Skip to content

DEPR: removal of deprecation warnings for float indexers #12246

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

DEPR: removal of deprecation warnings for float indexers #12246

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions doc/source/release.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,6 +52,7 @@ Highlights include:
- ``pd.test()`` top-level nose test runner is available (:issue:`4327`)
- Adding support for a ``RangeIndex`` as a specialized form of the ``Int64Index`` for memory savings, see :ref:`here <whatsnew_0180.enhancements.rangeindex>`.
- API breaking ``.resample`` changes to make it more ``.groupby`` like, see :ref:`here <whatsnew_0180.breaking.resample>`.
- Removal of support for deprecated float indexers; these will now raise a ``TypeError``, see :ref:`here <whatsnew_0180.enhancements.float_indexers>`.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe mention that it only is about positional indexing? Otherwise people could think FloatIndex is removed.


See the :ref:`v0.18.0 Whatsnew <whatsnew_0180>` overview for an extensive list
of all enhancements and bugs that have been fixed in 0.17.1.
Expand Down
39 changes: 38 additions & 1 deletion doc/source/whatsnew/v0.18.0.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ Highlights include:
- ``pd.test()`` top-level nose test runner is available (:issue:`4327`)
- Adding support for a ``RangeIndex`` as a specialized form of the ``Int64Index`` for memory savings, see :ref:`here <whatsnew_0180.enhancements.rangeindex>`.
- API breaking ``.resample`` changes to make it more ``.groupby`` like, see :ref:`here <whatsnew_0180.breaking.resample>`.
- Removal of support for deprecated float indexers; these will now raise a ``TypeError``, see :ref:`here <whatsnew_0180.enhancements.float_indexers>`.

Check the :ref:`API Changes <whatsnew_0180.api_breaking>` and :ref:`deprecations <whatsnew_0180.deprecations>` before updating.

Expand Down Expand Up @@ -865,9 +866,45 @@ Deprecations
is better handled by matplotlib's `style sheets`_ (:issue:`11783`).


.. _style sheets: http://matplotlib.org/users/style_sheets.html

.. _whatsnew_0180.float_indexers:

.. _style sheets: http://matplotlib.org/users/style_sheets.html
Removal of deprecated float indexers
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

In :issue:`4892` indexing with floating point numbers on a non-``Float64Index`` was deprecated (in version 0.14.0).
In 0.18.0, this deprecation warning is removed and these will now raise a ``TypeError``. (:issue:`12165`)

Previous Behavior:

.. code-block:

In [1]: s = Series([1,2,3])
In [2]: s[1.0]
FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
Out[2]: 2

In [3]: s.iloc[1.0]
FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
Out[3]: 2

In [4]: s.loc[1.0]
FutureWarning: scalar indexers for index type Int64Index should be integers and not floating point
Out[4]: 2

New Behavior:

.. code-block:

In [4]: s[1.0]
TypeError: cannot do label indexing on <class 'pandas.indexes.range.RangeIndex'> with these indexers [1.0] of <type 'float'>

In [4]: s.iloc[1.0]
TypeError: cannot do label indexing on <class 'pandas.indexes.range.RangeIndex'> with these indexers [1.0] of <type 'float'>

In [4]: s.loc[1.0]
TypeError: cannot do label indexing on <class 'pandas.indexes.range.RangeIndex'> with these indexers [1.0] of <type 'float'>

.. _whatsnew_0180.prior_deprecations:

Expand Down
38 changes: 30 additions & 8 deletions pandas/core/indexing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,7 +115,11 @@ def _get_setitem_indexer(self, key):

try:
return self._convert_to_indexer(key, is_setter=True)
except TypeError:
except TypeError as e:

# invalid indexer type vs 'other' indexing errors
if 'cannot do' in str(e):
raise
raise IndexingError(key)

def __setitem__(self, key, value):
Expand Down Expand Up @@ -312,6 +316,18 @@ def _setitem_with_indexer(self, indexer, value):
index = self.obj.index
new_index = index.insert(len(index), indexer)

# we have a coerced indexer, e.g. a float
# that matches in an Int64Index, so
# we will not create a duplicate index, rather
# index to that element
# e.g. 0.0 -> 0
# GH12246
if index.is_unique:
new_indexer = index.get_indexer([new_index[-1]])
if (new_indexer != -1).any():
return self._setitem_with_indexer(new_indexer,
value)

# this preserves dtype of the value
new_values = Series([value])._values
if len(self.obj._values):
Expand Down Expand Up @@ -1091,8 +1107,17 @@ def _convert_to_indexer(self, obj, axis=0, is_setter=False):
"""
labels = self.obj._get_axis(axis)

# if we are a scalar indexer and not type correct raise
obj = self._convert_scalar_indexer(obj, axis)
if isinstance(obj, slice):
return self._convert_slice_indexer(obj, axis)

# try to find out correct indexer, if not type correct raise
try:
obj = self._convert_scalar_indexer(obj, axis)
except TypeError:

# but we will allow setting
if is_setter:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Does this mean that series[1.0] will TypeError, but series[1.0] = 2 will succeed? Or does that raise later on?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

short answer is yes, though, series[1.0] = 2 works because its a label assignment.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm, this seems not great (on your branch)

In [10]: s = Series([1,2,3])

In [11]: s[1.] = 5

In [12]: s
Out[12]:
0    1
1    2
2    3
1    5
dtype: int64

In [13]: s.index
Out[13]: Int64Index([0, 1, 2, 1], dtype='int64')

So before this would have warned about the Floating scalar, cast to an int, and then overwritten the value at index=1. I guess that would make this an API change, but this seems like a pretty hairy edge-case

I think I see why you would want to allow setting of floats that could be cast as integers, but maybe you could explain your thinking a bit more?

EDIT: I accidentally left out a "not" in my first line, so now I sound like a sarcastic jerk 😄

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So I was fixing this bug

In [1]: s = Series([1,2,3])

In [2]: s
Out[2]: 
0    1
1    2
2    3
dtype: int64

In [3]: s.index
Out[3]: RangeIndex(start=0, stop=3, step=1)

In [4]: s['0'] = 5

In [5]: s
Out[5]: 
0    1
1    2
2    3
0    5
dtype: int64

In [6]: s.index
Out[6]: Index([0, 1, 2, u'0'], dtype='object')

so s[1.0] = 1 should change this to a Float64Index

but I couldn't get it to work and it seemed innocuous enough. Let me have another look. This is a giant rabbit hole.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more. 

In [1]: s = Series([1,2,3])

This is on current master. This should work and coerce to a Float64Index.

In [2]: s[1.1] = 1
pandas/core/indexing.py:1095: FutureWarning: scalar indexers for index type RangeIndex should be integers and not floating point
IndexingError: 1.1

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So I fixed everything, but had to change this:

What kind of index should this be:
Index([1., 2., 3.], dtype=int)
Index([1., 2, 3.], dtype=int)

w/o actually running it?

pass

# see if we are positional in nature
is_int_index = labels.is_integer()
Expand Down Expand Up @@ -1131,10 +1156,7 @@ def _convert_to_indexer(self, obj, axis=0, is_setter=False):

return obj

if isinstance(obj, slice):
return self._convert_slice_indexer(obj, axis)

elif is_nested_tuple(obj, labels):
if is_nested_tuple(obj, labels):
return labels.get_locs(obj)
elif is_list_like_indexer(obj):
if is_bool_indexer(obj):
Expand Down Expand Up @@ -1278,7 +1300,7 @@ def _get_slice_axis(self, slice_obj, axis=0):

labels = obj._get_axis(axis)
indexer = labels.slice_indexer(slice_obj.start, slice_obj.stop,
slice_obj.step)
slice_obj.step, kind=self.name)

if isinstance(indexer, slice):
return self._slice(indexer, axis=axis, kind='iloc')
Expand Down
106 changes: 67 additions & 39 deletions pandas/indexes/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,12 +21,12 @@
from pandas.core.missing import _clean_reindex_fill_method
from pandas.core.common import (isnull, array_equivalent,
is_object_dtype, is_datetimetz, ABCSeries,
ABCPeriodIndex,
ABCPeriodIndex, ABCMultiIndex,
_values_from_object, is_float, is_integer,
is_iterator, is_categorical_dtype,
_ensure_object, _ensure_int64, is_bool_indexer,
is_list_like, is_bool_dtype,
is_integer_dtype)
is_integer_dtype, is_float_dtype)
from pandas.core.strings import StringAccessorMixin

from pandas.core.config import get_option
Expand Down Expand Up @@ -162,7 +162,46 @@ def __new__(cls, data=None, dtype=None, copy=False, name=None,

if dtype is not None:
try:
data = np.array(data, dtype=dtype, copy=copy)

# we need to avoid having numpy coerce
# things that look like ints/floats to ints unless
# they are actually ints, e.g. '0' and 0.0
# should not be coerced
# GH 11836
if is_integer_dtype(dtype):
inferred = lib.infer_dtype(data)
if inferred == 'integer':
data = np.array(data, copy=copy, dtype=dtype)
elif inferred in ['floating', 'mixed-integer-float']:

# if we are actually all equal to integers
# then coerce to integer
from .numeric import Int64Index, Float64Index
try:
res = data.astype('i8')
if (res == data).all():
return Int64Index(res, copy=copy,
name=name)
except (TypeError, ValueError):
pass

# return an actual float index
return Float64Index(data, copy=copy, dtype=dtype,
name=name)

elif inferred == 'string':
pass
else:
data = data.astype(dtype)
elif is_float_dtype(dtype):
inferred = lib.infer_dtype(data)
if inferred == 'string':
pass
else:
data = data.astype(dtype)
else:
data = np.array(data, dtype=dtype, copy=copy)

except (TypeError, ValueError):
pass

Expand Down Expand Up @@ -930,35 +969,32 @@ def _convert_scalar_indexer(self, key, kind=None):
kind : optional, type of the indexing operation (loc/ix/iloc/None)

right now we are converting
floats -> ints if the index supports it
"""

def to_int():
ikey = int(key)
if ikey != key:
return self._invalid_indexer('label', key)
return ikey

if kind == 'iloc':
if is_integer(key):
return key
elif is_float(key):
key = to_int()
warnings.warn("scalar indexers for index type {0} should be "
"integers and not floating point".format(
type(self).__name__),
FutureWarning, stacklevel=5)
return key
return self._invalid_indexer('label', key)
else:

if is_float(key):
if isnull(key):
return self._invalid_indexer('label', key)
warnings.warn("scalar indexers for index type {0} should be "
"integers and not floating point".format(
type(self).__name__),
FutureWarning, stacklevel=3)
return to_int()
if len(self):

# we can safely disallow
# if we are not a MultiIndex
# or a Float64Index
# or have mixed inferred type (IOW we have the possiblity
# of a float in with say strings)
if is_float(key):
if not (isinstance(self, ABCMultiIndex,) or
self.is_floating() or self.is_mixed()):
return self._invalid_indexer('label', key)

# we can disallow integers with loc
# if could not contain and integer
elif is_integer(key) and kind == 'loc':
if not (isinstance(self, ABCMultiIndex,) or
self.holds_integer() or self.is_mixed()):
return self._invalid_indexer('label', key)

return key

Expand Down Expand Up @@ -991,14 +1027,6 @@ def f(c):
v = getattr(key, c)
if v is None or is_integer(v):
return v

# warn if it's a convertible float
if v == int(v):
warnings.warn("slice indexers when using iloc should be "
"integers and not floating point",
FutureWarning, stacklevel=7)
return int(v)

self._invalid_indexer('slice {0} value'.format(c), v)

return slice(*[f(c) for c in ['start', 'stop', 'step']])
Expand Down Expand Up @@ -1057,7 +1085,7 @@ def is_int(v):
indexer = key
else:
try:
indexer = self.slice_indexer(start, stop, step)
indexer = self.slice_indexer(start, stop, step, kind=kind)
except Exception:
if is_index_slice:
if self.is_integer():
Expand Down Expand Up @@ -1891,10 +1919,7 @@ def get_value(self, series, key):
s = _values_from_object(series)
k = _values_from_object(key)

# prevent integer truncation bug in indexing
if is_float(k) and not self.is_floating():
raise KeyError

k = self._convert_scalar_indexer(k, kind='getitem')
try:
return self._engine.get_value(s, k,
tz=getattr(series.dtype, 'tz', None))
Expand Down Expand Up @@ -2236,6 +2261,7 @@ def reindex(self, target, method=None, level=None, limit=None,
if self.equals(target):
indexer = None
else:

if self.is_unique:
indexer = self.get_indexer(target, method=method,
limit=limit,
Expand Down Expand Up @@ -2722,7 +2748,9 @@ def _maybe_cast_slice_bound(self, label, side, kind):
# datetimelike Indexes
# reject them
if is_float(label):
self._invalid_indexer('slice', label)
if not (kind in ['ix'] and (self.holds_integer() or
self.is_floating())):
self._invalid_indexer('slice', label)

# we are trying to find integer bounds on a non-integer based index
# this is rejected (generally .loc gets you here)
Expand Down
23 changes: 20 additions & 3 deletions pandas/indexes/numeric.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,9 +42,14 @@ def _maybe_cast_slice_bound(self, label, side, kind):
"""

# we are a numeric index, so we accept
# integer/floats directly
if not (com.is_integer(label) or com.is_float(label)):
self._invalid_indexer('slice', label)
# integer directly
if com.is_integer(label):
pass

# disallow floats only if we not-strict
elif com.is_float(label):
if not (self.is_floating() or kind in ['ix']):
self._invalid_indexer('slice', label)

return label

Expand Down Expand Up @@ -200,6 +205,18 @@ def __new__(cls, data=None, dtype=None, copy=False, name=None,

if dtype is None:
dtype = np.float64
dtype = np.dtype(dtype)

# allow integer / object dtypes to be passed, but coerce to float64
if dtype.kind in ['i', 'O']:
dtype = np.float64

elif dtype.kind in ['f']:
pass

else:
raise TypeError("cannot support {0} dtype in "
"Float64Index".format(dtype))

try:
subarr = np.array(data, dtype=dtype, copy=copy)
Expand Down
Loading