hiffi-artist-intelligence/tests/test_platform_invariants.py

"""
tests/test_platform_invariants.py
Rigorous mathematical invariance and governance test suite for HIFFI Artist Intelligence.
Guarantees zero-imputation, rank continuity, isolated simulation, and score bounds.
"""

from pathlib import Path
import sys
import pytest
import pandas as pd
import numpy as np

# Ensure project root is on sys.path for test discovery
PROJECT_ROOT = Path(__file__).resolve().parent.parent
if str(PROJECT_ROOT) not in sys.path:
    sys.path.insert(0, str(PROJECT_ROOT))

from core.cpi_engine import CommercialPowerEngine

DATA_PATH = PROJECT_ROOT / "data" / "hiffi_multiplatform_500.csv"


@pytest.fixture(scope="module")
def cohort_df():
    assert DATA_PATH.exists(), f"Master dataset not found at {DATA_PATH}"
    df = pd.read_csv(DATA_PATH)
    assert len(df) == 500, f"Expected 500 rows, found {len(df)}"
    return df


def test_zero_imputation_policy(cohort_df):
    """Enforces that missing streaming telemetry is strictly NaN, never synthetic zeroes."""
    if "spotify_monthly_listeners" in cohort_df.columns:
        zero_listeners = (cohort_df["spotify_monthly_listeners"] == 0).sum()
        assert zero_listeners == 0, f"Violation: Found {zero_listeners} synthetic zeroes in Spotify listeners!"


def test_effective_weight_distribution_sums(cohort_df):
    """Verifies that all dynamic weights sum to 100% (+/- floating point tolerance)."""
    assert "effective_weights" in cohort_df.columns, "Column 'effective_weights' missing."
    for idx, weights_str in enumerate(cohort_df["effective_weights"]):
        if pd.isna(weights_str) or weights_str == "NONE":
            continue
        # Extract percentage numbers: e.g. 'yt:50.0%,sp_monthly:25.0%'
        percentages = [float(p.split(":")[1].replace("%", "")) for p in str(weights_str).split(",")]
        total = sum(percentages)
        assert abs(total - 100.0) < 0.25, f"Row {idx} weights do not sum to 100%: {weights_str} (sum={total})"


def test_published_rank_continuity(cohort_df):
    """Verifies published ranks are unique, consecutive integers from 1 to 500."""
    ranks = cohort_df["hiffi_xp_published_rank"].dropna().astype(int).tolist()
    assert len(ranks) == 500, f"Expected 500 published ranks, got {len(ranks)}"
    expected_ranks = list(range(1, 501))
    assert sorted(ranks) == expected_ranks, "Published ranks contain gaps, duplicates, or out-of-order sequences!"


def test_score_normalization_bounds(cohort_df):
    """Ensures production and XP composite scores are bounded in [0.0, 100.0]."""
    for col in ["production_hiffi_score", "hiffi_xp_score"]:
        if col in cohort_df.columns:
            valid_vals = cohort_df[col].dropna()
            assert (valid_vals >= 0.0).all(), f"Found negative scores in {col}"
            assert (valid_vals <= 100.0).all(), f"Found scores exceeding 100.0 in {col}"


def test_sales_provenance_tiers_clean(cohort_df):
    """Verifies all 500 rows have valid, recognized provenance statuses."""
    valid_statuses = {"VERIFIED_CSPC", "VERIFIED_RIAA", "ESTIMATED_SEA"}
    actual_statuses = set(cohort_df["sales_data_status"].dropna().unique())
    invalid = actual_statuses - valid_statuses
    assert not invalid, f"Detected unrecognized sales statuses: {invalid}"
    assert cohort_df["total_eas_cspc"].isna().sum() == 0, "Found NaN in total_eas_cspc"


def test_cpi_simulation_isolation(cohort_df):
    """Confirms running simulation engine in-memory does not modify published baseline ranks."""
    engine = CommercialPowerEngine()
    sim_df = engine.simulate_ranking(cohort_df, commercial_weight=0.20, mode="global_redistribution")
    
    # Check that published baseline ranks remain identical
    assert (sim_df["hiffi_xp_published_rank"] == cohort_df["hiffi_xp_published_rank"]).all()
    # Check that simulated rank column was generated and populated
    assert "simulated_rank" in sim_df.columns
    assert (sim_df["simulated_rank"] >= 1).all()