fetch_ml/tests/integration/consistency/dataset_hash_test.go

//go:build cgo
// +build cgo

package consistency

import (
	"os"
	"path/filepath"
	"testing"
)

// TestDatasetHashConsistency verifies all implementations produce identical hashes
func TestDatasetHashConsistency(t *testing.T) {
	fixturesDir := filepath.Join("..", "..", "fixtures", "consistency")

	expected, err := LoadExpectedHashes(fixturesDir)
	if err != nil {
		t.Fatalf("Failed to load expected hashes: %v", err)
	}

	// Initialize implementations
	impls := []Implementation{
		NewNativeImpl(),
		NewGoImpl(),
		NewZigImpl(),
	}

	// Check which implementations are available
	availableCount := 0
	for _, impl := range impls {
		if impl.Available() {
			availableCount++
			t.Logf("Implementation available: %s", impl.Name())
		} else {
			t.Logf("Implementation not available: %s", impl.Name())
		}
	}

	if availableCount < 2 {
		t.Skip("Need at least 2 implementations for consistency testing")
	}

	// Test each fixture
	for _, fixture := range expected.Fixtures {
		t.Run(fixture.ID, func(t *testing.T) {
			testFixture(t, fixturesDir, &fixture, impls)
		})
	}
}

// TestDatasetHashSmoke runs a quick smoke test
func TestDatasetHashSmoke(t *testing.T) {
	fixturesDir := filepath.Join("..", "..", "fixtures", "consistency")

	// Just test single file fixture for quick validation
	fixturePath := filepath.Join(fixturesDir, "dataset_hash", "02_single_file", "input")

	// Verify fixture exists
	if _, err := os.Stat(fixturePath); os.IsNotExist(err) {
		t.Skipf("Fixture not found: %s", fixturePath)
	}

	impls := []Implementation{
		NewNativeImpl(),
		NewGoImpl(),
		NewZigImpl(),
	}

	results, err := ComputeAllHashes(fixturePath, impls)
	if err != nil {
		t.Logf("Errors during hash computation: %v", err)
	}

	expected := "6dd7e8e932ea9d58555d7fee44a9b01a9bd7448e986636b728ee3711b01f37ce"
	match, mismatches := CompareHashes(results, expected)

	t.Logf("\n%s", FormatHashComparison(results, expected))

	if !match {
		for _, m := range mismatches {
			t.Errorf("Mismatch: %s", m)
		}
	}
}

// TestCrossImplEquivalence compares implementations against each other
func TestCrossImplEquivalence(t *testing.T) {
	fixturesDir := filepath.Join("..", "..", "fixtures", "consistency")

	impls := []Implementation{
		NewGoImpl(),
		NewNativeImpl(),
		NewZigImpl(),
	}

	// Find first available implementation as reference
	var reference Implementation
	for _, impl := range impls {
		if impl.Available() {
			reference = impl
			break
		}
	}

	if reference == nil {
		t.Skip("No implementations available")
	}

	t.Logf("Using %s as reference implementation", reference.Name())

	// Test fixtures
	fixtures := []string{
		"02_single_file",
		"03_nested",
		"04_multiple_files",
	}

	for _, fixtureName := range fixtures {
		t.Run(fixtureName, func(t *testing.T) {
			fixturePath := filepath.Join(fixturesDir, "dataset_hash", fixtureName, "input")

			if _, err := os.Stat(fixturePath); os.IsNotExist(err) {
				t.Skipf("Fixture not found: %s", fixturePath)
			}

			// Get reference hash
			refHash, err := reference.HashDataset(fixturePath)
			if err != nil {
				t.Fatalf("Reference implementation failed: %v", err)
			}

			// Compare all other implementations
			for _, impl := range impls {
				if impl == reference || !impl.Available() {
					continue
				}

				hash, err := impl.HashDataset(fixturePath)
				if err != nil {
					t.Errorf("%s failed: %v", impl.Name(), err)
					continue
				}

				if hash != refHash {
					t.Errorf("%s mismatch: got %s, reference (%s) has %s",
						impl.Name(), hash, reference.Name(), refHash)
				} else {
					t.Logf("%s matches reference ✓", impl.Name())
				}
			}
		})
	}
}

// TestEmptyDirectory specifically tests empty directory handling
func TestEmptyDirectory(t *testing.T) {
	fixturesDir := filepath.Join("..", "..", "fixtures", "consistency")
	fixturePath := filepath.Join(fixturesDir, "dataset_hash", "01_empty_dir", "input")

	expected := "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"

	impls := []Implementation{
		NewGoImpl(),
		NewNativeImpl(),
		NewZigImpl(),
	}

	for _, impl := range impls {
		if !impl.Available() {
			continue
		}

		t.Run(impl.Name(), func(t *testing.T) {
			hash, err := impl.HashDataset(fixturePath)
			if err != nil {
				t.Fatalf("Failed to hash empty directory: %v", err)
			}

			if hash != expected {
				t.Errorf("Empty directory hash mismatch: got %s, expected %s", hash, expected)
			}
		})
	}
}

// testFixture tests a single fixture against all implementations
func testFixture(t *testing.T, fixturesDir string, fixture *Fixture, impls []Implementation) {
	fixturePath := filepath.Join(fixturesDir, "dataset_hash", fixture.ID, "input")

	// Verify fixture exists
	if _, err := os.Stat(fixturePath); os.IsNotExist(err) {
		t.Skipf("Fixture not found: %s", fixturePath)
	}

	// Skip fixtures with TODO expected hashes
	if fixture.ExpectedHash == "TODO_COMPUTE" {
		t.Skipf("Fixture %s has uncomputed expected hash", fixture.ID)
	}

	results, err := ComputeAllHashes(fixturePath, impls)
	if err != nil {
		t.Logf("Errors during hash computation: %v", err)
	}

	match, mismatches := CompareHashes(results, fixture.ExpectedHash)

	// Log comparison for debugging
	t.Logf("\nFixture: %s - %s", fixture.ID, fixture.Name)
	t.Logf("\n%s", FormatHashComparison(results, fixture.ExpectedHash))

	if !match {
		for _, m := range mismatches {
			t.Errorf("%s", m)
		}
	}
}