Python / Robot Scripting — Learning Plan

For: Engineer writing daily RCA scripts who wants them testable, typed, and maintainable

Goal: Write analysis scripts that future-you can understand in 6 months

Why This Track Exists

You write Python daily: knowledge_search.py, session_tracker.py, knowledge_builder.py, log analysis scripts. But they lack: - Type annotations (hard to refactor safely) - Tests (regressions silently break things) - Structured data parsing (ad-hoc string manipulation in many places)

Goal: After 3 weeks, every new script you write is typed, has pytest coverage, and uses dataclasses/Pydantic for data boundaries.

Weekly Schedule

Week 1: Type Hints + Dataclasses + Pydantic (2 hrs)

Type annotations: Optional, Union, TypedDict, Protocol
@dataclass vs NamedTuple vs Pydantic BaseModel — when to use each
Validating config files (.env → Pydantic settings)

Exercise: Refactor session_tracker.py session state into a typed Pydantic model

Week 2: Testing CLI Scripts with pytest (2 hrs)

pytest fixtures, tmp_path, monkeypatch
Mocking ADO API calls with pytest-mock
Testing scripts that read from .env

Exercise: Write tests for knowledge_search.py — test the scoring logic without hitting ADO

Week 3: Time-Series Log Analysis with polars (3 hrs)

polars vs pandas: why polars is faster for large bag CSVs
Lazy evaluation: scan_csv().filter().collect()
Time-window aggregations: rolling max, group-by time bin
Plotting with matplotlib: covariance vs time, velocity vs time

Exercise: Load a bag CSV and reproduce the covariance blow-up plot using polars + matplotlib