Extending Pandas using Apache Arrow and Numba

With the latest release of Pandas the ability to extend it with custom dtypes was introduced. Using Apache Arrow as the in-memory storage and Numba for fast, vectorized computations on these memory regions, it is possible to extend Pandas in pure Python while achieving the same performance of the built-in types. In the talk we implement a native string type as an example.

Transcript

1. 1
   PyData Berlin 2018
   Uwe L. Korn
   Extending Pandas
   using Apache Arrow and Numba
   

   View Slide

2. 2
   PyData Berlin 2018
   Uwe L. Korn
   Extending Pandas
   using Apache Arrow and Numba
   

   View Slide

3. 3
   PyData Berlin 2018
   Uwe L. Korn
   Strings, Strings, please give me Strings!
   

   View Slide

4. 4
   • Senior Data Scientist at Blue Yonder
   (@BlueYonderTech)
   • Apache {Arrow, Parquet} PMC
   • Data Engineer and Architect with heavy
   focus around Pandas
   About me
   xhochy
   mail@uwekorn.com
   

   View Slide

5. 5
   1. Shortcomings of Pandas
   2. ExtensionArrays
   3. Arrow for storage
   4. Numba for compute
   5. All the stuff
   Agenda
   

   View Slide

6. 6
   Pandas Series
   • Payload stored in a numpy.ndarray
   • Index for data alignment
   • Rich analytical API
   • Accessors like .dt or .str
   

   View Slide

7. 7
   Shortcomings
   • Limited to NumPy data types, otherwise object
   • NumPy’s focus is numerical data and tensors
   • Pandas performs well when NumPy performs well
   • Most popular:
   • no native variable-length strings
   • integers are non-nullable
   

   View Slide

8. 8
   What’s the problem?
   

   View Slide

9. 9
   What’s the problem?
   

   View Slide

10. 10
    Why are objects bad?
    Python Data Science Handbook, Jake VanderPlas; O’Reilly Media, Nov 2016
    https://jakevdp.github.io/PythonDataScienceHandbook/02.01-understanding-data-types.html
    

    View Slide

11. 11
    Extending Pandas (0.23+)
    • Two new interfaces:
    • ExtensionDtype
    • What type of scalars?
    • ExtensionArray
    • Implement basic array ops
    • Pandas provides algorithms on top
    

    View Slide

12. 10x !!1
    12
    

    View Slide

13. 13
    Extending Pandas (0.23+)
    • _from_sequence
    • _from_factorized
    • __getitem__
    • __len__
    • dtype
    • nbytes
    • isna
    • copy
    • _concat_same_type
    https://pandas.pydata.org/pandas-docs/stable/generated/pandas.api.extensions.ExtensionArray.html
    13
    

    View Slide

14. 14
    Apache Arrow
    • Specification for in-memory columnar data layout
    • No overhead for cross-system communication
    • Designed for efficiency (exploit SIMD, cache locality, ..)
    • Exchange data without conversion between Python, C++, C(glib),
    Ruby, Lua, R, JavaScript, Go, Rust, Matlab and the JVM
    • Brought Parquet to Pandas and made PySpark fast (@pandas_udf)
    

    View Slide

15. 15
    Nice properties
    • More native datatypes: string, date, nullable int, list of X, …
    • Everything is nullable
    • Memory can be chunked
    • Zero-copy to other ecosystems like Java / R
    • Highly efficient I/O
    

    View Slide

16. 16
    Not so nice properties
    • Still a young project
    • Not much analytic on top (yet!)
    • Core is in modern C++
    • Extremely fast but hard to extend in Python
    

    View Slide

17. 17
    Writing Algorithms in Python is easy!
    but slow
    

    View Slide

18. 18 Photo by Matthew Brodeur on Unsplash
    

    View Slide

19. 19
    Fast for-loops with Numba
    

    View Slide

20. 20
    Anatomy of an Arrow StringArray
    • 3 memory buffers
    • bitmap to indicate valid (non-null) entries
    • uint32 array of offsets: „where does the string start“
    • uint8 array of characters (UTF-8 encoded)
    • int64 offset
    • allows zero-copy slicing
    

    View Slide

21. 21
    Numba @jitclass
    

    View Slide

22. 22
    Numba @jitclass
    

    View Slide

23. 23 Photo by Niklas Tidbury on Unsplash
    

    View Slide

24. 24
    Fletcher
    https://github.com/xhochy/fletcher
    • Implements Extension{Array,Dtype} with Apache Arrow as storage
    • Uses Numba to implement the necessary analytic on top
    

    View Slide

25. Demo
    25
    

    View Slide

26. 26
    Fletcher Demo
    

    View Slide

27. 27
    Fletcher Demo
    

    View Slide

28. 28
    Fletcher Demo
    

    View Slide

29. 29
    Fletcher Demo
    

    View Slide

30. 30
    ExtensionArray Implementations
    https://github.com/ContinuumIO/cyberpandas
    IPArray
    (PR) https://github.com/geopandas/geopandas
    GeometryArray
    (WIP) https://github.com/xhochy/fletcher
    Apache Arrow + Numba backed Arrays
    

    View Slide

31. 31 Photo by Israel Sundseth on Unsplash
    pip install fletcher
    

    View Slide

32. 32
    By JOEXX (Own work) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia Commons
    By JOEXX (Own work) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia Commons
    24. - 26. October
    + 2 days of sprints (27/28.10.)
    ZKM Karlsruhe, DE
    Karlsruhe
    Call for Participation opens next week.
    

    View Slide

33. 33
    I’m Uwe Korn
    Twitter: @xhochy
    https://github.com/xhochy
    Thank you!
    

    View Slide