A personal practice repo deliberately structured as a small software project rather than a dump of one-off files.
Problem
Most algorithm-practice repos are append-only directories with no shared scaffolding, no tests, and no traceable mapping from problem to pattern. That makes them useless for revisiting, refactoring, or measuring coverage against curated study paths.
What I built
A Python repo with one directory per problem, a generated catalog driven by a single metadata.json source-of-truth, written pattern notes, and overlays that mark coverage against curated paths (Blind 75, NeetCode 150).
Technical components
- Language
- Python
- Tests
- pytest harness exercising public LeetCode signatures
- Structure
- Per-problem directory, per-pattern notes under
0000-notes/, curated paths underpaths/ - Generation
scripts/update_indexes.pyregeneratesCATALOG.md+TOPICS.md+ featured tables frommetadata.json- CI
- GitHub Actions "quality" workflow on every push
Evidence / outputs
Problems
99
reference solutions
Difficulty
28 / 52 / 19
easy / med / hard
Blind 75
10 / 75
tagged
Reviewed
13
AI-card showcases
Generated catalog and topic index in-repo, derived from the metadata.json source-of-truth.
Current status
Maintained. Expanded incrementally rather than in bursts.
Limitations
- Not a complete coverage of either curated list yet.
- Complexity analysis is prose only — no formal proofs.