rohdesamuel commented on a change in pull request #14569:
URL: https://github.com/apache/beam/pull/14569#discussion_r623402734
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -1928,22 +1968,31 @@ def pop(self, item):
@frame_base.args_to_kwargs(pd.DataFrame)
@frame_base.populate_defaults(pd.DataFrame)
def quantile(self, q, axis, **kwargs):
- if axis in (1, 'columns') and isinstance(q, list):
- raise frame_base.WontImplementError(
- "quantile(axis=columns) with multiple q values is not supported "
- "because it transposes the input DataFrame. Note computing "
- "an individual quantile across columns (e.g. "
- f"df.quantile(q={q[0]!r}, axis={axis!r}) is supported.",
- reason="non-deferred-columns")
+ if axis in (1, 'columns'):
+ if isinstance(q, list):
+ raise frame_base.WontImplementError(
+ "quantile(axis=columns) with multiple q values is not supported "
+ "because it transposes the input DataFrame. Note computing "
+ "an individual quantile across columns (e.g. "
+ f"df.quantile(q={q[0]!r}, axis={axis!r}) is supported.",
+ reason="non-deferred-columns")
+ else:
+ requires = partitionings.Arbitrary()
+ else: # axis='index'
+ # TODO(BEAM-12167): Provide an option for approximate distributed
+ # quantiles
+ requires = partitionings.Singleton(reason=(
+ "Computing quantiles across index cannot currently be parallelized. "
+ "See BEAM-12167 tracking the possible addition of an approximate, "
+ "parallizable implementation of quantile."
Review comment:
s/parallizable/parallelizable
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
