Skip to content

Latest commit

 

History

History
69 lines (47 loc) · 2.65 KB

workshop.md

File metadata and controls

69 lines (47 loc) · 2.65 KB

🚧🚨 DO NOT READ IF YOU ARE NOT A WORKSHOP INSTRUCTOR OR MENTOR!!! 🚨 🚧

% title: One step closer to unit testing in data science

% authors: Ellen König, Marielle Dado, Tereza Iofciu

-> Unit testing

-> Where to start?

  • Fork and clone the workshop repo: git clone https://github.com/wimlds/berlin-tdd-workshop

  • follow setup instructions in README

  • start environment (README)

  • run first test in terminal: pytest

  • let's look at the function in scripts/imputation.py and at the test in test/test_imputation.py

-> Missing Values - Exercise 1

  • start Jupyter Lab (terminal: jupyter lab)
  • open the EDA notebook
  • the data has missing pricing values
  • let's add different functions for imputing the data and more tests.
  • running, in env run pytest
  • change the imputation method and adapt test (example method is with mean)
    • some should do impute w max
    • some with min
    • think of test edge cases: calling this function on a column full of nans?
    • how will we name all the tests and functions
  • copy the methods we write into scripts/imputation.py and at add the tests in test/test_imputation.py

-> Transformations - Exercise 2

  • checking the data open the EDA notebook
  • method is in scripts/transformation.py and at add the tests in test/test_transformation.py
  • first data transformation method, it works in notebook but has error..
  • comment out the skip test, run test.. (test will fail)
    • fix method so that test passes
    • write a test that explicitly does a type check
    • add test with different input data.. all values the same
  • fix method (should return pd.Series(new_series))

-> EDA - Exercise 3 and Homework

  • Open the EDA notebook
  • let's write the test for the aggregate in the third exercise
    • here we add new python script and new test file
    • discuss the chicken and egg problem of notebooks and scripts..
  • discuss homework .. notify in which channel people should upload their solutions (ideally everybody has forked the repo and would push their solutions to their version of the workshop)

-> Closing