ENH: Implement array ops for pandas #1100
Conversation
|
@jreback can you review this when you get a chance? |
|
Going to pull out the conftest addition and timezone parameterized fixture refactor into a separate PR. |
cpcloud
force-pushed
the
pandas-array-ops
branch
2 times, most recently
from
05ab434
to
fc12974
Compare
| @@ -740,6 +741,9 @@ class MultiQuantile(Quantile): | |||
| name='interpolation', | |||
| default='linear')] | |||
|
|
|||
| def output_type(self): | |||
There was a problem hiding this comment.
maybe should consolidate all of the helper routines somewhere
There was a problem hiding this comment.
I think most of these are operation specific so I don't know if there's any consolidation possible here.
| @@ -2403,9 +2407,9 @@ class ArraySlice(ValueOp): | |||
| input_type = [ | |||
| rules.array(dt.any), | |||
| rules.integer(name='start'), | |||
| rules.integer(name='stop') | |||
| rules.integer(name='stop', optional=True) | |||
There was a problem hiding this comment.
step isn't supported yet. We could make the translation specific to the backend and have the translation rules raise for SQL backends and have pandas work. I'll make an issue.
There was a problem hiding this comment.
The downside to this approach is that you don't get a failure at expression creation time, instead you get a failure at compile time.
| raise ValueError( | ||
| 'Array value type must be a primitive type ' | ||
| '(e.g., number, string, or timestamp)' | ||
| ) |
There was a problem hiding this comment.
isn't this just .astype()? or is this basically np.array(ibis_expr)?
There was a problem hiding this comment.
No, because we're casting the elements of each list in the Series to a different value type.
cpcloud
force-pushed
the
pandas-array-ops
branch
from
fc12974
to
3bb4e80
Compare
cpcloud
force-pushed
the
pandas-array-ops
branch
from
3bb4e80
to
256ad7a
Compare
|
@jreback Ok to merge on green? |
|
yep lgtm. |
Closes #1099