Overload df.getitem with bool array idx #587
Conversation
densmirn
requested review from
AlexanderKalistratov,
akharche and
kozlov-alexey
densmirn
force-pushed
the
feature/df_getitem_bool_arr
branch
4 times, most recently
from
9a323a8 to
200c54a
Compare
densmirn
force-pushed
the
feature/df_getitem_bool_arr
branch
from
200c54a to
601411e
Compare
|
@kozlov-alexey could you please review it? |
| return gen_df_getitem_bool_series_idx_impl(self, idx) | ||
|
|
||
| ty_checker.raise_exc(idx, 'str', 'idx') | ||
| if isinstance(idx, types.Array) and isinstance(idx.dtype, types.Boolean): |
There was a problem hiding this comment.
I would propose to move it at the top, just after we checked that self is of DataFrameType, because this check should really intercept wrong types (for which we do not want compilation to happen). Otherwise we might compile (or at least try to compile) some specialization we do not want to compile (e.g. if some of the arg types can be type casted to another one).
|
|
||
|
|
||
| def df_index_codelines(self): | ||
| def df_index_codelines(self, with_length=False): |
There was a problem hiding this comment.
I think it's better for this function to do only index lines generation, so that we just make df_length_codelines call not inside it, but where this func is called with with_lenght=True. This simplifies things, as on the call site you see exactly, that you generate line for lenght, then line for index.
Alternatively we could use a format string and pass the actual var name that will hold df's len to it as a second argument, i.e.
def df_index_codelines(self, length):
...
func_lines += [f' res_index = numpy.arange({length})']
3 participants
No description provided.