2026-06-23 22:47:34 +01:00
|
|
|
#!/usr/bin/env python3
|
|
|
|
|
"""Diagnostic B (homemaker-py-erc.2, DESIGN.md §13.2): undersize-despite-slack
|
|
|
|
|
localization — construction-target vs inner-loop-fill.
|
|
|
|
|
|
|
|
|
|
GATES the plot-fill-construction (`erc.4`) vs inner-loop-slack-expansion (`erc.6`)
|
|
|
|
|
decision. The paradox from §12.3: plot utilisation is ~0.44 (over half the plot
|
|
|
|
|
empty) yet size fails are large (rooms UNDERSIZE). Where is the slack stranded,
|
|
|
|
|
and at which stage should it be spent?
|
|
|
|
|
|
|
|
|
|
Reads, does not change behaviour. For each programme x seed it builds the §12.2
|
|
|
|
|
constructive seed (adjacency- and proportion-aware) — whose geometry already sits
|
|
|
|
|
at the proportion-aware TARGET ratios (`_size_divisions_from_targets`, the inner
|
|
|
|
|
loop's warm-start), so this IS the "before inner loop" state — then runs
|
|
|
|
|
`innerloop.optimise` to get the "after inner loop" state, and measures at each:
|
|
|
|
|
|
|
|
|
|
1. Per sized-room leaf: achieved area vs target area (get_space_params size),
|
|
|
|
|
classified undersize / at-target / oversize, plus authoritative size-fail
|
|
|
|
|
count from `score_with_fails`.
|
|
|
|
|
2. Plot accounting: total plot area split into sized-room / circulation /
|
|
|
|
|
outside, and the room-target sum vs plot (could rooms even fill the plot?).
|
|
|
|
|
3. Whether the INNER LOOP moves any of it: size fails before vs after, util
|
|
|
|
|
before vs after, oversize leaves before vs after.
|
|
|
|
|
|
|
|
|
|
DECISION RULE: if rooms are parked at/under target with the slack sitting as
|
|
|
|
|
unused plot (rooms can't fill it / no oversize to rebalance) -> fix in
|
|
|
|
|
CONSTRUCTION (plot-fill, `erc.4`). If the inner loop has room to expand
|
|
|
|
|
(oversize coexists with undersize, slack is inside the sized leaves) but does not
|
|
|
|
|
spend it -> no objective gradient -> fix in the INNER LOOP (slack-expansion term,
|
|
|
|
|
`erc.6`).
|
|
|
|
|
|
|
|
|
|
Usage:
|
|
|
|
|
URB_NO_OCCLUSION=1 python3 experiments/diag_slack_localization.py
|
|
|
|
|
"""
|
|
|
|
|
|
|
|
|
|
from __future__ import annotations
|
|
|
|
|
|
|
|
|
|
import copy
|
|
|
|
|
import sys
|
|
|
|
|
from pathlib import Path
|
|
|
|
|
|
|
|
|
|
import numpy as np
|
|
|
|
|
|
|
|
|
|
sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
|
|
|
|
|
from homemaker_layout import ( # noqa: E402
|
|
|
|
|
dom, fitness, geometry, innerloop, operators, programme,
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
PROGRAMMES = ["harbor-house", "maple-court"]
|
|
|
|
|
SEEDS = (0, 1, 2)
|
|
|
|
|
BUDGET = 80 # constructed bootstrap seeds get child_budget (driver default 80)
|
|
|
|
|
AT_TARGET_TOL = 0.10 # |achieved/target - 1| <= tol counts as "at target"
|
|
|
|
|
ROOT = Path(__file__).resolve().parents[1]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _is_sized(leaf, reqs) -> bool:
|
|
|
|
|
return leaf.type in reqs and reqs[leaf.type].size > 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _t0(leaf) -> str:
|
|
|
|
|
return leaf.type[0].lower() if leaf.type else ""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _measure(tree, fit, reqs):
|
|
|
|
|
"""Per-leaf achieved-vs-target + plot accounting for one laid-out tree."""
|
|
|
|
|
geometry.clear_cache()
|
|
|
|
|
_score, fails = fit.score_with_fails(copy.deepcopy(tree))
|
|
|
|
|
size_fails = sum(1 for f in fails if f.endswith(" size"))
|
|
|
|
|
|
|
|
|
|
leaves = [lf for lvl in dom.levels(tree) for lf in lvl.leaves()]
|
|
|
|
|
geometry.clear_cache()
|
|
|
|
|
plot = sum(geometry.area(lvl) for lvl in dom.levels(tree))
|
|
|
|
|
|
|
|
|
|
sized_area = circ_area = out_area = 0.0
|
|
|
|
|
target_sum = 0.0
|
|
|
|
|
under = at = over = 0
|
|
|
|
|
ratios = []
|
|
|
|
|
for lf in leaves:
|
|
|
|
|
a = geometry.area(lf)
|
|
|
|
|
t0 = _t0(lf)
|
|
|
|
|
if _is_sized(lf, reqs):
|
|
|
|
|
target = fit.get_space_params(lf.type, "size")[0]
|
|
|
|
|
sized_area += a
|
|
|
|
|
target_sum += target
|
|
|
|
|
r = a / target if target else float("nan")
|
|
|
|
|
ratios.append(r)
|
|
|
|
|
if r < 1.0 - AT_TARGET_TOL:
|
|
|
|
|
under += 1
|
|
|
|
|
elif r > 1.0 + AT_TARGET_TOL:
|
|
|
|
|
over += 1
|
|
|
|
|
else:
|
|
|
|
|
at += 1
|
|
|
|
|
elif t0 == "c":
|
|
|
|
|
circ_area += a
|
|
|
|
|
else:
|
|
|
|
|
out_area += a
|
|
|
|
|
|
|
|
|
|
n_sized = len(ratios)
|
|
|
|
|
return {
|
|
|
|
|
"size_fails": size_fails,
|
|
|
|
|
"plot": plot,
|
|
|
|
|
"sized_area": sized_area,
|
|
|
|
|
"circ_area": circ_area,
|
|
|
|
|
"out_area": out_area,
|
|
|
|
|
"target_sum": target_sum,
|
|
|
|
|
"n_sized": n_sized,
|
|
|
|
|
"util": sized_area / plot if plot else float("nan"),
|
|
|
|
|
"target_fill": target_sum / plot if plot else float("nan"),
|
|
|
|
|
"mean_ratio": float(np.mean(ratios)) if ratios else float("nan"),
|
|
|
|
|
"under": under, "at": at, "over": over,
|
|
|
|
|
"pct_under": 100.0 * under / n_sized if n_sized else float("nan"),
|
|
|
|
|
"pct_over": 100.0 * over / n_sized if n_sized else float("nan"),
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _leaf_ratios(tree, fit, reqs):
|
|
|
|
|
"""(a/t, type, target, achieved) for every sized-room leaf."""
|
|
|
|
|
geometry.clear_cache()
|
|
|
|
|
out = []
|
|
|
|
|
for lvl in dom.levels(tree):
|
|
|
|
|
for lf in lvl.leaves():
|
|
|
|
|
if _is_sized(lf, reqs):
|
|
|
|
|
t = fit.get_space_params(lf.type, "size")[0]
|
|
|
|
|
a = geometry.area(lf)
|
|
|
|
|
out.append((a / t if t else float("nan"), lf.type, t, a))
|
|
|
|
|
return out
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _run_seed(pdir, fit, reqs, types, seed_root, s):
|
|
|
|
|
rng = np.random.default_rng(s)
|
|
|
|
|
topo = operators.constructive_topology(
|
|
|
|
|
seed_root, reqs, rng, types,
|
|
|
|
|
adjacency_aware=True, proportion_aware=True, circ_divisor=3)
|
|
|
|
|
before = _measure(topo, fit, reqs)
|
|
|
|
|
|
|
|
|
|
after_tree = copy.deepcopy(topo)
|
|
|
|
|
r = innerloop.optimise(after_tree, str(pdir), x0=None, budget=BUDGET,
|
Remove the Perl oracle
Owner's decision: "we need to abandon the perl oracle, this was only useful
when initially porting, but I suspect many of the remaining problems have been
carried in from the perl (such as the weird scoring of outdoor and circulation
space, which definitely needs fixing)".
39 supports that second clause. Every defect the section found is inherited,
not introduced: the two-sided crinkliness gaussian that double-charges surplus
daylight (39.14), quality as a product over a variable number of factors
(39.18), value_supported priced as value_inside so a terrace was worth more per
m2 than a room (39.19), and circulation returning 0.07 per unit cost (hxi).
So parity with the oracle was never a safety net -- it was a commitment to
reproduce those defects. Each of 39.14, 39.18 and 39.19 would have been a
parity failure had parity ever been checked, and keeping the tests would have
meant reverting the fixes or explaining the failures away.
Removed: oracle.py, test_oracle.py, the two parity tests and their fixture
machinery in test_dom_corpus.py, innerloop.OracleEvaluator with its use_native
and urb_root plumbing, the same plumbing through driver, and fourteen
experiments/ scripts that could only run against Perl. Several of those are
cited in earlier DESIGN sections; the citations now point into git history,
which is the honest state -- they had been unrunnable since the oracle root
(/home/bruno/src/urb) stopped being present. run_search is superseded by
run_search_scaled, which does the same job natively.
Kept: dump_areas.pl/.py, which validate GEOMETRY against Urb (4.1) rather than
fitness, and the prose in fitness_cmd.py and dom.py explaining why the
.score/.fails formats are shaped as they are. Provenance is worth keeping; a
dead code path is not.
CLAUDE.md updated: fitness.py is the only evaluator, and "Urb did it this way"
is no longer an argument that a constant is right. 39.16 is the standing
counterweight in the other direction -- the crinkliness target WAS right and
twice looked wrong only because the code reading it was misunderstood.
Inheritance is neither evidence for nor against.
410 passed. The 69 removed cases account exactly: 64 parity (all skipped, since
no oracle .score was ever committed), 4 in test_oracle.py, and the guard test
39.20 added as a stopgap.
Closes homemaker-py-118. Files homemaker-py-bk9 for the re-baseline that 39.19
made necessary.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MJ84Feep79Hhm3E4zZJmnB
2026-09-06 07:56:24 +00:00
|
|
|
method="nm")
|
2026-06-23 22:47:34 +01:00
|
|
|
after = _measure(after_tree, fit, reqs)
|
|
|
|
|
after["n_evals"] = r.n_evals
|
|
|
|
|
return before, after, _leaf_ratios(topo, fit, reqs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _avg(rows, key):
|
|
|
|
|
vals = [r[key] for r in rows if not (isinstance(r[key], float) and r[key] != r[key])]
|
|
|
|
|
return sum(vals) / len(vals) if vals else float("nan")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def main() -> int:
|
|
|
|
|
print("Diagnostic B — undersize-despite-slack localization (§13.2)\n")
|
|
|
|
|
print("BEFORE = constructive seed at proportion-aware TARGET ratios "
|
|
|
|
|
"(inner-loop warm start)")
|
|
|
|
|
print(f"AFTER = after innerloop.optimise (nm, budget={BUDGET})")
|
|
|
|
|
print(f"Seeds: {SEEDS} at-target tol: +/-{AT_TARGET_TOL:.0%}\n")
|
|
|
|
|
|
|
|
|
|
for name in PROGRAMMES:
|
|
|
|
|
pdir = ROOT / "examples" / name
|
|
|
|
|
reqs = programme.load_programme_dir(pdir)
|
|
|
|
|
types = sorted(reqs) + ["C", "O"]
|
|
|
|
|
conf, cost = fitness.load_config(pdir)
|
|
|
|
|
fit = fitness.Fitness(conf, cost)
|
|
|
|
|
seed_root = dom.load(str(pdir / "init.dom"))
|
|
|
|
|
|
|
|
|
|
befores, afters = [], []
|
|
|
|
|
detail0 = None
|
|
|
|
|
for s in SEEDS:
|
|
|
|
|
b, a, leaf_rows = _run_seed(pdir, fit, reqs, types, seed_root, s)
|
|
|
|
|
befores.append(b)
|
|
|
|
|
afters.append(a)
|
|
|
|
|
if detail0 is None:
|
|
|
|
|
detail0 = leaf_rows
|
|
|
|
|
|
|
|
|
|
n_sized = _avg(befores, "n_sized")
|
|
|
|
|
print(f"=== {name} (sized rooms/seed: {n_sized:.0f}) ===")
|
|
|
|
|
print(f"{'':16}{'sizeF':>7}{'util':>7}{'tgtFill':>8}{'a/t':>6}"
|
|
|
|
|
f"{'%und':>6}{'%ovr':>6}{'sized%':>7}{'circ%':>7}{'out%':>7}")
|
|
|
|
|
for label, rows in (("BEFORE (target)", befores), ("AFTER (innerloop)", afters)):
|
|
|
|
|
plot = _avg(rows, "plot")
|
|
|
|
|
print(f"{label:16}"
|
|
|
|
|
f"{_avg(rows,'size_fails'):>7.1f}"
|
|
|
|
|
f"{_avg(rows,'util'):>7.2f}"
|
|
|
|
|
f"{_avg(rows,'target_fill'):>8.2f}"
|
|
|
|
|
f"{_avg(rows,'mean_ratio'):>6.2f}"
|
|
|
|
|
f"{_avg(rows,'pct_under'):>6.0f}"
|
|
|
|
|
f"{_avg(rows,'pct_over'):>6.0f}"
|
|
|
|
|
f"{100*_avg(rows,'sized_area')/plot:>7.0f}"
|
|
|
|
|
f"{100*_avg(rows,'circ_area')/plot:>7.0f}"
|
|
|
|
|
f"{100*_avg(rows,'out_area')/plot:>7.0f}")
|
|
|
|
|
print(f" plot area ~ {_avg(befores,'plot'):.0f} m2; "
|
|
|
|
|
f"room-target sum ~ {_avg(befores,'target_sum'):.0f} m2 "
|
|
|
|
|
f"(tgtFill = target sum / plot)")
|
|
|
|
|
print(f" inner-loop evals/seed ~ {_avg(afters,'n_evals'):.0f}")
|
|
|
|
|
|
|
|
|
|
# Same-target maldistribution evidence (seed 0): a leaf's area is set by
|
|
|
|
|
# its SLICING POSITION, not its target. The same room TYPE/target lands
|
|
|
|
|
# at both extremes -> depth-driven, not unclaimed plot, not tiny-target.
|
|
|
|
|
detail0.sort(reverse=True)
|
|
|
|
|
print(" seed-0 extremes (type / target / achieved / a-over-t):")
|
|
|
|
|
for r, ty, t, a in detail0[:3]:
|
|
|
|
|
print(f" OVER {ty:5} t={t:5.1f} a={a:6.1f} a/t={r:5.2f}")
|
|
|
|
|
for r, ty, t, a in detail0[-3:]:
|
|
|
|
|
print(f" UNDER {ty:5} t={t:5.1f} a={a:6.1f} a/t={r:5.2f}")
|
|
|
|
|
print()
|
|
|
|
|
|
|
|
|
|
print("READ: util = sized-room area / plot. tgtFill = sum of room targets / plot:")
|
|
|
|
|
print(" if tgtFill << 1, rooms cannot fill the plot even at target -> slack is")
|
|
|
|
|
print(" structurally unassigned (CONSTRUCTION / erc.4). %ovr>0 alongside %und")
|
|
|
|
|
print(" with AFTER unchanged -> slack is inside leaves but inner loop won't spend")
|
|
|
|
|
print(" it (no gradient -> INNER LOOP / erc.6).")
|
|
|
|
|
return 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
|
sys.exit(main())
|