Github user BryanCutler commented on a diff in the pull request: https://github.com/apache/spark/pull/19459#discussion_r146619813 --- Diff: python/pyspark/sql/session.py --- @@ -414,6 +415,52 @@ def _createFromLocal(self, data, schema): data = [schema.toInternal(row) for row in data] return self._sc.parallelize(data), schema + def _createFromPandasWithArrow(self, pdf, schema): + """ + Create a DataFrame from a given pandas.DataFrame by slicing it into partitions, converting + to Arrow data, then sending to the JVM to parallelize. If a schema is passed in, the + data types will be used to coerce the data in Pandas to Arrow conversion. + """ + from pyspark.serializers import ArrowSerializer, _create_batch + from pyspark.sql.types import from_arrow_schema, to_arrow_type + import pyarrow as pa + + # Slice the DataFrame into batches + step = -(-len(pdf) // self.sparkContext.defaultParallelism) # round int up + pdf_slices = (pdf[start:start + step] for start in xrange(0, len(pdf), step)) + + if schema is None or isinstance(schema, list): + batches = [pa.RecordBatch.from_pandas(pdf_slice, preserve_index=False) + for pdf_slice in pdf_slices] + + # There will be at least 1 batch after slicing the pandas.DataFrame + schema_from_arrow = from_arrow_schema(batches[0].schema) + + # If passed schema as a list of names then rename fields + if isinstance(schema, list): + fields = [] + for i, field in enumerate(schema_from_arrow): + field.name = schema[i] + fields.append(field) + schema = StructType(fields) + else: + schema = schema_from_arrow + else: + if not isinstance(schema, StructType) and isinstance(schema, DataType): + schema = StructType().add("value", schema) --- End diff -- Sorry, I misunderstood. I'm fine with not supporting this case and falling back. Like you pointed out, as this is, it doesn't make much sense to specify a single type for a pd.DataFrame.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org