The problem
An analyzis script becomes hard to extend the moment input and output choices are coupled to transformation logic.
For: Engineers reviewing architecture fundamentals.
What I built
A layered pipeline with core contracts, pluggable CSV and Excel readers, console and graphics writers, and configuration-driven driver selection.
The hard part
Keeping dependencies pointed at core abstractions while leaving the pipeline simple to run.
The tradeoff
Accepted indirection in exchange for boundaries that make a new source a small change.
What this does not do
- A coursework-scale system. The architecture is the point, not the dataset.