Software design in data-oriented applications benefits from centralizing configuration values in a single, explicit location rather than scattering hard-coded parameters throughout implementation files, and from designing responsibilities around where data lives. This concept belongs to software design principles applied to machine learning/data science codebases, drawing on the information expert principle (assigning responsibility to the component that already holds the relevant data), and on minimizing the amount of data communicated between modules as a proxy for coupling. Type hints are presented as a complementary tool for making data flow and interfaces explicit and less error-prone.
Software design in data-oriented applications benefits from centralizing configuration values in a single, explicit location rather than scattering hard-coded parameters throughout implementation fil…