Implement Series.prod() in new style #227
Conversation
| if not isinstance(self, SeriesType): | ||
| raise TypingError('{} The object must be a pandas.series. Given: {}'.format(_func_name, self)) | ||
|
|
||
| if not isinstance(self.data.dtype, (types.Integer, types.Float)): |
There was a problem hiding this comment.
I would use self.dtype instead of self.data.dtype, but perhaps it's also ok.
| # TODO: ignore non-numerics | ||
| # get series prod output types | ||
| dtypes = tuple(hpat.hiframes.pd_series_ext.SeriesAttribute.resolve_prod( | ||
| dtypes = tuple(numba.typing.arraydecl.ArrayAttribute.resolve_prod( |
There was a problem hiding this comment.
What issue you trying to solve by this change?
There was a problem hiding this comment.
The issue is exception with 'attribute not found'
|
|
||
|
|
||
| @overload_method(SeriesType, 'prod') | ||
| def hpat_pandas_series_prod(self, axis=None, skipna=None, level=None, numeric_only=None, min_count=0): |
There was a problem hiding this comment.
| def hpat_pandas_series_prod(self, axis=None, skipna=None, level=None, numeric_only=None, min_count=0): | |
| def hpat_pandas_series_prod(self, axis=None, skipna=True, level=None, numeric_only=None, min_count=0): |
According to documentation:
skipna : bool, default True
There was a problem hiding this comment.
None in Python is usually used as default value because Python consider default values as global values (see this).
Example:
def f(a=[]):
a.append(1)
return a
>>> f()
[1]
>>> f()
[1,1]
So common approach is:
def f(a=None):
"""Docstring: a: list, default []"""
if a is None:
a = []
...
There was a problem hiding this comment.
In this case with bool True is global and not mutable. But implementation of Pandas takes None as argument. Moreover Pandas is ready to get None so we also should be ready to get None and set it to True in the function.
So I think skipna : bool, default True is about logical value. It meas if you drop out skipna it will be considered as True.
Example:
S.prod()
S.prod(True)
Both calls are equivalent.
| skipna: :obj:`bool` object, default :obj:`True` | ||
| self: :obj:`pandas.Series` | ||
| input series | ||
| axis: :obj:`int` |
There was a problem hiding this comment.
Actually axis could be a string ('index' or 'columns'), not only an integer. You could add string type here. Moreover you could check on typing stage of the implementation that axis type is able to be a string.
There was a problem hiding this comment.
I have created #235 for discussing documentation for parameters.
Improve exception message for min_count value for Series.prod() Improve test Add tests for skipna=None as default Fix prod(skipna=None) and use subTest() for better error indication when testing on a set of data Describe the parameters and add types to docstring Use SeriesType().dtype instead of SeriesType.data.dtype Use numba.typing.arraydecl.ArrayAttribute.resolve_prod() instead of hpat.hiframes.pd_series_ext.SeriesAttribute.resolve_prod() because SeriesAttribute does not have resolve_prod Update docstring for hpat_pandas_series_prod Fix docstring
3 participants
No description provided.