[
https://issues.apache.org/jira/browse/BEAM-9547?focusedWorklogId=501708&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-501708
]
ASF GitHub Bot logged work on BEAM-9547:
----------------------------------------
Author: ASF GitHub Bot
Created on: 16/Oct/20 23:34
Start Date: 16/Oct/20 23:34
Worklog Time Spent: 10m
Work Description: TheNeuralBit commented on a change in pull request
#13122:
URL: https://github.com/apache/beam/pull/13122#discussion_r506755598
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -32,9 +32,74 @@ def __array__(self, dtype=None):
raise frame_base.WontImplementError(
'Conversion to a non-deferred a numpy array.')
+ get = frame_base.not_implemented_method('get')
+
@frame_base.DeferredFrame._register_for(pd.Series)
class DeferredSeries(DeferredDataFrameOrSeries):
+ def __getitem__(self, key):
+ if _is_null_slice(key) or key is Ellipsis:
+ return self
+
+ elif (isinstance(key, int) or _is_integer_slice(key)
+ ) and self._expr.proxy().index._should_fallback_to_positional():
+ raise frame_base.WontImplementError('order sensitive')
+
+ elif isinstance(key, slice) or callable(key):
+ return frame_base.DeferredFrame.wrap(
+ expressions.ComputedExpression(
+ # yapf: disable
+ 'getitem',
+ lambda df: df[key],
+ [self._expr],
+ requires_partition_by=partitionings.Nothing(),
+ preserves_partition_by=partitionings.Singleton()))
+
+ elif isinstance(key, DeferredSeries):
+ return frame_base.DeferredFrame.wrap(
+ expressions.ComputedExpression(
+ # yapf: disable
+ 'getitem',
+ lambda df, indexer: df[indexer],
+ [self._expr, key._expr],
+ requires_partition_by=partitionings.Index(),
+ preserves_partition_by=partitionings.Singleton()))
+
+ elif pd.core.series.is_iterator(key) or
pd.core.common.is_bool_indexer(key):
+ raise frame_base.WontImplementError('order sensitive')
+
+ else:
+ # We could consider returning a deferred scalar, but that might
+ # be more surprising than a clear error.
+ raise frame_base.WontImplementError('non-deferred')
+
+ if isinstance(key, frame_base.DeferredBase):
+ # Fail early if key is a DeferredBase as it interacts surprisingly with
+ # key in self._expr.proxy().columns
+ raise NotImplementedError(
+ "Indexing with a deferred frame is not yet supported. Consider "
+ "using df.loc[...]")
+
+ if isinstance(key, slice):
+ types = set([type(key.start), type(key.stop), type(key.step)])
+ if types == {type(None)}:
+ # Empty slice is just a copy.
+ return frame_base.DeferredFrame.wrap(self._expr)
+ elif types in [{int}, {type(None), int}]:
Review comment:
You could use `_is_null_slice` and `_is_integer_slice` here for clarity.
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -32,9 +32,74 @@ def __array__(self, dtype=None):
raise frame_base.WontImplementError(
'Conversion to a non-deferred a numpy array.')
+ get = frame_base.not_implemented_method('get')
+
@frame_base.DeferredFrame._register_for(pd.Series)
class DeferredSeries(DeferredDataFrameOrSeries):
+ def __getitem__(self, key):
+ if _is_null_slice(key) or key is Ellipsis:
+ return self
+
+ elif (isinstance(key, int) or _is_integer_slice(key)
+ ) and self._expr.proxy().index._should_fallback_to_positional():
+ raise frame_base.WontImplementError('order sensitive')
+
+ elif isinstance(key, slice) or callable(key):
+ return frame_base.DeferredFrame.wrap(
+ expressions.ComputedExpression(
+ # yapf: disable
+ 'getitem',
+ lambda df: df[key],
+ [self._expr],
+ requires_partition_by=partitionings.Nothing(),
+ preserves_partition_by=partitionings.Singleton()))
+
+ elif isinstance(key, DeferredSeries):
+ return frame_base.DeferredFrame.wrap(
+ expressions.ComputedExpression(
+ # yapf: disable
+ 'getitem',
+ lambda df, indexer: df[indexer],
+ [self._expr, key._expr],
+ requires_partition_by=partitionings.Index(),
+ preserves_partition_by=partitionings.Singleton()))
+
+ elif pd.core.series.is_iterator(key) or
pd.core.common.is_bool_indexer(key):
+ raise frame_base.WontImplementError('order sensitive')
+
+ else:
+ # We could consider returning a deferred scalar, but that might
+ # be more surprising than a clear error.
+ raise frame_base.WontImplementError('non-deferred')
+
+ if isinstance(key, frame_base.DeferredBase):
+ # Fail early if key is a DeferredBase as it interacts surprisingly with
+ # key in self._expr.proxy().columns
+ raise NotImplementedError(
+ "Indexing with a deferred frame is not yet supported. Consider "
+ "using df.loc[...]")
+
+ if isinstance(key, slice):
+ types = set([type(key.start), type(key.stop), type(key.step)])
+ if types == {type(None)}:
+ # Empty slice is just a copy.
+ return frame_base.DeferredFrame.wrap(self._expr)
+ elif types in [{int}, {type(None), int}]:
+ # This depends on the contents of the index.
+ raise frame_base.WontImplementError(
+ 'Use iloc or loc with integer slices.')
Review comment:
Doesn't `iloc` only work for a null slice? I think we should just direct
users to `loc` iff the proxy has an integer index. If it has a non-integer
index we could tell the user they're going to have a bad time (maybe eventually
it would link to some documentation about the dangers of integer-location based
indexing in DataframeTransform).
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -453,17 +518,31 @@ def __getattr__(self, name):
def __getitem__(self, key):
# TODO: Replicate pd.DataFrame.__getitem__ logic
- if isinstance(key, frame_base.DeferredBase):
+ if isinstance(key, DeferredSeries) and key._expr.proxy().dtype == bool:
Review comment:
Should we have this logic in `DeferredSeries.__getitem__` too?
----------------------------------------------------------------
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]
Issue Time Tracking
-------------------
Worklog Id: (was: 501708)
Time Spent: 26h 40m (was: 26.5h)
> Implement all pandas operations (or raise WontImplementError)
> -------------------------------------------------------------
>
> Key: BEAM-9547
> URL: https://issues.apache.org/jira/browse/BEAM-9547
> Project: Beam
> Issue Type: Improvement
> Components: sdk-py-core
> Reporter: Brian Hulette
> Assignee: Robert Bradshaw
> Priority: P2
> Time Spent: 26h 40m
> Remaining Estimate: 0h
>
> We should have an implementation for every DataFrame, Series, and GroupBy
> method. Everything that's not actually implemented should get a default
> implementation that raises WontImplementError
> See https://github.com/apache/beam/pull/10757#discussion_r389132292
> Progress at the individual operation level is tracked in a
> [spreadsheet|https://docs.google.com/spreadsheets/d/1hHAaJ0n0k2tw465ORs5tfdy4Lg0DnGWIQ53cLjAhel0/edit],
> consider requesting edit access if you'd like to help out.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
