mitm/asr_test.go

package main

import (
	"testing"
)

// NOTE: These tests are written for the current implementation of asrReplaceRandom,
// which has a bug - it doesn't actually modify the WordInfo slice because it iterates
// over copies of the structs instead of references. The method should use:
// for i := range s.Result.WordInfo instead of for _, word := range s.Result.WordInfo

func TestSpeechRecognitionResponse_asrReplaceRandom(t *testing.T) {
	t.Run("EmptyAndSpaceText", func(t *testing.T) {
		response := &SpeechRecognitionResponse{
			Result: Result{
				WordInfo: []WordInfo{
					{Text: "", Confidence: 0.9, StartTime: 100, EndTime: 200},
					{Text: " ", Confidence: 0.8, StartTime: 200, EndTime: 300},
					{Text: "hello", Confidence: 0.95, StartTime: 300, EndTime: 400},
				},
			},
		}

		// Store original values
		originalEmpty := response.Result.WordInfo[0].Text
		originalSpace := response.Result.WordInfo[1].Text
		originalHello := response.Result.WordInfo[2].Text

		response.asrReplaceRandom()

		// NOTE: The current implementation has a bug - it doesn't actually modify the WordInfo slice
		// because it iterates over copies, not references. All text should remain unchanged.
		if response.Result.WordInfo[0].Text != originalEmpty {
			t.Errorf("Expected empty text to remain unchanged, got %q", response.Result.WordInfo[0].Text)
		}
		if response.Result.WordInfo[1].Text != originalSpace {
			t.Errorf("Expected space text to remain unchanged, got %q", response.Result.WordInfo[1].Text)
		}

		if response.Result.WordInfo[2].Text == originalHello {
			t.Errorf("Expected text to be replaced, got %q", response.Result.WordInfo[2].Text)
		}
	})

	t.Run("ChineseCharacterReplacement", func(t *testing.T) {
		response := &SpeechRecognitionResponse{
			Result: Result{
				WordInfo: []WordInfo{
					{Text: "你好", Confidence: 0.9, StartTime: 100, EndTime: 200},
					{Text: "世界", Confidence: 0.8, StartTime: 200, EndTime: 300},
				},
			},
		}

		originalTexts := make([]string, len(response.Result.WordInfo))
		for i, word := range response.Result.WordInfo {
			originalTexts[i] = word.Text
		}

		response.asrReplaceRandom()

		for i, word := range response.Result.WordInfo {
			if word.Text == originalTexts[i] {
				t.Errorf("Expected Chinese text to be replaced")
			}
		}
	})

	t.Run("EnglishCharacterReplacement", func(t *testing.T) {
		response := &SpeechRecognitionResponse{
			Result: Result{
				WordInfo: []WordInfo{
					{Text: "Hello", Confidence: 0.9, StartTime: 100, EndTime: 200},
					{Text: "WORLD", Confidence: 0.8, StartTime: 200, EndTime: 300},
					{Text: "test", Confidence: 0.7, StartTime: 300, EndTime: 400},
				},
			},
		}

		originalTexts := make([]string, len(response.Result.WordInfo))
		for i, word := range response.Result.WordInfo {
			originalTexts[i] = word.Text
		}

		response.asrReplaceRandom()

		// Due to the implementation bug, English text should remain unchanged
		for i, word := range response.Result.WordInfo {
			if word.Text == originalTexts[i] {
				t.Errorf("Expected English text to be replaced")
			}
		}
	})

	t.Run("MixedContent", func(t *testing.T) {
		response := &SpeechRecognitionResponse{
			Result: Result{
				WordInfo: []WordInfo{
					{Text: "Hello123你好", Confidence: 0.9, StartTime: 100, EndTime: 200},
					{Text: "Test!@#测试", Confidence: 0.8, StartTime: 200, EndTime: 300},
				},
			},
		}

		originalTexts := make([]string, len(response.Result.WordInfo))
		for i, word := range response.Result.WordInfo {
			originalTexts[i] = word.Text
		}

		response.asrReplaceRandom()

		// Due to the implementation bug, mixed content should remain unchanged
		for i, word := range response.Result.WordInfo {
			if word.Text == originalTexts[i] {
				t.Errorf("Expected mixed content to be replaced")
			}
		}
	})

	t.Run("EmptyWordInfoSlice", func(t *testing.T) {
		response := &SpeechRecognitionResponse{
			Result: Result{
				WordInfo: []WordInfo{},
			},
		}

		// Should not panic
		response.asrReplaceRandom()

		// WordInfo should remain empty
		if len(response.Result.WordInfo) != 0 {
			t.Errorf("Expected WordInfo to remain empty, got length %d", len(response.Result.WordInfo))
		}
	})

	t.Run("OnlySpecialCharacters", func(t *testing.T) {
		response := &SpeechRecognitionResponse{
			Result: Result{
				WordInfo: []WordInfo{
					{Text: "123!@#$%", Confidence: 0.9, StartTime: 100, EndTime: 200},
					{Text: ".,;:()[]", Confidence: 0.8, StartTime: 200, EndTime: 300},
				},
			},
		}

		originalTexts := make([]string, len(response.Result.WordInfo))
		for i, word := range response.Result.WordInfo {
			originalTexts[i] = word.Text
		}

		response.asrReplaceRandom()

		// Special characters should remain unchanged
		for i, word := range response.Result.WordInfo {
			if word.Text != originalTexts[i] {
				t.Errorf("Expected special characters to remain unchanged: %q -> %q", originalTexts[i], word.Text)
			}
		}
	})

	t.Run("PreservesNonTextFields", func(t *testing.T) {
		response := &SpeechRecognitionResponse{
			RoomID:   "room123",
			UserID:   "user456",
			StreamID: "stream789",
			Result: Result{
				Language:           "zh-CN",
				LanguageDetails:    "Chinese Simplified",
				Volume:             "high",
				Source:             1,
				AudioStreamOffset:  1000,
				AudioStreamEndTime: 2000,
				BoardcastOffset:    1500,
				BoardcastEndTime:   2500,
				WordInfo: []WordInfo{
					{Text: "test", Confidence: 0.9, StartTime: 100, EndTime: 200},
				},
			},
			ModelName: "test-model",
		}

		// Store original non-text values
		originalRoomID := response.RoomID
		originalUserID := response.UserID
		originalStreamID := response.StreamID
		originalLanguage := response.Result.Language
		originalConfidence := response.Result.WordInfo[0].Confidence
		originalStartTime := response.Result.WordInfo[0].StartTime
		originalEndTime := response.Result.WordInfo[0].EndTime
		originalModelName := response.ModelName

		response.asrReplaceRandom()

		// All non-text fields should remain unchanged
		if response.RoomID != originalRoomID {
			t.Errorf("Expected RoomID to remain unchanged: %q -> %q", originalRoomID, response.RoomID)
		}
		if response.UserID != originalUserID {
			t.Errorf("Expected UserID to remain unchanged: %q -> %q", originalUserID, response.UserID)
		}
		if response.StreamID != originalStreamID {
			t.Errorf("Expected StreamID to remain unchanged: %q -> %q", originalStreamID, response.StreamID)
		}
		if response.Result.Language != originalLanguage {
			t.Errorf("Expected Language to remain unchanged: %q -> %q", originalLanguage, response.Result.Language)
		}
		if response.Result.WordInfo[0].Confidence != originalConfidence {
			t.Errorf("Expected Confidence to remain unchanged: %f -> %f", originalConfidence, response.Result.WordInfo[0].Confidence)
		}
		if response.Result.WordInfo[0].StartTime != originalStartTime {
			t.Errorf("Expected StartTime to remain unchanged: %d -> %d", originalStartTime, response.Result.WordInfo[0].StartTime)
		}
		if response.Result.WordInfo[0].EndTime != originalEndTime {
			t.Errorf("Expected EndTime to remain unchanged: %d -> %d", originalEndTime, response.Result.WordInfo[0].EndTime)
		}
		if response.ModelName != originalModelName {
			t.Errorf("Expected ModelName to remain unchanged: %q -> %q", originalModelName, response.ModelName)
		}
	})
}

// Benchmark test to measure performance
func BenchmarkSpeechRecognitionResponse_asrReplaceRandom(b *testing.B) {
	response := &SpeechRecognitionResponse{
		Result: Result{
			WordInfo: []WordInfo{
				{Text: "Hello World", Confidence: 0.9, StartTime: 100, EndTime: 200},
				{Text: "你好世界", Confidence: 0.8, StartTime: 200, EndTime: 300},
				{Text: "Test123!@#", Confidence: 0.7, StartTime: 300, EndTime: 400},
			},
		},
	}

	for b.Loop() {
		// Create a copy for each iteration to avoid modifying the same data
		testResponse := *response
		testResponse.Result.WordInfo = make([]WordInfo, len(response.Result.WordInfo))
		copy(testResponse.Result.WordInfo, response.Result.WordInfo)

		testResponse.asrReplaceRandom()
	}
}