Datasets:
File size: 5,162 Bytes
a42d019 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 | #!/usr/bin/env python3
"""Semantic edge and mutation checks; no network or model calls."""
import copy
from decimal import Decimal
import json
import pathlib
import unittest
from verify import evaluate, normalized_decimal
from baselines import symbolic
from model_input import model_input
from score import value_equal, score_records, reject_constant
def cell(value, year="2000", **updates):
row = dict(cell_id="c" + year, value=value, year=year, unit_code="persons", measure_kind="quantity",
economy_code="X", indicator_code="SP.POP.TOTL")
row.update(updates)
return row
class ContractTests(unittest.TestCase):
def test_decimal_half_up_and_zero(self):
self.assertEqual(normalized_decimal("1.2345675", True), "1.234568")
self.assertEqual(normalized_decimal("-1.2345675", True), "-1.234568")
self.assertEqual(normalized_decimal("-0.0000001", True), "0")
self.assertEqual(normalized_decimal("12000000"), "12000000")
def test_missing_is_not_zero(self):
self.assertEqual(evaluate("lookup", [cell(None)])["status"], "missing")
self.assertEqual(evaluate("lookup", [cell("0")])["status"], "ok")
with self.assertRaises(AssertionError):
evaluate("difference", [cell(None), cell("1", "2001")])
def test_signed_and_relative_change(self):
rows = [cell("100"), cell("75", "2001")]
self.assertEqual(evaluate("difference", rows)["value"], "-25")
self.assertEqual(evaluate("relative_change", rows)["value"], "-25")
with self.assertRaises(AssertionError):
evaluate("relative_change", [cell("0"), cell("2", "2001")])
with self.assertRaises(AssertionError):
evaluate("relative_change", [cell("-1"), cell("2", "2001")])
def test_percentage_points_not_percent(self):
rows = [cell("5", measure_kind="rate"), cell("7", "2001", measure_kind="rate")]
self.assertEqual(evaluate("percentage_point_difference", rows), {"status": "ok", "value": "2", "unit": "percentage_points"})
with self.assertRaises(AssertionError):
evaluate("relative_change", rows)
def test_tied_and_second_distinct_rank(self):
rows = [cell("10"), cell("10", "2001"), cell("8", "2002"), cell("8", "2003"), cell("2", "2004")]
self.assertEqual(evaluate("maximum_years", rows)["value"], ["2000", "2001"])
self.assertEqual(evaluate("second_highest_years", rows)["value"], ["2002", "2003"])
def test_semantic_join_and_units(self):
rows = [cell("10", indicator_code="NY.GDP.MKTP.CD", unit_code="current_USD"), cell("3", indicator_code="SP.POP.TOTL", cell_id="pop")]
self.assertEqual(evaluate("gdp_per_capita_join", rows)["value"], "3.333333")
bad = copy.deepcopy(rows)
bad[1]["year"] = "2001"
with self.assertRaises(AssertionError):
evaluate("gdp_per_capita_join", bad)
self.assertEqual(evaluate("incompatible_sum", rows)["status"], "incompatible_units")
def test_symbolic_never_reads_gold(self):
q = {"question_ar": "بحسب الخلية المعروضة، ما القيمة المسجلة؟", "context": json.dumps({"cells": [cell("3")]}),
"gold_answer": '{"value":"999"}', "operation": "malicious_gold", "program": "hidden"}
item = model_input(q)
self.assertEqual(set(item), {"question_ar", "table"})
self.assertEqual(symbolic(item)["value"], "3")
item["table"]["cells"][0]["value"] = "4"
self.assertEqual(symbolic(item)["value"], "4")
with self.assertRaises(ValueError):
symbolic({"question_ar": "سؤال غير مدعوم", "table": {"cells": [cell("3")]}})
def test_scorer_boolean_is_not_number(self):
self.assertFalse(value_equal(1, True))
self.assertFalse(value_equal(False, "0"))
self.assertTrue(value_equal(True, True))
self.assertTrue(value_equal("3.000", "3"))
self.assertFalse(value_equal("NaN", "3"))
self.assertFalse(value_equal("Infinity", "3"))
with self.assertRaises(ValueError):
reject_constant("NaN")
def test_scorer_rejects_duplicate_unknown_ids(self):
q = dict(id="q", gold_answer='{"status":"ok","value":"3","unit":"persons"}',
gold_evidence_ids=["c"], split="test", operation="lookup", structure_holdout_diagnostic=False)
p = dict(id="q", answer={"status": "ok", "value": "3.0", "unit": "persons"}, evidence_ids=["c"])
self.assertEqual(score_records([q], [p])["groups"]["all"]["strict_joint_rate"], 1)
with self.assertRaises(AssertionError):
score_records([q], [p, p])
with self.assertRaises(AssertionError):
score_records([q], [dict(p, id="unknown")])
self.assertEqual(score_records([q], [])["groups"]["all"]["answer_exact_rate"], 0)
self.assertEqual(score_records([q], [dict(p, evidence_ids=["c", "unknown"])])["groups"]["all"]["strict_joint_rate"], 0)
self.assertEqual(score_records([q], [dict(p, evidence_ids=["c", "c"])])["groups"]["all"]["strict_joint_rate"], 0)
if __name__ == "__main__":
unittest.main()
|