Python rewrite of the Urb/Homemaker stack
genome.py (homemaker-py-k2g): Genome = base-floor GNode tree + per-storey StoreyDelta (undivides, divide subtrees, leaf retypes, height) + base metadata. encode/decode round-trips dom.py Node trees. Key empirical finding baked into the design: upper-storey nodes carry heavily drifted DEAD fields (97 inherited-cut divisions, 187 rotations differ from the owning node below across the corpus) — dead because geometry delegates to below before reading them. decode canonicalises them; encode stores only owned state, so genomes from drifted sources compare equal (fixed-point test). Acceptance: 35/35 corpus files fitness-identical after round-trip through the oracle (experiments/genome_parity.py, URB_NO_OCCLUSION=1); owned-cut projection + genome fixed-point + storey counts in tests/test_genome.py (16 tests pass). Closes homemaker-py-k2g. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|---|---|---|
| .beads | ||
| .claude | ||
| experiments | ||
| src/homemaker | ||
| tests | ||
| .gitignore | ||
| AGENTS.md | ||
| CLAUDE.md | ||
| DESIGN.md | ||
| pyproject.toml | ||
| README.md | ||
homemaker-py
Programme-driven building-layout search over slicing trees. A clean-room Python successor to the Perl Urb project, intended to eventually be 100% Python.
Why a rewrite
Urb represents a building as a binary slicing tree where room sizes are derived top-down from division ratios. That makes room area an emergent property of every cut above it, which:
- gives the genome low locality (a cut near the root rescales every descendant),
- makes target room sizes nearly impossible to hit, so the gaussian size penalty dominates fitness, and
- defeats crossover (transplanted subtrees lose their proportions).
homemaker inverts this: leaves carry target dimensions from the programme and division ratios are solved bottom-up for a fixed topology. The evolutionary search then only explores topology + types + adjacency.
Phase plan
- Solver experiment (current): port Urb's geometry, re-solve ratios from programme targets, score the result against the original via the Perl oracle.
- Native Python fitness (retire the Perl oracle).
- Canonical slicing encoding (normalized Polish expression) + memetic search.
Layout
src/homemaker/dom.py— read/write Urb.domYAML into aNodetree.src/homemaker/geometry.py— faithful port of Urb's top-down geometry.src/homemaker/programme.py— parsepatterns.configspace requirements.src/homemaker/solver.py— bottom-up ratio solve (scipy).src/homemaker/oracle.py— Phase-1 scaffold: score a.domvia Urb'surb-fitness.pl.
The Perl oracle is the only throwaway component; everything else is permanent.