You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@beam.apache.org by GitBox <gi...@apache.org> on 2020/04/03 17:39:00 UTC
[GitHub] [beam] TheNeuralBit commented on a change in pull request #10760:
[BEAM-9545] Dataframe transforms
TheNeuralBit commented on a change in pull request #10760: [BEAM-9545] Dataframe transforms
URL: https://github.com/apache/beam/pull/10760#discussion_r403112100
##########
File path: sdks/python/apache_beam/dataframe/transforms.py
##########
@@ -0,0 +1,246 @@
+#
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements. See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+
+import pandas as pd
+
+import apache_beam as beam
+from apache_beam import transforms
+from apache_beam.dataframe import expressions
+from apache_beam.dataframe import frame_base
+from apache_beam.dataframe import frames # pylint: disable=unused-import
+
+
+class DataframeTransform(transforms.PTransform):
+ """A PTransform for applying function that takes and returns dataframes
+ to one or more PCollections.
+
+ For example, if pcoll is a PCollection of dataframes, one could write::
+
+ pcoll | DataframeTransform(lambda df: df.group_by('key').sum(), proxy=...)
+
+ To pass multiple PCollections, pass a tuple of PCollections wich will be
+ passed to the callable as positional arguments, or a dictionary of
+ PCollections, in which case they will be passed as keyword arguments.
+ """
+ def __init__(self, func, proxy):
+ self._func = func
+ self._proxy = proxy
+
+ def expand(self, input_pcolls):
+ def wrap_as_dict(values):
+ if isinstance(values, dict):
+ return values
+ elif isinstance(values, tuple):
+ return dict(enumerate(values))
+ else:
+ return {None: values}
+
+ # TODO: Infer the proxy from the input schema.
+ def proxy(key):
+ if key is None:
+ return self._proxy
+ else:
+ return self._proxy[key]
+
+ # The input can be a dictionary, tuple, or plain PCollection.
+ # Wrap as a dict for homogeneity.
+ # TODO: Possibly inject batching here.
+ input_dict = wrap_as_dict(input_pcolls)
+ placeholders = {
+ key: frame_base.DeferredFrame.wrap(
+ expressions.PlaceholderExpression(proxy(key)))
+ for key in input_dict.keys()
+ }
+
+ # The calling convention of the user-supplied func varies according to the
+ # type of the input.
+ if isinstance(input_pcolls, dict):
+ result_frames = self._func(**placeholders)
+ elif isinstance(input_pcolls, tuple):
+ result_frames = self._func(
+ *(value for _, value in sorted(placeholders.items())))
+ else:
+ result_frames = self._func(placeholders[None])
+
+ # Likewise the output may be a dict, tuple, or raw (deferred) Dataframe.
+ result_dict = wrap_as_dict(result_frames)
+
+ result_pcolls = self._apply_deferred_ops(
+ {placeholders[key]._expr: pcoll
+ for key, pcoll in input_dict.items()},
+ {key: df._expr
+ for key, df in result_dict.items()})
+
+ # Convert the result back into a set of PCollections.
+ if isinstance(result_frames, dict):
+ return result_pcolls
+ elif isinstance(result_frames, tuple):
+ return tuple((value for _, value in sorted(result_pcolls.items())))
+ else:
+ return result_pcolls[None]
+
+ def _apply_deferred_ops(
+ self,
+ inputs, # type: Dict[PlaceholderExpr, PCollection]
+ outputs, # type: Dict[Any, Expression]
+ ): # -> Dict[Any, PCollection]
+ """Construct a Beam graph that evaluates a set of expressions on a set of
+ input PCollections.
+
+ :param inputs: A mapping of placeholder expressions to PCollections.
+ :param outputs: A mapping of keys to expressions defined in terms of the
+ placeholders of inputs.
+
+ Returns a dictionary whose keys are those of outputs, and whose values are
+ PCollections corresponding to the values of outputs evaluated at the
+ values of inputs.
+
+ Logically, `_apply_deferred_ops({x: a, y: b}, {f: F(x, y), g: G(x, y)})`
+ returns `{f: F(a, b), g: G(a, b)}`.
+ """
+ class ComputeStage(beam.PTransform):
+ """A helper transform that computes a single stage of operations.
+ """
+ def __init__(self, stage):
+ self.stage = stage
+
+ def default_label(self):
+ return '%s:%s' % (self.stage.ops, id(self))
+
+ def expand(self, pcolls):
+ if self.stage.is_grouping:
+ # Arrange such that partitioned_pcoll is properly partitioned.
+ input_pcolls = {
+ k: pcoll | 'Flat%s' % k >> beam.FlatMap(partition_by_index)
+ for k,
+ pcoll in pcolls.items()
+ }
+ partitioned_pcoll = input_pcolls | beam.CoGroupByKey(
+ ) | beam.MapTuple(
+ lambda _, inputs: {k: pd.concat(vs)
+ for k, vs in inputs.items()})
+ else:
+ # Already partitioned, or no partitioning needed.
+ (k, pcoll), = pcolls.items()
+ partitioned_pcoll = pcoll | beam.Map(lambda df: {k: df})
+
+ # Actually evaluate the expressions.
+ def evaluate(partition, stage=self.stage):
+ session = expressions.Session(
+ {expr: partition[expr._id]
+ for expr in stage.inputs})
+ for expr in stage.outputs:
+ yield beam.pvalue.TaggedOutput(expr._id, expr.evaluate_at(session))
+
+ return partitioned_pcoll | beam.FlatMap(evaluate).with_outputs()
+
+ class Stage(object):
+ """Used to build up a set of operations that can be fused together.
+ """
+ def __init__(self, inputs, is_grouping):
+ self.inputs = set(inputs)
+ self.is_grouping = is_grouping or len(self.inputs) > 1
+ self.ops = []
+ self.outputs = set()
+
+ # First define some helper functions.
+ def output_is_partitioned_by_index(expr, stage):
+ if expr in stage.inputs:
+ return stage.is_grouping
+ elif expr.preserves_partition_by_index():
+ if expr.requires_partition_by_index():
+ return True
+ else:
+ return all(
+ output_is_partitioned_by_index(arg, stage) for arg in expr.args())
+ else:
+ return False
+
+ def partition_by_index(df, levels=None, parts=10):
+ if levels is None:
+ levels = list(range(df.index.nlevels))
+ elif isinstance(levels, (int, str)):
+ levels = [levels]
+ hashes = sum(
+ pd.util.hash_array(df.index.get_level_values(level))
+ for level in levels)
+ for k in range(parts):
+ yield k, df[hashes % parts == k]
+
+ def common_stages(stage_lists):
+ if stage_lists:
+ for stage in stage_lists[0]:
+ if all(stage in other for other in stage_lists[1:]):
+ yield stage
+
+ @memoize
+ def expr_to_stages(expr):
+ assert expr not in inputs
+ # First attempt to compute this expression as part of an existing stage,
+ # if possible.
+ for stage in common_stages([expr_to_stages(arg) for arg in expr.args()
+ if arg not in inputs]):
+ if (not expr.requires_partition_by_index() or
+ all(output_is_partitioned_by_index(arg, stage)
+ for arg in expr.args())):
+ break
Review comment:
A comment clarifying this logic would be helpful. My understanding: "If expr does not require partitioning, just grab any stage, else grab the first stage where all of expr's inputs are partitioned as required. In either case, use the first such stage because earlier stages are closer to the inputs (have fewer intermediate stages)"
----------------------------------------------------------------
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:
users@infra.apache.org
With regards,
Apache Git Services