TheNeuralBit commented on a change in pull request #14274:
URL: https://github.com/apache/beam/pull/14274#discussion_r605768355
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -344,6 +344,57 @@ def equals(self, other):
requires_partition_by=partitionings.Singleton(),
preserves_partition_by=partitionings.Singleton()))
+ @frame_base.args_to_kwargs(pd.DataFrame)
+ @frame_base.populate_defaults(pd.DataFrame)
+ @frame_base.maybe_inplace
+ def where(self, cond, other, errors, **kwargs):
+ requires = partitionings.Arbitrary()
+ deferred_args = {}
+ actual_args = {}
+
+ # TODO(bhulette): This is very similar to the logic in
+ # frame_base.elementwise_method, can we unify it?
+ if isinstance(cond, frame_base.DeferredFrame):
+ deferred_args['cond'] = cond
+ requires = partitionings.Index()
+ else:
+ actual_args['cond'] = cond
+
+ if isinstance(other, frame_base.DeferredFrame):
+ deferred_args['other'] = other
+ requires = partitionings.Index()
+ else:
+ actual_args['other'] = other
+
+ if errors == "ignore":
+ # We need all data in order to ignore errors and propagate the original
+ # data.
+ requires = partitionings.Singleton()
Review comment:
I think the intention is that if you have `try_cast=True`, this
parameter will determine what to do if we encounter an element that can't be
casted. It does not seem to affect "construction time" errors, like passing in
a DataFrame with non-matching columns.
Based on the [pandas
docs](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.where.html)
it seems this parameter doesn't actually affect anything right now since "the
results and will always coerce to a suitable dtype", but I figure we may as
well handle it this way to future-proof.
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -344,6 +344,57 @@ def equals(self, other):
requires_partition_by=partitionings.Singleton(),
preserves_partition_by=partitionings.Singleton()))
+ @frame_base.args_to_kwargs(pd.DataFrame)
+ @frame_base.populate_defaults(pd.DataFrame)
+ @frame_base.maybe_inplace
+ def where(self, cond, other, errors, **kwargs):
+ requires = partitionings.Arbitrary()
+ deferred_args = {}
+ actual_args = {}
+
+ # TODO(bhulette): This is very similar to the logic in
+ # frame_base.elementwise_method, can we unify it?
+ if isinstance(cond, frame_base.DeferredFrame):
+ deferred_args['cond'] = cond
+ requires = partitionings.Index()
+ else:
+ actual_args['cond'] = cond
+
+ if isinstance(other, frame_base.DeferredFrame):
+ deferred_args['other'] = other
+ requires = partitionings.Index()
+ else:
+ actual_args['other'] = other
+
+ if errors == "ignore":
+ # We need all data in order to ignore errors and propagate the original
+ # data.
+ requires = partitionings.Singleton()
+
+ actual_args['errors'] = errors
+
+ def where(self, *args):
+ runtime_values = {
+ name: value
+ for (name, value) in zip(deferred_args.keys(), args)
+ }
+ return self.where(**runtime_values, **actual_args, **kwargs)
Review comment:
Yeah your confusion is understandable, I just pushed a commit to clarify
the naming a bit. You're right this is the function that will be used at
execution time, it's called with the outputs from the input expressions:
```
[self._expr] + [df._expr for df in deferred_args.values()],
```
So in this function, the first argument, `self`, will be a pandas DataFrame
(the execution time result of `self._expr`)
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -344,6 +344,57 @@ def equals(self, other):
requires_partition_by=partitionings.Singleton(),
preserves_partition_by=partitionings.Singleton()))
+ @frame_base.args_to_kwargs(pd.DataFrame)
+ @frame_base.populate_defaults(pd.DataFrame)
+ @frame_base.maybe_inplace
+ def where(self, cond, other, errors, **kwargs):
+ requires = partitionings.Arbitrary()
+ deferred_args = {}
+ actual_args = {}
+
+ # TODO(bhulette): This is very similar to the logic in
+ # frame_base.elementwise_method, can we unify it?
+ if isinstance(cond, frame_base.DeferredFrame):
+ deferred_args['cond'] = cond
+ requires = partitionings.Index()
+ else:
+ actual_args['cond'] = cond
+
+ if isinstance(other, frame_base.DeferredFrame):
+ deferred_args['other'] = other
+ requires = partitionings.Index()
+ else:
+ actual_args['other'] = other
+
+ if errors == "ignore":
+ # We need all data in order to ignore errors and propagate the original
+ # data.
+ requires = partitionings.Singleton()
+
+ actual_args['errors'] = errors
+
+ def where(self, *args):
+ runtime_values = {
+ name: value
+ for (name, value) in zip(deferred_args.keys(), args)
+ }
+ return self.where(**runtime_values, **actual_args, **kwargs)
+
+ return frame_base.DeferredFrame.wrap(
+ expressions.ComputedExpression(
+ "where",
+ where,
+ [self._expr] + [df._expr for df in deferred_args.values()],
+ requires_partition_by=requires,
+ preserves_partition_by=partitionings.Index(),
Review comment:
Yes you have that right. It's also possible that an upstream expression
will require `Singleton` partitioning, which could propagate into this one.
I tried to document these semantics with some pictures over in
expressions.py:
https://github.com/apache/beam/blob/7963cd3329f5349cb3ad93e0bbdebebdeeb3b86f/sdks/python/apache_beam/dataframe/expressions.py#L180-L188
If the input is any partitioning that's "less than" the `preserves` spec,
the output will have that partitioning. Otherwise we make no guarantees.
##########
File path: sdks/python/apache_beam/dataframe/frames.py
##########
@@ -344,6 +344,57 @@ def equals(self, other):
requires_partition_by=partitionings.Singleton(),
preserves_partition_by=partitionings.Singleton()))
+ @frame_base.args_to_kwargs(pd.DataFrame)
Review comment:
Good point, I will add these :)
--
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]
