minstrel/internal/playlists/foryou_test.go

package playlists

import (
	"testing"
	"time"

	"github.com/jackc/pgx/v5/pgtype"

	"git.fabledsword.com/bvandeusen/minstrel/internal/db/dbq"
	"git.fabledsword.com/bvandeusen/minstrel/internal/recommendation"
)

// makeCand constructs a Candidate with distinct track/album/artist IDs
// and a SimilarityScore that drives distinguishable scores when passed
// through recommendation.Score. Using SimilarityScore alone is sufficient
// because systemMixWeights.SimilarityWeight = 1.5 (non-zero), so
// different similarity values produce different scores.
func makeCand(trackN, albumN, artistN int, similarity float64) recommendation.Candidate {
	var tID, aID, arID pgtype.UUID
	tID.Valid, aID.Valid, arID.Valid = true, true, true
	tID.Bytes[15] = byte(trackN)
	aID.Bytes[15] = byte(albumN)
	arID.Bytes[15] = byte(artistN)
	return recommendation.Candidate{
		Track: dbq.Track{ID: tID, AlbumID: aID, ArtistID: arID},
		Inputs: recommendation.ScoringInputs{
			SimilarityScore: similarity,
		},
	}
}

func TestCapCandidatesByAlbumAndArtist_AlbumCap(t *testing.T) {
	// discoverMaxTracksPerAlbum = 2: third track on album 10 must be dropped.
	in := []recommendation.Candidate{
		makeCand(1, 10, 100, 1.0),
		makeCand(2, 10, 101, 0.9),
		makeCand(3, 10, 102, 0.8), // 3rd from album 10 → drop
		makeCand(4, 11, 103, 0.7),
	}
	got := capCandidatesByAlbumAndArtist(in)
	if len(got) != 3 {
		t.Errorf("len = %d, want 3 (album 10 capped at 2)", len(got))
	}
}

func TestCapCandidatesByAlbumAndArtist_ArtistCap(t *testing.T) {
	// discoverMaxTracksPerArtist = 3: fourth track by artist 100 must be dropped.
	in := []recommendation.Candidate{
		makeCand(1, 10, 100, 1.0),
		makeCand(2, 11, 100, 0.9),
		makeCand(3, 12, 100, 0.8),
		makeCand(4, 13, 100, 0.7), // 4th by artist 100 → drop
		makeCand(5, 14, 101, 0.6),
	}
	got := capCandidatesByAlbumAndArtist(in)
	if len(got) != 4 {
		t.Errorf("len = %d, want 4 (artist 100 capped at 3)", len(got))
	}
}

func TestCapCandidatesByAlbumAndArtist_PreservesOrder(t *testing.T) {
	// All distinct albums and artists: all kept, original order preserved.
	in := []recommendation.Candidate{
		makeCand(3, 10, 100, 0.5),
		makeCand(1, 11, 101, 0.9),
		makeCand(2, 12, 102, 0.7),
	}
	got := capCandidatesByAlbumAndArtist(in)
	if len(got) != 3 {
		t.Fatalf("len = %d, want 3", len(got))
	}
	for i, want := range []byte{3, 1, 2} {
		if got[i].Track.ID.Bytes[15] != want {
			t.Errorf("got[%d].ID = %d, want %d", i, got[i].Track.ID.Bytes[15], want)
		}
	}
}

func TestCapCandidatesByAlbumAndArtist_Empty(t *testing.T) {
	got := capCandidatesByAlbumAndArtist(nil)
	if len(got) != 0 {
		t.Errorf("len = %d, want 0", len(got))
	}
}

func TestPickHeadAndTail_SmallPool(t *testing.T) {
	// Pool < headN+tailN → return up to total entries, no tail split.
	in := []recommendation.Candidate{
		makeCand(1, 10, 100, 1.0),
		makeCand(2, 11, 101, 0.9),
		makeCand(3, 12, 102, 0.8),
	}
	got := pickHeadAndTail(in, "2026-05-07", time.Now(), 5, 2)
	if len(got) != 3 {
		t.Errorf("len = %d, want 3 (pool too small for head/tail split)", len(got))
	}
}

func TestPickHeadAndTail_ExactlyTotal(t *testing.T) {
	// Pool == headN+tailN: no tail to sample from, returns all.
	in := make([]recommendation.Candidate, 0, 7)
	for i := 0; i < 7; i++ {
		in = append(in, makeCand(i+1, i+1, i+1, float64(7-i)/10.0))
	}
	got := pickHeadAndTail(in, "2026-05-07", time.Now(), 5, 2)
	if len(got) != 7 {
		t.Errorf("len = %d, want 7 (pool == total)", len(got))
	}
}

func TestPickHeadAndTail_HeadAndTailSplit(t *testing.T) {
	// Pool of 100 distinct (album, artist) pairs; no caps trim.
	// With headN=20, tailN=5, expect 25 entries total.
	in := make([]recommendation.Candidate, 0, 100)
	for i := 0; i < 100; i++ {
		in = append(in, makeCand(i+1, i+1, i+1, float64(100-i)/100.0))
	}
	got := pickHeadAndTail(in, "2026-05-07", time.Now(), 20, 5)
	if len(got) != 25 {
		t.Errorf("len = %d, want 25 (20 head + 5 tail)", len(got))
	}
}

func TestPickHeadAndTail_Determinism(t *testing.T) {
	// Same inputs, same dateStr → identical output both times.
	in := make([]recommendation.Candidate, 0, 100)
	for i := 0; i < 100; i++ {
		in = append(in, makeCand(i+1, i+1, i+1, float64(100-i)/100.0))
	}
	now := time.Now()
	got1 := pickHeadAndTail(in, "2026-05-07", now, 20, 5)
	got2 := pickHeadAndTail(in, "2026-05-07", now, 20, 5)
	if len(got1) != len(got2) {
		t.Fatalf("len mismatch: %d vs %d", len(got1), len(got2))
	}
	for i := range got1 {
		if got1[i].TrackID.Bytes[15] != got2[i].TrackID.Bytes[15] {
			t.Errorf("position %d differs across calls (not deterministic within day)", i)
			break
		}
	}
}

func TestPickHeadAndTail_HeadStable_TailVariesAcrossDays(t *testing.T) {
	// Head (positions 0..headN-1) must match across different date strings.
	// Tail (positions headN..headN+tailN-1) should differ for different dates
	// because tieBreakHash incorporates the date.
	in := make([]recommendation.Candidate, 0, 100)
	for i := 0; i < 100; i++ {
		in = append(in, makeCand(i+1, i+1, i+1, float64(100-i)/100.0))
	}
	now := time.Now()
	day1 := pickHeadAndTail(in, "2026-05-07", now, 20, 5)
	day2 := pickHeadAndTail(in, "2026-05-08", now, 20, 5)
	if len(day1) != 25 || len(day2) != 25 {
		t.Fatalf("len mismatch: day1=%d day2=%d", len(day1), len(day2))
	}

	// Head (score-sorted, deterministic): positions 0..19 must match.
	for i := 0; i < 20; i++ {
		if day1[i].TrackID.Bytes[15] != day2[i].TrackID.Bytes[15] {
			t.Errorf("head position %d differs across date strings (should be score-stable)", i)
		}
	}

	// Tail (tieBreakHash order): at least one position should differ.
	tailDiffers := false
	for i := 20; i < 25; i++ {
		if day1[i].TrackID.Bytes[15] != day2[i].TrackID.Bytes[15] {
			tailDiffers = true
			break
		}
	}
	if !tailDiffers {
		t.Errorf("tail identical across different date strings (tieBreakHash not using dateStr)")
	}
}

func TestPickHeadAndTail_TailFromBeyond2xHeadN(t *testing.T) {
	// With headN=5 and a 50-candidate pool, tailStart = 2*5 = 10.
	// Tail samples come from positions >=10, not from positions 5..9.
	// Verify: the first 5 positions contain tracks from the top-5 by score,
	// and positions 5..9 (the "buffer zone") do not appear in the tail.
	//
	// Tracks 1..50 have similarity 0.99..0.01 (descending), so
	// after scoring the sorted order is tracks 1,2,3,...,50.
	// Top 5 head = tracks 1..5.
	// Buffer zone (positions 5..9) = tracks 6..10.
	// Tail pool = tracks 11..50; tail sample = 3 from that pool.
	in := make([]recommendation.Candidate, 0, 50)
	for i := 0; i < 50; i++ {
		// trackN = i+1, score descends: track 1 scores highest.
		sim := float64(50-i) / 50.0
		in = append(in, makeCand(i+1, i+1, i+1, sim))
	}
	got := pickHeadAndTail(in, "2026-05-07", time.Now(), 5, 3)
	if len(got) != 8 {
		t.Fatalf("len = %d, want 8 (5 head + 3 tail)", len(got))
	}

	// Head tracks must be tracks 1..5 (highest scorers).
	for i := 0; i < 5; i++ {
		if got[i].TrackID.Bytes[15] != byte(i+1) {
			t.Errorf("head[%d] = track %d, want track %d",
				i, got[i].TrackID.Bytes[15], i+1)
		}
	}

	// Buffer-zone tracks 6..10 must NOT appear in the tail positions 5..7.
	bufferSet := map[byte]bool{6: true, 7: true, 8: true, 9: true, 10: true}
	for i := 5; i < 8; i++ {
		if bufferSet[got[i].TrackID.Bytes[15]] {
			t.Errorf("tail[%d] = track %d (buffer zone); should come from positions >=10",
				i-5, got[i].TrackID.Bytes[15])
		}
	}
}

func TestPickTopN_DiversityCap(t *testing.T) {
	// Verify pickTopN now enforces the diversity cap.
	// 5 tracks from the same artist: only 3 should survive the cap.
	in := []recommendation.Candidate{
		makeCand(1, 10, 100, 1.0),
		makeCand(2, 11, 100, 0.9),
		makeCand(3, 12, 100, 0.8),
		makeCand(4, 13, 100, 0.7), // 4th by artist → dropped by cap
		makeCand(5, 14, 100, 0.6), // 5th by artist → dropped by cap
	}
	got := pickTopN(in, "2026-05-07", time.Now(), 25)
	if len(got) != 3 {
		t.Errorf("len = %d, want 3 (artist 100 capped at 3)", len(got))
	}
}