Designing Functional Data Pipelines for Reproducibility and Maintainability

Designing Functional Data Pipelines for Reproducibility and Maintainability
Event: EuroPython 2021
Date: 29 July 2021
Location: Online

When building data pipelines at scale, it is crucial to design data pipelines that are reliable, scalable and extensible according to evolving business needs. Designing data pipelines for reproducibility and maintainability is a challenge, as testing and debugging across compute units (threads/cores/computes) are often complex and time-consuming due to dependencies and shared states at runtime. In this talk, I will be sharing about common challenges in designing reproducible and maintainable data pipelines at scale, and exploring the use of functional programming in Python to build scalable production-ready data pipelines that are designed for reproducibility and maintainability. Through analogies and realistic examples inspired by data pipeline designs in production environments, you will learn about:

1. What is Functional Programming, and how it differs from other programming paradigms
2. Key Principles of Functional Programming
3. How "control flow" is implemented in Functional Programming
4. Functional design patterns for data pipeline design in Python, and how they improve reproducibility and maintainability
5. Whether it is possible to write a purely functional program in Python

This talk assumes basic understanding of building data pipelines with functions and classes/objects. While the main target audience are data scientists/engineers and developers building data-intensive applications, anyone with hands-on experience in imperative programming (including Python) would be able to understand the key concepts and use-cases in functional programming.

Transcript

1. Designing Functional Data Pipelines
   for Reproducibility and Maintainability
   By: Chin Hwee Ong (@ongchinhwee)
   29 July 2021
   

   View Slide

2. About me
   Ong Chin Hwee 王敬惠
   ● Data Engineer @ DT One
   ● Aerospace Engineering +
   Computational Modelling
   ● Speaker and (occasional) writer on
   data processing
   @ongchinhwee
   Slides link:
   bit.ly/ep2021-design-fp-data
   

   View Slide

3. Basic Design Pattern: Data Pipeline
   @ongchinhwee
   

   View Slide

4. Designing a Data Pipeline at Scale
   ● Reliable
   ○ Data pipeline must produce the desired output → Reproducibility
   ● Scalable
   ○ Data pipeline must run independently across multiple nodes →
   Parallelism
   ● Extensible
   ○ Able to extend data pipeline with changing business logic →
   Maintainability
   @ongchinhwee
   

   View Slide

5. Challenges in Designing Data Pipelines at Scale
   @ongchinhwee
   ● Reproducibility during Testing
   

   View Slide

6. Challenges in Designing Data Pipelines at Scale
   @ongchinhwee
   ● Reproducibility during Testing
   

   View Slide

7. Challenges in Designing Data Pipelines at Scale
   ● Reproducibility during Testing
   ○ Dependencies in data pipeline design
   ■ Data source
   ■ Computation logic
   @ongchinhwee
   

   View Slide

8. Challenges in Designing Data Pipelines at Scale
   ● Reproducibility during Testing
   ○ Challenge: Given the same data source, how do we ensure
   that we replicate the same result every time we re-run the
   same process?
   @ongchinhwee
   

   View Slide

9. Challenges in Designing Data Pipelines at Scale
   @ongchinhwee
   ● Reproducibility in Production
   

   View Slide

10. Challenges in Designing Data Pipelines at Scale
    @ongchinhwee
    ● Reproducibility in Production
    

    View Slide

11. Challenges in Designing Data Pipelines at Scale
    @ongchinhwee
    ● Reproducibility in Production
    

    View Slide

12. Challenges in Designing Data Pipelines at Scale
    @ongchinhwee
    ● Reproducibility in Production
    ○ Debugging parallel/concurrent code at runtime due to
    shared states
    ■ E.g. What is the current state of the data source?
    

    View Slide

13. Challenges in Designing Data Pipelines at Scale
    @ongchinhwee
    ● Reproducibility in Production
    ○ Challenge: How do we design data pipelines that run the
    same computation logic across multiple nodes and
    reproduce predictable results every time?
    

    View Slide

14. Challenges in Designing Data Pipelines at Scale
    ● Maintainability during Debugging
    ○ “Works in testing, breaks in production” 😔
    ■ Edge cases and inefficiencies not detected in test cases
    causing performance issues and/or failures in production
    ■ Complexities in debugging and logging for parallelism
    @ongchinhwee
    

    View Slide

15. Challenges in Designing Data Pipelines at Scale
    ● Maintainability during Debugging
    ○ Challenge: How do we design data pipelines that are
    readable and maintainable at its core to reduce
    inefficiencies in production debugging at scale?
    @ongchinhwee
    

    View Slide

16. Challenges in Designing Data Pipelines at Scale
    ● Maintainability when Adding New Features
    ○ Adding new features to an evolving (growing) codebase
    ■ Code reasoning becomes more challenging with increasing
    code complexity
    ■ Risk of introducing unintended behaviour due to
    dependencies
    @ongchinhwee
    

    View Slide

17. Challenges in Designing Data Pipelines at Scale
    ● Maintainability when Adding New Features
    ○ Challenge: How do we design data pipelines that adapts
    well to changing business and technical requirements and
    ensures developer productivity?
    @ongchinhwee
    

    View Slide

18. Data Pipelines as Functions
    @ongchinhwee
    

    View Slide

19. What is Functional Programming?
    ● Declarative style of programming that emphasizes writing
    software using only:
    ○ Pure functions; and
    ○ Immutable values.
    @ongchinhwee
    

    View Slide

20. 3 Key Principles of Functional Programming
    ● Pure functions and avoid side effects
    ● Immutability
    ● Referential transparency
    @ongchinhwee
    

    View Slide

21. Pure Function and Avoid Side Effects
    @ongchinhwee
    

    View Slide

22. The concept of a “pure function”
    ● Pure function
    ○ Output depends only on its input parameters and its
    internal algorithm
    ○ No side effects
    ⇒ same function f, same input parameter x → same result y
    regardless of number of invocations
    @ongchinhwee
    

    View Slide

23. Pure Function: Making Pizza
    160°C, 10 mins
    P
    U
    T
    T
    H
    E
    M
    T
    O
    G
    ET
    H
    ER
    @ongchinhwee
    

    View Slide

24. “Impure” Function: Making Pizza with Side Effects
    160°C, 10 mins
    P
    U
    T
    T
    H
    E
    M
    T
    O
    G
    ET
    H
    ER
    @ongchinhwee
    Side Effect:
    Radiation Heat
    

    View Slide

25. “Impure” Function: Making Pizza with Side Effects
    180°C, 10 mins
    P
    U
    T
    T
    H
    E
    M
    T
    O
    G
    ET
    H
    ER
    @ongchinhwee
    Side Effect:
    Oven Overheat,
    Burnt Pizza! 😖
    

    View Slide

26. What is a side effect?
    ● A function with side effects changes state outside the
    local function scope
    ○ Examples:
    ■ modifying a variable or data structure in place
    ■ modifying a global state
    ■ performing any I/O operation
    ■ throwing an exception with an error
    @ongchinhwee
    

    View Slide

27. The concept of Immutability
    @ongchinhwee
    

    View Slide

28. The concept of Immutability
    ● Immutability of an assigned variable
    ○ Once a value is assigned to a variable, the state of the
    variable cannot be changed.
    ⇒ Disciplined state management
    ⇒ Prevents side effect resulting from state change → “pure
    function”
    @ongchinhwee
    

    View Slide

29. The concept of Immutability: Key Implication
    ● Key implication: Ease of writing parallel/concurrent programs
    @ongchinhwee
    

    View Slide

30. @ongchinhwee
    

    View Slide

31. The concept of Referential Transparency
    A function is referentially transparent when an expression
    can be substituted by its equivalent algorithm without
    affecting the program logic for all programs
    @ongchinhwee
    

    View Slide

32. Conditions for Referential Transparency
    ● Pure function
    ● Deterministic
    ○ Expression always returns the same output given the same input
    @ongchinhwee
    

    View Slide

33. @ongchinhwee
    

    View Slide

34. Conditions for Referential Transparency
    ● Pure function
    ● Deterministic
    ○ Expression returns the same output given the same input
    ● Idempotent
    ○ Expression can be applied multiple times without changing the
    result beyond its initial application
    @ongchinhwee
    

    View Slide

35. @ongchinhwee
    

    View Slide

36. Equational Reasoning
    ● A key consequent of referential transparency
    ○ Expression can be replaced with its equivalent result
    @ongchinhwee
    

    View Slide

37. Functional Control Flow
    @ongchinhwee
    

    View Slide

38. Function Composition
    @ongchinhwee
    

    View Slide

39. Functions are Values
    ● In Python, functions are first-class objects.
    ● A function can be:
    ○ assigned to a variable
    ○ passed as a parameter to other functions
    ○ returned as a value from other functions
    @ongchinhwee
    

    View Slide

40. Higher-order Functions
    ● Key consequent of first-class functions
    ● A higher-order function has at least one of these
    properties:
    ○ Accepts functions as parameters
    ○ Returns a function as a value
    @ongchinhwee
    

    View Slide

41. Anonymous Functions
    ● Also known as “lambda expressions” in Python
    ● Using function as input without defining named function object
    @ongchinhwee
    

    View Slide

42. map
    @ongchinhwee
    

    View Slide

43. filter
    @ongchinhwee
    

    View Slide

44. reduce
    @ongchinhwee
    

    View Slide

45. map/filter/reduce vs for-loops
    @ongchinhwee
    Managing state changes of
    mutable variables in a for-loop
    

    View Slide

46. Recursion as a form of “functional iteration”
    ● Recursion is a form of self-referential function composition
    ○ Takes the results of itself as inputs into another instance of
    itself
    ○ To prevent infinite recursive loop, base case required as
    terminating condition
    @ongchinhwee
    

    View Slide

47. Recursion as a form of “functional iteration”
    @ongchinhwee
    

    View Slide

48. Recursion as a form of “functional iteration”
    ● Tail-call optimization
    ○ Objective: reduce stack frame consumption in call stack
    ○ Tail call: does nothing other than returning the value of
    function call
    ○ Identify tail calls and compile them to iterative loops
    @ongchinhwee
    

    View Slide

49. Recursion as a form of “functional iteration”
    @ongchinhwee
    

    View Slide

50. Functional Design Patterns for
    Data Pipeline Design
    (in Python)
    @ongchinhwee
    

    View Slide

51. Built-in Higher-order Functions
    ● map/filter vs list comprehensions
    @ongchinhwee
    

    View Slide

52. Built-in Higher-order Functions
    ● map/filter vs list comprehensions
    ○ List comprehensions are syntactic sugar for map/filter
    operations in a data collection (list)
    @ongchinhwee
    

    View Slide

53. Built-in Higher-order Functions
    ● Using map/filter in data transformations
    @ongchinhwee
    

    View Slide

54. Built-in Higher-order Functions
    ● Benefits of using map/filter in data transformations
    ○ Keeping data and transformation logic separate
    ■ Improved code reusability with better transparency
    of transformation logic
    @ongchinhwee
    

    View Slide

55. Extending map/filter to parallel/concurrent programming
    @ongchinhwee
    

    View Slide

56. Extending map/filter to parallel/concurrent programming
    ● Using multiprocessing.Pool or concurrent.futures
    ○ Generate iterator using map, then filter results to a
    collection (list)
    @ongchinhwee
    More details on parallel processing and concurrent.futures: My EuroPython 2020
    talk "Speed Up Your Data Processing"
    

    View Slide

57. Immutable Data Structures
    ● Once an immutable data structure is created, it cannot be
    changed
    ● Benefits:
    ○ Easier to reason - “what you see is what you get”
    ○ Easier to test - worry about the logic, not the state
    ○ Thread-safe - easier for parallelism
    @ongchinhwee
    

    View Slide

58. Tuple vs List
    @ongchinhwee
    

    View Slide

59. Tuple vs List
    @ongchinhwee
    

    View Slide

60. Namedtuple vs Class vs Dictionary
    @ongchinhwee
    

    View Slide

61. Namedtuple vs Class vs Dictionary
    @ongchinhwee
    

    View Slide

62. Namedtuple vs Class vs Dictionary
    @ongchinhwee
    

    View Slide

63. Namedtuple vs Class vs Dictionary
    @ongchinhwee
    

    View Slide

64. Structural Pattern Matching (PEP 634)
    ● Python 3.10 feature inspired by similar syntax with Scala
    ● Especially useful for conditional matching of data structure
    patterns
    match Item:
    case Something:
    do_something()
    @ongchinhwee
    

    View Slide

65. Structural Pattern Matching (PEP 634)
    ● match/case expressions vs if/elif/else
    @ongchinhwee
    

    View Slide

66. Structural Pattern Matching (PEP 634)
    ● Pattern matching for maintainability of data schema
    @ongchinhwee
    Note: Example based on
    case classes and pattern
    matching syntax in Scala
    Dataclasses used as the
    Python equivalent of Scala
    case classes
    

    View Slide

67. Recursions in Python
    ● Tail-call optimization not supported in Python
    ○ Optimization has to be implemented manually
    ● Recursion limit of 1000 (by default) as a prevention
    mechanism against call stack overflow in CPython
    implementation
    @ongchinhwee
    

    View Slide

68. Type Systems
    ● Python has support for type hints (though not enforced in
    runtime)
    @ongchinhwee
    

    View Slide

69. Type Systems
    ● Type checking with mypy
    @ongchinhwee
    

    View Slide

70. Type Systems
    ● Type checking with mypy
    ● Preventing bugs at runtime by ensuring type safety and
    consistency across the data pipeline
    @ongchinhwee
    

    View Slide

71. Can we write a purely functional data
    pipeline in Python?
    @ongchinhwee
    

    View Slide

72. Short Answer: Not really.
    Can we write a purely functional data
    pipeline in Python?
    @ongchinhwee
    

    View Slide

73. “Functional Core, Imperative Shell”
    ● I/O operations still needed for reading
    and writing data outside of the
    application domain
    ● Keeping core domain logic and
    infrastructure code separate
    Ref: Gary Bernhardt's PyCon 2013 talk on "Boundaries"
    @ongchinhwee
    

    View Slide

74. “Functional Core, Imperative Shell”
    @ongchinhwee
    

    View Slide

75. Key Takeaways
    ● Adopt functional design patterns when designing data
    pipelines at scale (parallel and distributed workflows)
    ○ Reproducible
    ○ Scalable
    ○ Maintainable
    ● “Functional Core, Imperative Shell” to manage side effects
    separately from data pipeline logic
    @ongchinhwee
    

    View Slide

76. Reach out to
    me!
    : ongchinhwee
    : @ongchinhwee
    : hweecat
    : https://ongchinhwee.me
    And check out my ongoing
    series on Functional
    Programming at:
    https://ongchinhwee.me/tag/functional
    -programming
    @ongchinhwee
    

    View Slide