ueshin opened a new pull request, #41149:
URL: https://github.com/apache/spark/pull/41149
### What changes were proposed in this pull request?
Supports struct type in `createDataFrame` from pandas DataFrame.
With Arrow optimization, it works without the fallback:
```py
>>> import pandas as pd
>>> from pyspark.sql.types import Row
>>>
>>> pdf = pd.DataFrame(
... {"a": [Row(1, "a"), Row(2, "b")], "b": [{"s": 3, "t": "x"}, {"s": 4,
"t": "y"}]}
... )
>>> schema = "a struct<x int, y string>, b struct<s int, t string>"
>>>
>>> df = spark.createDataFrame(pdf, schema)
>>> df.show()
+------+------+
| a| b|
+------+------+
|{1, a}|{3, x}|
|{2, b}|{4, y}|
+------+------+
```
and Spark Connect also works.
### Why are the changes needed?
In vanilla PySpark without Arrow optimization, `Row` object or `dict` can be
handled as struct type if the schema is provided:
```py
>>> import pandas as pd
>>> from pyspark.sql.types import *
>>>
>>> pdf = pd.DataFrame(
... {"a": [Row(1, "a"), Row(2, "b")], "b": [{"s": 3, "t": "x"}, {"s": 4,
"t": "y"}]}
... )
>>> schema = "a struct<x int, y string>, b struct<s int, t string>"
>>>
>>> df = spark.createDataFrame(pdf, schema)
>>> df.show()
+------+------+
| a| b|
+------+------+
|{1, a}|{3, x}|
|{2, b}|{4, y}|
+------+------+
```
Whereas with Arrow, it uses a fallback to make it:
```py
>>> spark.conf.set('spark.sql.execution.arrow.pyspark.enabled', True)
>>> spark.createDataFrame(pdf, schema).show()
/.../pyspark/sql/pandas/conversion.py:329: UserWarning: createDataFrame
attempted Arrow optimization because
'spark.sql.execution.arrow.pyspark.enabled' is set to true; however, failed by
the reason below:
A field of type StructType expects a pandas.DataFrame, but got: <class
'pandas.core.series.Series'>
Attempting non-optimization as
'spark.sql.execution.arrow.pyspark.fallback.enabled' is set to true.
warn(msg)
+------+------+
| a| b|
+------+------+
|{1, a}|{3, x}|
|{2, b}|{4, y}|
+------+------+
```
and Spark Connect fails:
```py
>>> df = spark.createDataFrame(pdf, schema)
Traceback (most recent call last):
...
ValueError: A field of type StructType expects a pandas.DataFrame, but got:
<class 'pandas.core.series.Series'>
```
### Does this PR introduce _any_ user-facing change?
`Row` object or `dict` in pandas DataFrame works as struct type when
`createDataFrame` if the schema is provided.
### How was this patch tested?
Added the related test.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
