DEPR: Int64Index, UInt64Index & Float64Index #43028
Conversation
pandas/core/indexes/numeric.py
Outdated
| Notes | ||
| ----- | ||
| An NumericIndex instance can **only** contain numpy int64/32/16/8, uint64/32/16/8 or | ||
| float64/32/16 dtype. In Particulat, ``NumericIndex`` *can not* hold Pandas numeric |
topper-123
force-pushed
the
deprecate_old_numeric_indexes
branch
from
fcf5cd6
to
5afecb2
Compare
|
xref #43002 |
topper-123
force-pushed
the
deprecate_old_numeric_indexes
branch
from
098f9b0
to
ce2c306
Compare
lithomas1
added
Deprecate
|
IMHO, we probably should emit some sort of warning to users. I'm a little worried since 2.0 should be the next release after 1.4, and this is a somewhat big change. We can always suppress the deprecation warnings internally. |
|
I've made a new version where accessing >>> import pandas as pd
>>> idx = pd.Int64Index([8, 9])
FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.NumericIndex with the appropriate dtype instead.
>>> isinstance(idx, pd.Int64Index)
FutureWarning: pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.NumericIndex with the appropriate dtype instead.
Truebut this does not: >>> idx = pd.Index([8, 9])
>>> idx
Int64Index([8, 9], dtype='int64')
>>> isinstance(idx, pd.NumericIndex)
TrueThis last example doesn’t give warnings, because this access pattern is not deprecated, but will only change meaning in Pandas 2.0 (will return a NumericIndex rather than a Int64Index, + Int64Index is a subclass of NumericIndex). These warnings should IMO give users a good hint on where they need to update their code to stay compatible in Pandas 2.0. |
|
Got some failures still, but will work on them. The idea is completed, so commentd welcome. |
topper-123
force-pushed
the
deprecate_old_numeric_indexes
branch
from
d5a0689
to
33bb44f
Compare
doc/source/user_guide/advanced.rst
Outdated
| In pandas 2.0, :class:`NumericIndex` will become the default index type for numeric types | ||
| instead of ``Int64Index``, ``Float64Index`` and ``UInt64Index`` and those index types | ||
| will be removed. See :ref:`here <advanced.numericindex>` for more. | ||
| are therefore deprecated and will be removed in Pandas 2.0. |
There was a problem hiding this comment.
instead of saying Pandas 2.0, just a future version (we don't explicitly list the removal versions anywhere else i believe)
There was a problem hiding this comment.
same comment applies to all docs below
| __deprecated_num_index_names = ["Float64Index", "Int64Index", "UInt64Index"] | ||
|
|
||
|
|
||
| def __dir__(): |
There was a problem hiding this comment.
This is so Int64Index etc. show up in dir calls.
Int64Index etc. have been quite prominent parts of the public API, so I think it's better that they still show up for the duration of Pandas 1.x. This should be removed in Pandas 2.0.
There was a problem hiding this comment.
ok can you add a comment to this effect
|
Updated. |
|
Gentle ping… |
doc/source/user_guide/advanced.rst
Outdated
| will be removed. | ||
| ``RangeIndex`` however, will not be removed, as it represents an optimized version of an integer index. | ||
| instead of :class:`Int64Index`, :class:`Float64Index` and :class:`UInt64Index` and those index types | ||
| are therefore deprecated and will be removed in a futureversion. |
topper-123
force-pushed
the
deprecate_old_numeric_indexes
branch
from
60d16df
to
7ca2ef7
Compare
|
thanks @topper-123 very nice! |
|
@topper-123 want to enforce this deprecation and rip out some code? |
Deprecates Int64Index, UInt64Index & Float64Index.
I don't think it will be sensible to emit a deprecation warning for these because they have a lot of internal usage as is, so I've just added texts explaining that these are deprecated and will be removed in Pandas 2.0.
EDIT: it does now emit an deprecation warning when accessed from the main namespace. See changes to
pandas.__init__.pyfor implementation.