package semantic

import (
	"testing"

	"github.com/kami/maven/internal/router"
)

func TestValidRoute(t *testing.T) {
	for _, r := range AllRoutes {
		if !ValidRoute(r) {
			t.Errorf("ValidRoute(%q) = false, want true", r)
		}
	}
	if ValidRoute("bogus") {
		t.Error("ValidRoute(\"bogus\") = true, want false")
	}
}

func TestIntentToRouteMapping(t *testing.T) {
	cases := []struct {
		intent router.Intent
		route  SemanticRoute
	}{
		{router.IntentChat, RouteConversation},
		{router.IntentQuery, RouteKnowledge},
		{router.IntentAct, RouteAction},
		{router.IntentReminder, RouteAction},
		{router.IntentFact, RouteMemoryWrite},
		{router.IntentNote, RouteMemoryWrite},
		{router.IntentSystem, RouteSystem},
		{router.Intent("unknown"), RouteUncertain},
		{router.Intent(""), RouteUncertain},
	}
	for _, c := range cases {
		got := IntentToRoute(c.intent)
		if got != c.route {
			t.Errorf("IntentToRoute(%q) = %q, want %q", c.intent, got, c.route)
		}
	}
}

func TestLoadCorpus(t *testing.T) {
	exs, err := LoadCorpus()
	if err != nil {
		t.Fatal(err)
	}
	if len(exs) == 0 {
		t.Fatal("corpus is empty")
	}
	for _, e := range exs {
		if !ValidRoute(e.Route) {
			t.Errorf("example %q has invalid route %q", e.SourceID, e.Route)
		}
		if e.Source == "" {
			t.Errorf("example %q has empty source", e.SourceID)
		}
		if e.SplitGroup == "" {
			t.Errorf("example %q has empty split_group", e.SourceID)
		}
	}
	fastPath, residual := SplitCounts(exs)
	if fastPath == 0 {
		t.Error("no fast_path_resolved examples in corpus")
	}
	if residual == 0 {
		t.Error("no residual examples in corpus")
	}
	t.Logf("corpus: %d examples (%d fast-path, %d residual)", len(exs), fastPath, residual)
}

func TestCorpusValidation(t *testing.T) {
	exs, err := LoadCorpus()
	if err != nil {
		t.Fatal(err)
	}
	stats := CorpusStatsFrom(exs)

	// Total matches.
	if stats.Total != len(exs) {
		t.Errorf("stats.Total=%d, want %d", stats.Total, len(exs))
	}

	// Route counts sum to total.
	routeSum := 0
	for _, c := range stats.RouteCounts {
		routeSum += c
	}
	if routeSum != stats.Total {
		t.Errorf("route counts sum to %d, want %d", routeSum, stats.Total)
	}

	// Source counts sum to total.
	sourceSum := 0
	for _, c := range stats.SourceCounts {
		sourceSum += c
	}
	if sourceSum != stats.Total {
		t.Errorf("source counts sum to %d, want %d", sourceSum, stats.Total)
	}

	// Fast-path + residual == total.
	if stats.FastPath+stats.Residual != stats.Total {
		t.Errorf("fast_path(%d) + residual(%d) = %d, want %d",
			stats.FastPath, stats.Residual, stats.FastPath+stats.Residual, stats.Total)
	}

	// All SourceIDs are valid (non-empty, checked by ValidateCorpus).
	// All SplitGroups are non-empty (checked by ValidateCorpus).
	// No duplicate corpus identity (checked by ValidateCorpus).
	// No identical Text with conflicting labels (checked by ValidateCorpus).

	t.Logf("validation passed: %d rows, route_hash=%s", stats.Total, stats.DatasetHash)

	// Verify the exact counts match what we expect.
	if stats.Total != 136 {
		t.Errorf("total = %d, want 136", stats.Total)
	}
	if stats.SourceCounts["ru_routing_v1"] != 96 {
		t.Errorf("ru_routing_v1 count = %d, want 96", stats.SourceCounts["ru_routing_v1"])
	}
	if stats.SourceCounts["contrastive"] != 40 {
		t.Errorf("contrastive count = %d, want 40", stats.SourceCounts["contrastive"])
	}
}

func TestCorpusByRoute(t *testing.T) {
	exs, err := LoadCorpus()
	if err != nil {
		t.Fatal(err)
	}
	byRoute := ByRoute(exs)
	for _, route := range AllRoutes {
		count := len(byRoute[route])
		if count == 0 {
			t.Errorf("no examples for route %q", route)
		}
		t.Logf("  %s: %d examples", route, count)
	}
}

func TestSplitByFamily(t *testing.T) {
	exs, err := LoadCorpus()
	if err != nil {
		t.Fatal(err)
	}
	train, eval := SplitByFamily(exs, 0.8)
	total := len(train) + len(eval)
	if total != len(exs) {
		t.Errorf("split lost examples: train=%d eval=%d total=%d, want %d",
			len(train), len(eval), total, len(exs))
	}
	trainGroups := map[string]bool{}
	for _, e := range train {
		trainGroups[e.SplitGroup] = true
	}
	for _, e := range eval {
		if trainGroups[e.SplitGroup] {
			t.Errorf("split_group %q leaked across train/eval (example %q)",
				e.SplitGroup, e.SourceID)
		}
	}
	t.Logf("split: %d train, %d eval", len(train), len(eval))
}

func TestFrozenHoldoutSplit(t *testing.T) {
	exs, err := LoadCorpus()
	if err != nil {
		t.Fatal(err)
	}
	frozen, dev, hash := FrozenHoldoutSplit(exs)
	total := len(frozen) + len(dev)
	if total != len(exs) {
		t.Errorf("split lost examples: frozen=%d dev=%d total=%d, want %d",
			len(frozen), len(dev), total, len(exs))
	}
	if len(frozen) == 0 {
		t.Error("frozen holdout is empty")
	}
	if len(dev) == 0 {
		t.Error("development pool is empty")
	}

	// No split_group leakage.
	frozenGroups := map[string]bool{}
	for _, e := range frozen {
		frozenGroups[e.SplitGroup] = true
	}
	for _, e := range dev {
		if frozenGroups[e.SplitGroup] {
			t.Errorf("split_group %q leaked across frozen/dev (example %q)",
				e.SplitGroup, e.SourceID)
		}
	}

	// Determinism: same split produces same hash.
	frozen2, dev2, hash2 := FrozenHoldoutSplit(exs)
	if hash != hash2 {
		t.Errorf("non-deterministic: hash=%s vs %s", hash, hash2)
	}
	if len(frozen) != len(frozen2) || len(dev) != len(dev2) {
		t.Errorf("non-deterministic: frozen=%d/%d dev=%d/%d",
			len(frozen), len(frozen2), len(dev), len(dev2))
	}

	// Log route coverage in both splits (not all routes need to appear
	// in the frozen holdout — at 15% of 136 rows, small routes may miss).
	frozenRoutes := map[SemanticRoute]bool{}
	for _, e := range frozen {
		frozenRoutes[e.Route] = true
	}
	devRoutes := map[SemanticRoute]bool{}
	for _, e := range dev {
		devRoutes[e.Route] = true
	}
	for _, r := range AllRoutes {
		if !frozenRoutes[r] {
			t.Logf("note: route %q missing from frozen holdout (expected at 15%%)", r)
		}
		if !devRoutes[r] {
			t.Errorf("route %q missing from development pool", r)
		}
	}

	t.Logf("frozen: %d examples, dev: %d examples, holdout_hash=%s",
		len(frozen), len(dev), hash)
}

func TestGroupedCVFolds(t *testing.T) {
	exs, err := LoadCorpus()
	if err != nil {
		t.Fatal(err)
	}
	_, dev, _ := FrozenHoldoutSplit(exs)
	folds := GroupedCVFolds(dev, 5)
	if len(folds) != 5 {
		t.Fatalf("expected 5 folds, got %d", len(folds))
	}

	for i, f := range folds {
		total := len(f.Train) + len(f.Eval)
		if total != len(dev) {
			t.Errorf("fold %d: train=%d + eval=%d = %d, want %d",
				i, len(f.Train), len(f.Eval), total, len(dev))
		}
		if len(f.Eval) == 0 {
			t.Errorf("fold %d: empty eval set", i)
		}

		// No split_group leakage within a fold.
		evalGroups := map[string]bool{}
		for _, e := range f.Eval {
			evalGroups[e.SplitGroup] = true
		}
		for _, e := range f.Train {
			if evalGroups[e.SplitGroup] {
				t.Errorf("fold %d: split_group %q leaked (example %q)",
					i, e.SplitGroup, e.SourceID)
			}
		}
	}

	// Every example appears in exactly one eval set across all folds.
	evalCount := 0
	for _, f := range folds {
		evalCount += len(f.Eval)
	}
	if evalCount != len(dev) {
		t.Errorf("total eval examples across folds: %d, want %d", evalCount, len(dev))
	}

	t.Logf("grouped CV: %d folds on %d development examples", len(folds), len(dev))
}

func TestFamilyID(t *testing.T) {
	id1 := FamilyID("ru-act-002", "negation")
	id2 := FamilyID("ru-act-002", "question")
	id3 := FamilyID("ru-act-002", "")
	id4 := FamilyID("ru-act-001", "negation")
	if id1 == id2 {
		t.Error("different transforms produced same FamilyID")
	}
	if id1 == id3 {
		t.Error("transform and base produced same FamilyID")
	}
	if id1 == id4 {
		t.Error("different base IDs produced same FamilyID")
	}
}

func TestContrastFamiliesShareSplitGroup(t *testing.T) {
	exs, err := LoadCorpus()
	if err != nil {
		t.Fatal(err)
	}

	// Group contrastive examples by their base source_id.
	baseGroups := map[string]map[string]bool{} // source_id → {split_group: true}
	for _, e := range exs {
		if e.Source != "contrastive" {
			continue
		}
		if baseGroups[e.SourceID] == nil {
			baseGroups[e.SourceID] = map[string]bool{}
		}
		baseGroups[e.SourceID][e.SplitGroup] = true
	}

	// Every base seed's contrastive variants must share exactly one
	// split_group (the base seed's own group).
	for sourceID, groups := range baseGroups {
		if len(groups) != 1 {
			t.Errorf("base %q has contrastive variants in %d split groups: %v",
				sourceID, len(groups), groups)
		}
	}

	// Also verify that contrastive variants share the split_group with
	// their base seed from ru_routing_v1.
	ruGroups := map[string]string{} // source_id → split_group
	for _, e := range exs {
		if e.Source == "ru_routing_v1" {
			ruGroups[e.SourceID] = e.SplitGroup
		}
	}
	for sourceID, groups := range baseGroups {
		var contrastGroup string
		for g := range groups {
			contrastGroup = g
		}
		if ruGroup, ok := ruGroups[sourceID]; ok && contrastGroup != ruGroup {
			t.Errorf("base %q: ru_routing_v1 split_group=%q, contrastive split_group=%q",
				sourceID, ruGroup, contrastGroup)
		}
	}

	t.Logf("contrast family leakage check passed: %d base seeds, all groups consistent",
		len(baseGroups))
}