business-sim-openenv / test_grade.py
ychanakya12
GRANULARITY FIX: extremely granular graders, ASCII-only logs, 3-decimal [END] lines
31f0e47
Raw History Blame Contribute Delete
5.16 kB
"""
Exhaustive edge case test for all 3 graders.
Simulates every possible boundary condition.
"""
import sys
sys.path.insert(0, '.')
# ── Mock objects ──────────────────────────────────────────────────────────────
class Team:
def __init__(self, burnout=0.0, skill=0.5, size=3):
self.burnout = burnout
self.skill = skill
self.size = size
class MockEnv:
def __init__(self, budget, reputation, burnout=0.0, done=False, history=None, quarter=1):
self.budget = budget
self.reputation = reputation
self.team = Team(burnout=burnout)
self.done = done
self.history = history or []
self.quarter = quarter
self.task_id = "test"
# ── Import graders ────────────────────────────────────────────────────────────
from src.tasks import task_easy, task_medium, task_hard
# ── Test cases ────────────────────────────────────────────────────────────────
def clamp_check(name, score):
strict_ok = 0.0 < score < 1.0
status = "OK " if strict_ok else "FAIL *"
print(f" [{status}] {name}: {score}")
return strict_ok
print("\n=== TASK EASY (single_quarter_survival) ===")
tests_easy = [
("budget=0 (bankrupt exact)", MockEnv(0, 0.7)),
("budget=-1 (barely bankrupt)", MockEnv(-1, 0.7)),
("budget=-49999", MockEnv(-49999, 0.7)),
("budget=-50000", MockEnv(-50000, 0.7)),
("budget=-100000 (total loss)", MockEnv(-100000, 0.7)),
("budget=100000 (break-even)", MockEnv(100000, 0.7)),
("budget=100001 (1 dollar up)", MockEnv(100001, 0.7)),
("budget=150000 (mid)", MockEnv(150000, 0.7)),
("budget=150000.0 (exact float)", MockEnv(150000.0,0.7)),
("budget=200000 (target)", MockEnv(200000, 0.7)),
("budget=999999 (massive win)", MockEnv(999999, 0.7)),
]
results_easy = [clamp_check(n, task_easy.grade(e)) for n, e in tests_easy]
print("\n=== TASK MEDIUM (four_quarter_growth) ===")
tests_medium = [
("no history, budget=100000", MockEnv(100000, 0.7, history=[])),
("no history, budget=0", MockEnv(0, 0.0, history=[])),
("no history, budget=200000", MockEnv(200000, 1.0, history=[])),
("history, budget=100000", MockEnv(100000, 0.7, history=["q1"])),
("history, budget=130000", MockEnv(130000, 0.7, history=["q1"])),
("history, rep=0.0", MockEnv(130000, 0.0, history=["q1"])),
("history, rep=1.0, budget=200k", MockEnv(200000, 1.0, history=["q1"])),
("history, budget=0, rep=0.0", MockEnv(0, 0.0, history=["q1"])),
("history, budget=-50000", MockEnv(-50000, 0.0, history=["q1"])),
]
results_medium = [clamp_check(n, task_medium.grade(e)) for n, e in tests_medium]
print("\n=== TASK HARD (adversarial_resilience) ===")
tests_hard = [
("budget=0, rep=0.0, burnout=1.0", MockEnv(0, 0.0, burnout=1.0)),
("budget=0, rep=0.0, burnout=0.0", MockEnv(0, 0.0, burnout=0.0)),
("budget=200000, rep=1.0, burnout=0",MockEnv(200000, 1.0, burnout=0.0)),
("budget=200000, rep=0.5, burnout=0",MockEnv(200000, 0.5, burnout=0.0)),
("budget=-1, rep=0.6, burnout=0.5", MockEnv(-1, 0.6, burnout=0.5)),
("budget=100000, rep=0.7, burnout=0",MockEnv(100000, 0.7, burnout=0.0)),
("budget=100000, rep=0.0, burnout=0",MockEnv(100000, 0.0, burnout=0.0)),
("budget=100000, rep=1.0, burnout=1",MockEnv(100000, 1.0, burnout=1.0)),
("budget=999999, rep=1.0, burnout=0",MockEnv(999999, 1.0, burnout=0.0)),
("budget=-99999, rep=0.0, burnout=1",MockEnv(-99999, 0.0, burnout=1.0)),
]
results_hard = [clamp_check(n, task_hard.grade(e)) for n, e in tests_hard]
# ── Also test the server-level clamp via HTTP ─────────────────────────────────
print("\n=== LIVE /grade ENDPOINT (via HTTP) ===")
import httpx
BASE = "http://localhost:7860"
tasks = ["single_quarter_survival", "four_quarter_growth", "adversarial_resilience"]
results_live = []
for task_id in tasks:
r = httpx.post(f"{BASE}/reset", params={"task_id": task_id})
sid = r.json()["session_id"]
rg = httpx.get(f"{BASE}/grade", params={"session_id": sid})
score = rg.json()["score"]
ok = clamp_check(f"{task_id} (after reset)", score)
results_live.append(ok)
# Summary
all_results = results_easy + results_medium + results_hard + results_live
all_ok = all(all_results)
total = len(all_results)
passed = sum(all_results)
print(f"\n{'='*50}")
print(f" {passed}/{total} edge cases passed")
if all_ok:
print(" ALL EDGE CASES PASS - scores are strictly in (0.0, 1.0)")
else:
print(" *** SOME EDGE CASES FAILED - scores are NOT strictly in (0.0, 1.0) ***")
print(f"{'='*50}")