-                other = pd.Timestamp(other)
+                other = Timestamp(other)
 
             res_values = dispatch_to_index_op(op, self, other, pd.DatetimeIndex)
 
@@ -1890,7 +1890,7 @@ def wrapper(self, other, axis=None):
                 res_values, index=self.index, name=res_name
             ).rename(res_name)
 
-        elif isinstance(other, (np.ndarray, pd.Index)):
+        elif isinstance(other, (np.ndarray, ABCIndexClass)):
             # do not check length of zerodim array
             # as it will broadcast
             if other.ndim != 0 and len(self) != len(other):

pandas/tests/computation/test_eval.py

@@ -411,7 +411,7 @@ def check_single_invert_op(self, lhs, cmp1, rhs):
                 )
 
     def check_compound_invert_op(self, lhs, cmp1, rhs):
-        skip_these = "in", "not in"
+        skip_these = ["in", "not in"]
         ex = "~(lhs {0} rhs)".format(cmp1)
 
         msg = (

pandas/tests/io/json/test_pandas.py

@@ -1432,6 +1432,7 @@ def test_to_jsonl(self):
         assert result == expected
         assert_frame_equal(pd.read_json(result, lines=True), df)
 
+    # TODO: there is a near-identical test for pytables; can we share?
     def test_latin_encoding(self):
         # GH 13774
         pytest.skip("encoding not implemented in .to_json(), xref #13774")

pandas/tests/series/test_arithmetic.py

@@ -24,7 +24,7 @@ class TestSeriesFlexArithmetic:
         ],
     )
     @pytest.mark.parametrize(
-        "opname", ["add", "sub", "mul", "floordiv", "truediv", "div", "pow"]
+        "opname", ["add", "sub", "mul", "floordiv", "truediv", "pow"]
     )
     def test_flex_method_equivalence(self, opname, ts):
         # check that Series.{opname} behaves like Series.__{opname}__,
@@ -34,15 +34,8 @@ def test_flex_method_equivalence(self, opname, ts):
         other = ts[1](tser)
         check_reverse = ts[2]
 
-        if opname == "div":
-            pytest.skip("div test only for Py3")
-
         op = getattr(Series, opname)
-
-        if op == "div":
-            alt = operator.truediv
-        else:
-            alt = getattr(operator, opname)
+        alt = getattr(operator, opname)
 
         result = op(series, other)
         expected = alt(series, other)