bpo-37999: No longer use __int__ in implicit integer conversions. #15636
Conversation
Only __index__ should be used to make integer conversions lossless.
serhiy-storchaka
requested review from
rhettinger,
tiran and
a team
as code owners
serhiy-storchaka
changed the title
Lib/datetime.py
Outdated
| @@ -2539,7 +2508,7 @@ def _name_from_offset(delta): | |||
| # Clean up unused names | |||
| del (_DAYNAMES, _DAYS_BEFORE_MONTH, _DAYS_IN_MONTH, _DI100Y, _DI400Y, | |||
| _DI4Y, _EPOCH, _MAXORDINAL, _MONTHNAMES, _build_struct_time, | |||
| _check_date_fields, _check_int_field, _check_time_fields, | |||
| _check_date_fields, _index, _check_time_fields, | |||
There was a problem hiding this comment.
Looks like there's an effort to keep things in alphabetical order here; maybe move _index to the right place with respect to that order?
serhiy-storchaka
force-pushed
the
int-implicit-lossless
branch
from
656c19e to
49aa158
Compare
There was a problem hiding this comment.
LGTM; I'm happy to see this finally reach completion.
Not sure I like the EDIT: This has been changedOverflowError for large negative values for math.factorial: that seems like a case of the implementation details leaking into the API.
Just to be clear, this PR doesn't change the fact that PyNumber_Index can still return something that's an instance of a strict subclass of int. Is that correct?
Yes, it is correct. |
…ted as integers Prior to pythongh-11952, several standard library functions that expected integer arguments would nevertheless silently accept (and truncate) non-integer arguments. This behaviour was deprecated in pythongh-11952, and removed in pythongh-15636. However, it may be possible to interpret some non-integer numeric types (such as `decimal.Decimal`s) as integers if they contain no fractional part. Implement `__index__` for `decimal.Decimal`, returning an integer representation of the value if it does not contain a fractional part or raising a `TypeError` if it does.
Only
__index__should be used to make integer conversions lossless.https://bugs.python.org/issue37999