PDEP-4: Consistent datetime parsing
The suggestion is that:
to_datetimebecomes strict and uses the same datetime format to parse all elements in its input. The format will either be inferred from the first non-NaN element (if
formatis not provided by the user), or from
infer_datetime_formatbe deprecated (as a strict version of it will become the default);
- an easy workaround for non-strict parsing be clearly documented.
Motivation and Scope
Pandas date parsing is very flexible, but arguably too much so - see https://github.com/pandas-dev/pandas/issues/12585 and linked issues for how much confusion this causes. Pandas can swap format midway, and though this is documented, it regularly breaks users' expectations.
In : pd.to_datetime(['12-01-2000 00:00:00', '13-01-2000 00:00:00']) Out: DatetimeIndex(['2000-12-01', '2000-01-13'], dtype='datetime64[ns]', freq=None)
The user was almost certainly intending the data to be read as "12th of January, 13th of January". However, it's read as "1st of December, 13th of January". No warning or error is thrown.
Currently, the only way to ensure consistent parsing is by explicitly passing
format=. The argument
isn't strict, can be called together with
format, and can still break users' expectations:
In : pd.to_datetime(['12-01-2000 00:00:00', '13-01-2000 00:00:00'], infer_datetime_format=True) Out: DatetimeIndex(['2000-12-01', '2000-01-13'], dtype='datetime64[ns]', freq=None)
Concretely, the suggestion is:
- if no
pandaswill guess the format from the first non-NaN row and parse the rest of the input according to that format. Errors will be handled according to the
errorsargument - there will be no silent switching of format;
infer_datetime_formatwill be deprecated;
yearfirstwill continue working as they currently do;
- if the format cannot be guessed from the first non-NaN row, a
UserWarningwill be thrown, encouraging users to explicitly pass in a format. Note that this should only happen for invalid inputs such as
'a'(which would later throw a
ParserErroranyway), or inputs such as
'00:12:13', which would currently get converted to
If a user has dates in a mixed format, they can still use flexible parsing and accept the risks that poses, e.g.:
In : pd.to_datetime(['12-01-2000 00:00:00', '13-01-2000 00:00:00'], format='mixed') Out: DatetimeIndex(['2000-12-01', '2000-01-13'], dtype='datetime64[ns]', freq=None)
or, if their dates are all ISO8601,
In : pd.to_datetime(['2020-01-01', '2020-01-01 03:00'], format='ISO8601') Out: DatetimeIndex(['2020-01-01 00:00:00', '2020-01-01 03:00:00'], dtype='datetime64[ns]', freq=None)
Usage and Impact
My expectation is that the impact would be a net-positive:
- potentially severe bugs in people's code will be caught early;
- users who actually want mixed formats can still parse them, but now they'd be forced to be very explicit about it;
- the codebase would be noticeably simplified.
As far as I can tell, there is no chance of introducing bugs.
The whatsnew notes read
In the next major version release, 2.0, several larger API changes are being considered without a formal deprecation.
I'd suggest making this change as part of the above, because:
- it would only help prevent bugs, not introduce any;
- given the severity of bugs that can result from the current behaviour, waiting another 2 years until pandas 3.0.0 would potentially cause a lot of damage.
Note that this wouldn't mean getting rid of
dateutil.parser, as that would still be used within
guess_datetime_format. With this proposal, however, subsequent rows would be parsed with the guessed format rather than repeatedly calling
dateutil.parser and risk having it silently switch format
Finally, the function
from pandas._libs.tslibs.parsing import guess_datetime_format would be made public, under
Out of scope
We could make
guess_datetime_format smarter by using a random sample of elements to infer the format.
- 18 September 2022: Initial draft
- 25 January 2023: Amended to mention