NOTE: Spark 3.0 introduced a new pandas UDF. You can find more details in the following blog post: New Pandas UDFs and Python Type Hints in the Upcoming Release of Apache Spark 3.0
This is a guest community post from Li Jin, a software engineer at Two Sigma Investments, LP in New York. This blog is also posted on Two SigmaTry this notebook in Databricks
UPDATE: This blog was updated on Feb 22, 2018, to include some changes.
This blog post introduces the Pandas UDFs (a.k.a. Vectorized UDFs) feature in the upcoming Apache Spark 2.3 release that substantially improves the performance and usability of user-defined functions (UDFs) in Python.
Over the past few years, Python has become the default language for data scientists. Packages such as pandas, numpy, statsmodel, and scikit-learn have gained great adoption and become the mainstream toolkits. At the same time, Apache Spark has become the de facto standard in processing big data. To enable data scientists to leverage the value of big data, Spark added a Python API in version 0.7, with support for user-defined functions. These user-defined functions operate one-row-at-a-time, and thus suffer from high serialization and invocation overhead. As a result, many data pipelines define UDFs in Java and Scala and then invoke them from Python.
Pandas UDFs built on top of Apache Arrow bring you the best of both worlds—the ability to define low-overhead, high-performance UDFs entirely in Python.
In Spark 2.3, there will be two types of Pandas UDFs: scalar and grouped map. Next, we illustrate their usage using four example programs: Plus One, Cumulative Probability, Subtract Mean, Ordinary Least Squares Linear Regression.
Scalar Pandas UDFs
Scalar Pandas UDFs are used for vectorizing scalar operations. To define a scalar Pandas UDF, simply use
@pandas_udf to annotate a Python function that takes in
pandas.Series as arguments and returns another
pandas.Series of the same size. Below we illustrate using two examples: Plus One and Cumulative Probability.
Computing v + 1 is a simple example for demonstrating differences between row-at-a-time UDFs and scalar Pandas UDFs. Note that built-in column operators can perform much faster in this scenario.
Using row-at-a-time UDFs:
from pyspark.sql.functions import udf # Use udf to define a row-at-a-time udf @udf('double') # Input/output are both a single double value def plus_one(v): return v + 1 df.withColumn('v2', plus_one(df.v))
Using Pandas UDFs:
from pyspark.sql.functions import pandas_udf, PandasUDFType # Use pandas_udf to define a Pandas UDF @pandas_udf('double', PandasUDFType.SCALAR) # Input/output are both a pandas.Series of doubles def pandas_plus_one(v): return v + 1 df.withColumn('v2', pandas_plus_one(df.v))
The examples above define a row-at-a-time UDF “plus_one” and a scalar Pandas UDF “pandas_plus_one” that performs the same “plus one” computation. The UDF definitions are the same except the function decorators: “udf” vs “pandas_udf”.
In the row-at-a-time version, the user-defined function takes a double “v” and returns the result of “v + 1” as a double. In the Pandas version, the user-defined function takes a
pandas.Series “v” and returns the result of “v + 1” as a
pandas.Series. Because “v + 1” is vectorized on
pandas.Series, the Pandas version is much faster than the row-at-a-time version.
Note that there are two important requirements when using scalar pandas UDFs:
- The input and output series must have the same size.
- How a column is split into multiple
pandas.Seriesis internal to Spark, and therefore the result of user-defined function must be independent of the splitting.
import pandas as pd from scipy import stats @pandas_udf('double') def cdf(v): return pd.Series(stats.norm.cdf(v)) df.withColumn('cumulative_probability', cdf(df.v))
stats.norm.cdfworks both on a scalar value and
pandas.Series, and this example can be written with the row-at-a-time UDFs as well. Similar to the previous example, the Pandas version runs much faster, as shown later in the “Performance Comparison” section.
Grouped Map Pandas UDFs
Python users are fairly familiar with the split-apply-combine pattern in data analysis. Grouped map Pandas UDFs are designed for this scenario, and they operate on all the data for some group, e.g., “for each date, apply this operation”.
Grouped map Pandas UDFs first splits a Spark
DataFrame into groups based on the conditions specified in the groupby operator, applies a user-defined function (
pandas.DataFrame) to each group, combines and returns the results as a new Spark
Grouped map Pandas UDFs uses the same function decorator
pandas_udf as scalar Pandas UDFs, but they have a few differences:
- Input of the user-defined function:
- Grouped map:
- Output of the user-defined function:
- Grouped map:
- Grouping semantics:
- Scalar: no grouping semantics
- Grouped map: defined by “groupby” clause
- Output size:
- Scalar: same as input size
- Grouped map: any size
- Return types in the function decorator:
- Scalar: a
DataTypethat specifies the type of the returned
- Grouped map: a
StructTypethat specifies each column name and type of the returned
- Scalar: a
Next, let us walk through two examples to illustrate the use cases of grouped map Pandas UDFs.
This example shows a simple use of grouped map Pandas UDFs: subtracting mean from each value in the group.
@pandas_udf(df.schema, PandasUDFType.GROUPED_MAP) # Input/output are both a pandas.DataFrame def subtract_mean(pdf): return pdf.assign(v=pdf.v - pdf.v.mean()) df.groupby('id').apply(subtract_mean)
In this example, we subtract mean of v from each value of v for each group. The grouping semantics is defined by the “groupby” function, i.e, each input
pandas.DataFrame to the user-defined function has the same “id” value. The input and output schema of this user-defined function are the same, so we pass “df.schema” to the decorator
pandas_udf for specifying the schema.
Grouped map Pandas UDFs can also be called as standalone Python functions on the driver. This is very useful for debugging, for example:
sample = df.filter(id == 1).toPandas() # Run as a standalone function on a pandas.DataFrame and verify result subtract_mean.func(sample) # Now run with Spark df.groupby('id').apply(substract_mean)
In the example above, we first convert a small subset of Spark
DataFrame to a
pandas.DataFrame, and then run subtract_mean as a standalone Python function on it. After verifying the function logics, we can call the UDF with Spark over the entire dataset.
Ordinary Least Squares Linear Regression
The last example shows how to run OLS linear regression for each group using statsmodels. For each group, we calculate beta b = (b1, b2) for X = (x1, x2) according to statistical model Y = bX + c.
import statsmodels.api as sm # df has four columns: id, y, x1, x2 group_column = 'id' y_column = 'y' x_columns = ['x1', 'x2'] schema = df.select(group_column, *x_columns).schema @pandas_udf(schema, PandasUDFType.GROUPED_MAP) # Input/output are both a pandas.DataFrame def ols(pdf): group_key = pdf[group_column].iloc y = pdf[y_column] X = pdf[x_columns] X = sm.add_constant(X) model = sm.OLS(y, X).fit() return pd.DataFrame([[group_key] + [model.params[i] for i in x_columns]], columns=[group_column] + x_columns) beta = df.groupby(group_column).apply(ols)
This example demonstrates that grouped map Pandas UDFs can be used with any arbitrary python function:
pandas.DataFrame -> pandas.DataFrame. The returned
pandas.DataFrame can have different number rows and columns as the input.
Lastly, we want to show performance comparison between row-at-a-time UDFs and Pandas UDFs. We ran micro benchmarks for three of the above examples (plus one, cumulative probability and subtract mean).
Configuration and Methodology
We ran the benchmark on a single node Spark cluster on Databricks community edition.
Data: A 10M-row DataFrame with a Int column and a Double column
Cluster: 6.0 GB Memory, 0.88 Cores, 1 DBU
Databricks runtime version: Latest RC (4.0, Scala 2.11)
For the detailed implementation of the benchmark, check the Pandas UDF Notebook.
As shown in the charts, Pandas UDFs perform much better than row-at-a-time UDFs across the board, ranging from 3x to over 100x.
Conclusion and Future Work
The upcoming Spark 2.3 release lays down the foundation for substantially improving the capabilities and performance of user-defined functions in Python. In the future, we plan to introduce support for Pandas UDFs in aggregations and window functions. The related work can be tracked in SPARK-22216.
Pandas UDFs is a great example of the Spark community effort. We would like to thank Bryan Cutler, Hyukjin Kwon, Jeff Reback, Liang-Chi Hsieh, Leif Walsh, Li Jin, Reynold Xin, Takuya Ueshin, Wenchen Fan, Wes McKinney, Xiao Li and many others for their contributions. Finally, special thanks to Apache Arrow community for making this work possible.