Keep/keep-notes/tests/unit/rrf.test.ts

import { test, expect } from '@playwright/test';
import { calculateRRFK } from '../../lib/utils';

test.describe('RRF K Calculation Tests', () => {
  test('should return minimum k=20 for small datasets', () => {
    // For small datasets (< 200 notes), k should be 20
    expect(calculateRRFK(0)).toBe(20);
    expect(calculateRRFK(10)).toBe(20);
    expect(calculateRRFK(50)).toBe(20);
    expect(calculateRRFK(100)).toBe(20);
    expect(calculateRRFK(199)).toBe(20);
  });

  test('should return k=20 for exactly 200 notes', () => {
    expect(calculateRRFK(200)).toBe(20);
  });

  test('should scale k for larger datasets', () => {
    // k = max(20, totalNotes / 10)
    expect(calculateRRFK(500)).toBe(50);  // 500/10 = 50
    expect(calculateRRFK(1000)).toBe(100); // 1000/10 = 100
    expect(calculateRRFK(250)).toBe(25);   // 250/10 = 25
  });

  test('should handle edge cases', () => {
    expect(calculateRRFK(201)).toBe(20);   // 201/10 = 20.1 → floor → 20, max(20,20) = 20
    expect(calculateRRFK(210)).toBe(21);   // 210/10 = 21
    expect(calculateRRFK(10000)).toBe(1000); // 10000/10 = 1000
  });

  test('k should always be at least 20', () => {
    // Even for very small datasets, k should not be below 20
    for (let i = 0; i <= 200; i++) {
      expect(calculateRRFK(i)).toBeGreaterThanOrEqual(20);
    }
  });

  test('should be lower than old value for typical datasets', () => {
    // For typical user datasets (< 500 notes), new k should be lower than old k=60
    expect(calculateRRFK(100)).toBeLessThan(60);
    expect(calculateRRFK(200)).toBeLessThan(60);
    expect(calculateRRFK(300)).toBe(30); // 300/10 = 30
    expect(calculateRRFK(500)).toBe(50); // 500/10 = 50
  });

  test('should surpass old value for very large datasets', () => {
    // For very large datasets (> 600 notes), k should be higher than 60
    expect(calculateRRFK(700)).toBe(70); // 700/10 = 70
    expect(calculateRRFK(1000)).toBe(100);
  });
});

test.describe('RRF Ranking Behavior Tests', () => {
  test('RRF with lower k penalizes low ranks more', () => {
    // Simulate RRF scores for a note at rank 5
    // Formula: score = 1 / (k + rank)

    const rank = 5;
    const scoreWithK20 = 1 / (20 + rank);  // 1/25 = 0.04
    const scoreWithK60 = 1 / (60 + rank);  // 1/65 = 0.015

    // Lower k gives higher score to low ranks
    expect(scoreWithK20).toBeGreaterThan(scoreWithK60);
  });

  test('RRF favors items ranked high in both lists', () => {
    // Note A: rank 1 in both lists
    const scoreA_K20 = 1 / (20 + 1) + 1 / (20 + 1); // 2/21 ≈ 0.095

    // Note B: rank 1 in one list, rank 10 in other
    const scoreB_K20 = 1 / (20 + 1) + 1 / (20 + 10); // 1/21 + 1/30 ≈ 0.081

    // Note C: rank 5 in both lists
    const scoreC_K20 = 1 / (20 + 5) + 1 / (20 + 5); // 2/25 = 0.08

    // Note A should have highest score (consistently high)
    expect(scoreA_K20).toBeGreaterThan(scoreB_K20);
    expect(scoreA_K20).toBeGreaterThan(scoreC_K20);
  });

  test('k=20 vs k=60 ranking difference', () => {
    // Note at rank 20
    const rank = 20;

    const scoreWithK20 = 1 / (20 + rank); // 1/40 = 0.025
    const scoreWithK60 = 1 / (60 + rank); // 1/80 = 0.0125

    // With k=20, rank 20 is scored 2x higher than with k=60
    expect(scoreWithK20 / scoreWithK60).toBeCloseTo(2.0, 1);
  });

  test('RRF score should decrease as rank increases', () => {
    const k = 20;

    const scoreRank1 = 1 / (k + 1);
    const scoreRank5 = 1 / (k + 5);
    const scoreRank10 = 1 / (k + 10);
    const scoreRank50 = 1 / (k + 50);

    expect(scoreRank1).toBeGreaterThan(scoreRank5);
    expect(scoreRank5).toBeGreaterThan(scoreRank10);
    expect(scoreRank10).toBeGreaterThan(scoreRank50);
  });

  test('RRF handles missing ranks gracefully', () => {
    // If a note is not in a list, it gets the max rank
    const k = 20;
    const totalNotes = 100;
    const missingRank = totalNotes; // Treated as worst rank

    const scoreWithMissing = 1 / (k + missingRank);
    const scoreWithRank50 = 1 / (k + 50);

    // Missing rank should give much lower score
    expect(scoreWithMissing).toBeLessThan(scoreWithRank50);
  });
});

test.describe('RRF Adaptive Behavior Tests', () => {
  test('adaptive k provides better rankings for small datasets', () => {
    // For 50 notes: k=20 (adaptive) vs k=60 (old)
    const totalNotes = 50;
    const kAdaptive = calculateRRFK(totalNotes); // 20
    const kOld = 60;

    // Compare scores for rank 10 (20% of dataset)
    const rank = 10;
    const scoreAdaptive = 1 / (kAdaptive + rank);
    const scoreOld = 1 / (kOld + rank);

    // Adaptive k gives higher score to mid-rank items in small datasets
    expect(scoreAdaptive).toBeGreaterThan(scoreOld);
  });

  test('adaptive k scales appropriately', () => {
    const datasets = [10, 50, 100, 200, 500, 1000];

    datasets.forEach(notes => {
      const k = calculateRRFK(notes);

      // k should always be at least 20
      expect(k).toBeGreaterThanOrEqual(20);

      // k should scale with dataset size
      if (notes < 200) {
        expect(k).toBe(20);
      } else {
        expect(k).toBe(Math.floor(notes / 10));
      }
    });
  });
});