84 lines
2.7 KiB
Python
84 lines
2.7 KiB
Python
|
# retoor <retoor@molodetz.nl>
|
||
|
|
|
||
|
|
from devplacepy.services.jobs.isslop.analysis.scoring import (
|
||
|
|
CATEGORY_HUMAN_CLEAN,
|
||
|
|
CATEGORY_HUMAN_MESSY,
|
||
|
|
CATEGORY_SLOP,
|
||
|
|
CATEGORY_SOPHISTICATED,
|
||
|
|
CATEGORY_UNCERTAIN,
|
||
|
|
categorize,
|
||
|
|
compose_slop_score,
|
||
|
|
grade_for,
|
||
|
|
)
|
||
|
|
|
||
|
|
|
||
|
|
def test_compose_slop_score_bounds():
|
||
|
|
assert compose_slop_score(0.0, 0.0) == 0.0
|
||
|
|
assert compose_slop_score(100.0, 100.0) == 100.0
|
||
|
|
assert 0.0 <= compose_slop_score(50.0, 50.0) <= 100.0
|
||
|
|
|
||
|
|
|
||
|
|
def test_compose_slop_score_weighs_ai_heavier_than_quality():
|
||
|
|
ai_heavy = compose_slop_score(80.0, 20.0)
|
||
|
|
quality_heavy = compose_slop_score(20.0, 80.0)
|
||
|
|
assert ai_heavy > quality_heavy
|
||
|
|
|
||
|
|
|
||
|
|
def test_grade_for_is_monotonic():
|
||
|
|
grades = [grade_for(value) for value in (0.0, 20.0, 40.0, 60.0, 80.0, 100.0)]
|
||
|
|
order = "ABCDF"
|
||
|
|
positions = [order.index(grade) for grade in grades]
|
||
|
|
assert positions == sorted(positions)
|
||
|
|
assert grades[0] == "A"
|
||
|
|
assert grades[-1] == "F"
|
||
|
|
|
||
|
|
|
||
|
|
def test_categorize_corners():
|
||
|
|
assert categorize(95.0, 95.0) == CATEGORY_SLOP
|
||
|
|
assert categorize(95.0, 5.0) == CATEGORY_SOPHISTICATED
|
||
|
|
assert categorize(5.0, 5.0) == CATEGORY_HUMAN_CLEAN
|
||
|
|
assert categorize(5.0, 95.0) == CATEGORY_HUMAN_MESSY
|
||
|
|
assert categorize(50.0, 50.0) == CATEGORY_UNCERTAIN
|
||
|
|
|
||
|
|
|
||
|
|
def test_ai_fraction_is_continuous_and_monotonic():
|
||
|
|
from devplacepy.services.jobs.isslop.analysis.scoring import ai_fraction
|
||
|
|
|
||
|
|
assert ai_fraction(0.0) == 0.0
|
||
|
|
assert ai_fraction(35.0) == 0.0
|
||
|
|
assert ai_fraction(50.0) == 0.5
|
||
|
|
assert ai_fraction(65.0) == 1.0
|
||
|
|
assert ai_fraction(100.0) == 1.0
|
||
|
|
samples = [ai_fraction(v) for v in range(30, 71)]
|
||
|
|
assert samples == sorted(samples)
|
||
|
|
steps = [b - a for a, b in zip(samples, samples[1:])]
|
||
|
|
assert max(steps) < 0.06
|
||
|
|
|
||
|
|
|
||
|
|
def test_adjust_for_images_recomputes_grade_consistently():
|
||
|
|
from devplacepy.services.jobs.isslop.analysis.scoring import (
|
||
|
|
RepoScores,
|
||
|
|
adjust_for_images,
|
||
|
|
compose_slop_score,
|
||
|
|
grade_for,
|
||
|
|
)
|
||
|
|
|
||
|
|
scores = RepoScores(
|
||
|
|
origin_score=10.0,
|
||
|
|
quality_deficit=10.0,
|
||
|
|
slop_score=compose_slop_score(2.0, 10.0),
|
||
|
|
grade=grade_for(compose_slop_score(2.0, 10.0)),
|
||
|
|
category="human-clean",
|
||
|
|
human_percent=98.0,
|
||
|
|
ai_percent=2.0,
|
||
|
|
confidence="medium",
|
||
|
|
strong_signal_count=0,
|
||
|
|
medium_signal_count=0,
|
||
|
|
files_scored=50,
|
||
|
|
)
|
||
|
|
adjusted = adjust_for_images(scores, 90.0)
|
||
|
|
assert adjusted.ai_percent == round(0.75 * 2.0 + 0.25 * 90.0, 1)
|
||
|
|
assert adjusted.human_percent == round(100.0 - adjusted.ai_percent, 1)
|
||
|
|
assert adjusted.slop_score == compose_slop_score(adjusted.ai_percent, scores.quality_deficit)
|
||
|
|
assert adjusted.grade == grade_for(adjusted.slop_score)
|