a36956b1e1
The truth files used `determinableFromResponseDocument`; the schema column
and gateEvaluableFromResponse() use `determinableFromResponse`. A silent
name drift means the flag reads as its default (true), and R-B6C — "two bid
copies submitted", which no response document can ever establish — starts
failing bidders on a packaging detail.
Renamed in rfp.truth.json and validate-responses.py, and added the mirror
check to validate-truth.py so the drift cannot recur unnoticed:
gates 13, evaluable from the response 12, excluded 1
excluded: R-B6C Two bid copies submitted.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
104 lines
4.8 KiB
Python
104 lines
4.8 KiB
Python
#!/usr/bin/env python3
|
|
"""Validate a corpus truth file's internal consistency.
|
|
|
|
Run: python3 corpus/validate-truth.py corpus/<dir>/rfp.truth.json
|
|
These files are hand-authored ground truth, so nothing else checks them.
|
|
"""
|
|
import json, sys
|
|
|
|
path = sys.argv[1] if len(sys.argv) > 1 else 'friendship-pcs/rfp.truth.json'
|
|
t = json.load(open(path))
|
|
errs, warns = [], []
|
|
|
|
crit = {c['id']: c for c in t['criteria']}
|
|
reqs = {r['id']: r for r in t['requirements']}
|
|
kms = {k['id']: k for k in t['knownMisses']}
|
|
nest = t['criteriaNotes']['nestingIsInconsistent']
|
|
|
|
for r in t['requirements']:
|
|
for cid in r.get('criterionIds', []):
|
|
if cid not in crit: errs.append(f"{r['id']} -> unknown criterion {cid}")
|
|
if 'knownMissId' in r and r['knownMissId'] not in kms:
|
|
errs.append(f"{r['id']} -> unknown knownMiss {r['knownMissId']}")
|
|
|
|
for k in t['knownMisses']:
|
|
if k['requirementId'] not in reqs:
|
|
errs.append(f"{k['id']} -> unknown requirement {k['requirementId']}")
|
|
|
|
for c in t['criteria']:
|
|
if c['parentId'] and c['parentId'] not in crit:
|
|
errs.append(f"{c['id']} -> unknown parent {c['parentId']}")
|
|
|
|
for pid in nest['weightedParents']:
|
|
kids = [c for c in t['criteria'] if c['parentId'] == pid]
|
|
s = sum(c['maxPoints'] for c in kids)
|
|
if s != crit[pid]['maxPoints']:
|
|
errs.append(f"{pid}: children sum {s} != parent {crit[pid]['maxPoints']}")
|
|
else:
|
|
print(f" ok {pid:14} weighted children sum {s} == parent")
|
|
|
|
for pid in nest['guidanceOnlyParents']:
|
|
kids = [c for c in t['criteria'] if c['parentId'] == pid]
|
|
if kids: errs.append(f"{pid} declared guidance-only but has {len(kids)} child criteria")
|
|
else: print(f" ok {pid:14} guidance-only, no child criteria")
|
|
|
|
top = sum(c['maxPoints'] for c in t['criteria'] if c['parentId'] is None)
|
|
tot = t['criteriaNotes']['totalPoints']
|
|
print(f" {'ok ' if top == tot else 'ERR '} top-level points sum {top} vs declared {tot}")
|
|
if top != tot: errs.append(f"top-level sum {top} != {tot}")
|
|
|
|
pts = sorted(a['points'] for a in t['ratingScale']['anchors'])
|
|
if pts != list(range(t['ratingScale']['min'], t['ratingScale']['max'] + 1)):
|
|
errs.append(f"rating scale not contiguous: {pts}")
|
|
else:
|
|
print(f" ok rating scale {pts[0]}-{pts[-1]}, {len(pts)} anchors")
|
|
|
|
# The scoring-level rule, mirrored from src/scoring/routing.ts: a score commits
|
|
# at any criterion with NO WEIGHTED CHILDREN. Printed so the corpus and the code
|
|
# cannot silently disagree about which criteria a committee actually scores.
|
|
scoreable = [c for c in t['criteria']
|
|
if not any(k['parentId'] == c['id'] and (k.get('maxPoints') or 0) > 0
|
|
for k in t['criteria'])]
|
|
sc_total = sum(c['maxPoints'] for c in scoreable)
|
|
print(f" ok scoreable criteria {len(scoreable)} of {len(t['criteria'])}, "
|
|
f"points sum {sc_total}")
|
|
if sc_total != tot:
|
|
errs.append(f"scoreable points sum {sc_total} != declared total {tot}")
|
|
print(" " + ", ".join(f"{c['id']}({c['maxPoints']})" for c in scoreable))
|
|
|
|
mapped = {c for r in t['requirements'] for c in r.get('criterionIds', [])}
|
|
for cid, c in crit.items():
|
|
kids = [x for x in t['criteria'] if x['parentId'] == cid]
|
|
if cid not in mapped and not kids:
|
|
warns.append(f"criterion {cid} ({c['label'][:38]}) has no requirement mapped — unscoreable")
|
|
|
|
sec = {s['mapsToCriterion'] for s in t['bindingMechanism']['requiredSections']} - {None}
|
|
for cid in sec:
|
|
if cid not in crit: errs.append(f"bindingMechanism section -> unknown criterion {cid}")
|
|
|
|
gates = [r for r in t['requirements'] if r['kind'] == 'mandatory_gate']
|
|
|
|
# Mirrors gateEvaluableFromResponse() in src/scoring/routing.ts. The key read
|
|
# here is the SCHEMA COLUMN NAME on purpose — the two names drifted once, and a
|
|
# silent drift means the flag reads as its default and a bidder is disqualified
|
|
# over something no document could have proven.
|
|
evaluable = [r for r in gates if r.get('determinableFromResponse', True)]
|
|
excluded = [r for r in gates if not r.get('determinableFromResponse', True)]
|
|
print(f" ok gates {len(gates)}, evaluable from the response {len(evaluable)}, "
|
|
f"excluded {len(excluded)}")
|
|
for r in excluded:
|
|
print(f" excluded: {r['id']:14} {r['text'][:50]}")
|
|
for r in t['requirements']:
|
|
if r.get('determinableFromResponse') is False and r['kind'] != 'mandatory_gate':
|
|
warns.append(f"{r['id']} is not determinable but is not a gate — the flag "
|
|
f"only changes gate evaluation, so this has no effect")
|
|
|
|
kinds = {}
|
|
for r in t['requirements']: kinds[r['kind']] = kinds.get(r['kind'], 0) + 1
|
|
print(f"\n requirements {len(t['requirements'])} criteria {len(t['criteria'])} knownMisses {len(t['knownMisses'])}")
|
|
print(f" by kind: " + ", ".join(f"{k}={v}" for k, v in sorted(kinds.items())))
|
|
for w in warns: print(f" WARN {w}")
|
|
for e in errs: print(f" ERR {e}")
|
|
print("\nVALID" if not errs else "\nINVALID")
|
|
sys.exit(1 if errs else 0)
|