refactor: move LoadImage

internal/builtin/utils.go

@@ -6,9 +6,13 @@ import (
 	"context"
 	"crypto/hmac"
 	"crypto/sha256"
+	"encoding/base64"
 	"encoding/csv"
 	builtinJSON "encoding/json"
 	"fmt"
+	"image"
+	"image/jpeg"
+	"image/png"
 	"math"
 	"math/rand"
 	"net"
@@ -26,6 +30,7 @@ import (
 
 	"github.com/httprunner/httprunner/v5/code"
 	"github.com/httprunner/httprunner/v5/internal/json"
+	"github.com/httprunner/httprunner/v5/uixt/types"
 )
 
 func Dump2JSON(data interface{}, path string) error {
@@ -484,3 +489,41 @@ func RunCommandWithCallback(cmdName string, args []string, callback LineCallback
 		return err
 	}
 }
+
+// LoadImage loads image file and returns base64 encoded string and image size
+func LoadImage(imagePath string) (base64Str string, size types.Size, err error) {
+	// Read the image file
+	imageFile, err := os.OpenFile(imagePath, os.O_RDONLY, 0o600)
+	if err != nil {
+		return "", types.Size{}, fmt.Errorf("failed to open image file: %w", err)
+	}
+	defer imageFile.Close()
+
+	// Decode the image to get its resolution
+	imageData, format, err := image.Decode(imageFile)
+	if err != nil {
+		return "", types.Size{}, fmt.Errorf("failed to decode image: %w", err)
+	}
+
+	// Get the resolution of the image
+	width := imageData.Bounds().Dx()
+	height := imageData.Bounds().Dy()
+	size = types.Size{Width: width, Height: height}
+
+	// Convert image to base64
+	buf := new(bytes.Buffer)
+	if format == "jpeg" || format == "jpg" {
+		if err := jpeg.Encode(buf, imageData, nil); err != nil {
+			return "", types.Size{}, fmt.Errorf("failed to encode image to buffer: %w", err)
+		}
+	} else {
+		// default use png
+		if err := png.Encode(buf, imageData); err != nil {
+			return "", types.Size{}, fmt.Errorf("failed to encode image to buffer: %w", err)
+		}
+	}
+	base64Str = fmt.Sprintf("data:image/%s;base64,%s", format,
+		base64.StdEncoding.EncodeToString(buf.Bytes()))
+
+	return base64Str, size, nil
+}

internal/version/VERSION

@@ -1 +1 @@
-v5.0.0-beta-2504301521
+v5.0.0-beta-2504301621

uixt/ai/asserter_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"testing"
 
+	"github.com/httprunner/httprunner/v5/internal/builtin"
 	"github.com/httprunner/httprunner/v5/uixt/option"
 	"github.com/httprunner/httprunner/v5/uixt/types"
 	"github.com/stretchr/testify/assert"
@@ -50,7 +51,7 @@ func TestValidAssertions(t *testing.T) {
 
 	for _, tc := range testCases {
 		t.Run(tc.name, func(t *testing.T) {
-			imageBase64, size, err := loadImage(tc.imagePath)
+			imageBase64, size, err := builtin.LoadImage(tc.imagePath)
 			require.NoError(t, err)
 
 			result, err := asserter.Assert(&AssertOptions{

uixt/ai/planner_test.go

@@ -1,18 +1,12 @@
 package ai
 
 import (
-	"bytes"
 	"context"
-	"encoding/base64"
-	"fmt"
-	"image"
-	"image/jpeg"
-	"image/png"
-	"os"
 	"testing"
 
 	"github.com/cloudwego/eino/schema"
 	"github.com/httprunner/httprunner/v5/code"
+	"github.com/httprunner/httprunner/v5/internal/builtin"
 	"github.com/httprunner/httprunner/v5/uixt/option"
 	"github.com/httprunner/httprunner/v5/uixt/types"
 	"github.com/stretchr/testify/assert"
@@ -20,7 +14,7 @@ import (
 )
 
 func TestVLMPlanning(t *testing.T) {
-	imageBase64, size, err := loadImage("testdata/llk_1.png")
+	imageBase64, size, err := builtin.LoadImage("testdata/llk_1.png")
 	require.NoError(t, err)
 
 	userInstruction := `连连看是一款经典的益智消除类小游戏，通常以图案或图标为主要元素。以下是连连看的基本规则说明：
@@ -104,7 +98,7 @@ func TestVLMPlanning(t *testing.T) {
 }
 
 func TestXHSPlanning(t *testing.T) {
-	imageBase64, size, err := loadImage("testdata/xhs-feed.jpeg")
+	imageBase64, size, err := builtin.LoadImage("testdata/xhs-feed.jpeg")
 	require.NoError(t, err)
 
 	userInstruction := "点击第二个帖子的作者头像"
@@ -177,7 +171,7 @@ func TestXHSPlanning(t *testing.T) {
 }
 
 func TestChatList(t *testing.T) {
-	imageBase64, size, err := loadImage("testdata/chat_list.jpeg")
+	imageBase64, size, err := builtin.LoadImage("testdata/chat_list.jpeg")
 	require.NoError(t, err)
 
 	userInstruction := "请结合图片的文字信息，请告诉我一共有多少个群聊，哪些群聊右下角有绿点"
@@ -232,7 +226,7 @@ func TestHandleSwitch(t *testing.T) {
 	}
 
 	for _, tc := range testCases {
-		imageBase64, size, err := loadImage(tc.imageFile)
+		imageBase64, size, err := builtin.LoadImage(tc.imageFile)
 		require.NoError(t, err)
 
 		opts := &PlanningOptions{
@@ -263,7 +257,7 @@ func TestHandleSwitch(t *testing.T) {
 }
 
 func TestValidateInput(t *testing.T) {
-	imageBase64, size, err := loadImage("testdata/popup_risk_warning.png")
+	imageBase64, size, err := builtin.LoadImage("testdata/popup_risk_warning.png")
 	require.NoError(t, err)
 
 	tests := []struct {
@@ -388,87 +382,18 @@ func TestProcessVLMResponse(t *testing.T) {
 	}
 }
 
-func TestSavePositionImg(t *testing.T) {
-	imageBase64, _, err := loadImage("testdata/popup_risk_warning.png")
-	require.NoError(t, err)
-
-	params := struct {
-		InputImgBase64 string
-		Rect           struct {
-			X float64
-			Y float64
-		}
-		OutputPath string
-	}{
-		InputImgBase64: imageBase64,
-		Rect: struct {
-			X float64
-			Y float64
-		}{
-			X: 100,
-			Y: 100,
-		},
-		OutputPath: "testdata/output.png",
-	}
-
-	err = SavePositionImg(params)
-	require.NoError(t, err)
-
-	// cleanup
-	defer os.Remove(params.OutputPath)
-}
-
 func TestLoadImage(t *testing.T) {
 	// Test PNG image
-	pngBase64, pngSize, err := loadImage("testdata/llk_1.png")
+	pngBase64, pngSize, err := builtin.LoadImage("testdata/llk_1.png")
 	require.NoError(t, err)
 	assert.NotEmpty(t, pngBase64)
 	assert.Greater(t, pngSize.Width, 0)
 	assert.Greater(t, pngSize.Height, 0)
 
 	// Test JPEG image
-	jpegBase64, jpegSize, err := loadImage("testdata/xhs-feed.jpeg")
+	jpegBase64, jpegSize, err := builtin.LoadImage("testdata/xhs-feed.jpeg")
 	require.NoError(t, err)
 	assert.NotEmpty(t, jpegBase64)
 	assert.Greater(t, jpegSize.Width, 0)
 	assert.Greater(t, jpegSize.Height, 0)
 }
-
-// loadImage loads image and returns base64 encoded string
-func loadImage(imagePath string) (base64Str string, size types.Size, err error) {
-	// Read the image file
-	imageFile, err := os.OpenFile(imagePath, os.O_RDONLY, 0o600)
-	if err != nil {
-		return "", types.Size{}, fmt.Errorf("failed to open image file: %w", err)
-	}
-	defer imageFile.Close()
-
-	// Decode the image to get its resolution
-	imageData, format, err := image.Decode(imageFile)
-	if err != nil {
-		return "", types.Size{}, fmt.Errorf("failed to decode image: %w", err)
-	}
-
-	// Get the resolution of the image
-	width := imageData.Bounds().Dx()
-	height := imageData.Bounds().Dy()
-	size = types.Size{Width: width, Height: height}
-
-	// Convert image to base64
-	buf := new(bytes.Buffer)
-	// 根据图像格式选择正确的编码器
-	if format == "jpeg" || format == "jpg" {
-		if err := jpeg.Encode(buf, imageData, nil); err != nil {
-			return "", types.Size{}, fmt.Errorf("failed to encode image to buffer: %w", err)
-		}
-	} else {
-		// 默认使用 PNG 编码
-		if err := png.Encode(buf, imageData); err != nil {
-			return "", types.Size{}, fmt.Errorf("failed to encode image to buffer: %w", err)
-		}
-	}
-	base64Str = fmt.Sprintf("data:image/%s;base64,%s", format,
-		base64.StdEncoding.EncodeToString(buf.Bytes()))
-
-	return base64Str, size, nil
-}

uixt/ai/session.go

@@ -1,14 +1,6 @@
 package ai
 
 import (
-	"bytes"
-	"encoding/base64"
-	"fmt"
-	"image"
-	"image/color"
-	"image/draw"
-	"image/png"
-	"os"
 	"strings"
 
 	"github.com/cloudwego/eino/schema"
@@ -109,70 +101,3 @@ func logResponse(resp *schema.Message) {
 	}
 	logger.Msg("log response message")
 }
-
-// SavePositionImg saves an image with position markers
-func SavePositionImg(params struct {
-	InputImgBase64 string
-	Rect           struct {
-		X float64
-		Y float64
-	}
-	OutputPath string
-}) error {
-	// 解码Base64图像
-	imgData := params.InputImgBase64
-	// 如果包含了数据URL前缀，去掉它
-	if strings.HasPrefix(imgData, "data:image/") {
-		parts := strings.Split(imgData, ",")
-		if len(parts) > 1 {
-			imgData = parts[1]
-		}
-	}
-
-	// 解码Base64
-	unbased, err := base64.StdEncoding.DecodeString(imgData)
-	if err != nil {
-		return fmt.Errorf("无法解码Base64图像: %w", err)
-	}
-
-	// 解码图像
-	reader := bytes.NewReader(unbased)
-	img, _, err := image.Decode(reader)
-	if err != nil {
-		return fmt.Errorf("无法解码图像数据: %w", err)
-	}
-
-	// 创建一个可以在其上绘制的图像
-	bounds := img.Bounds()
-	rgba := image.NewRGBA(bounds)
-	draw.Draw(rgba, bounds, img, bounds.Min, draw.Src)
-
-	// 在点击/拖动位置绘制标记
-	markRadius := 10
-	x, y := int(params.Rect.X), int(params.Rect.Y)
-
-	// 绘制红色圆圈
-	for i := -markRadius; i <= markRadius; i++ {
-		for j := -markRadius; j <= markRadius; j++ {
-			if i*i+j*j <= markRadius*markRadius {
-				if x+i >= 0 && x+i < bounds.Max.X && y+j >= 0 && y+j < bounds.Max.Y {
-					rgba.Set(x+i, y+j, color.RGBA{255, 0, 0, 255})
-				}
-			}
-		}
-	}
-
-	// 保存图像
-	outFile, err := os.Create(params.OutputPath)
-	if err != nil {
-		return fmt.Errorf("无法创建输出文件: %w", err)
-	}
-	defer outFile.Close()
-
-	// 编码为PNG并保存
-	if err := png.Encode(outFile, rgba); err != nil {
-		return fmt.Errorf("无法编码和保存图像: %w", err)
-	}
-
-	return nil
-}

uixt/driver_ext_screenshot.go

@@ -2,8 +2,11 @@ package uixt
 
 import (
 	"bytes"
+	"encoding/base64"
 	"fmt"
 	"image"
+	"image/color"
+	"image/draw"
 	"image/gif"
 	"image/jpeg"
 	"image/png"
@@ -294,3 +297,70 @@ func compressImageBuffer(raw *bytes.Buffer) (compressed *bytes.Buffer, err error
 	// return compressed image buffer
 	return &buf, nil
 }
+
+// SavePositionImg saves an image with position markers
+func SavePositionImg(params struct {
+	InputImgBase64 string
+	Rect           struct {
+		X float64
+		Y float64
+	}
+	OutputPath string
+}) error {
+	// 解码Base64图像
+	imgData := params.InputImgBase64
+	// 如果包含了数据URL前缀，去掉它
+	if strings.HasPrefix(imgData, "data:image/") {
+		parts := strings.Split(imgData, ",")
+		if len(parts) > 1 {
+			imgData = parts[1]
+		}
+	}
+
+	// 解码Base64
+	unbased, err := base64.StdEncoding.DecodeString(imgData)
+	if err != nil {
+		return fmt.Errorf("无法解码Base64图像: %w", err)
+	}
+
+	// 解码图像
+	reader := bytes.NewReader(unbased)
+	img, _, err := image.Decode(reader)
+	if err != nil {
+		return fmt.Errorf("无法解码图像数据: %w", err)
+	}
+
+	// 创建一个可以在其上绘制的图像
+	bounds := img.Bounds()
+	rgba := image.NewRGBA(bounds)
+	draw.Draw(rgba, bounds, img, bounds.Min, draw.Src)
+
+	// 在点击/拖动位置绘制标记
+	markRadius := 30
+	x, y := int(params.Rect.X), int(params.Rect.Y)
+
+	// 绘制红色圆圈
+	for i := -markRadius; i <= markRadius; i++ {
+		for j := -markRadius; j <= markRadius; j++ {
+			if i*i+j*j <= markRadius*markRadius {
+				if x+i >= 0 && x+i < bounds.Max.X && y+j >= 0 && y+j < bounds.Max.Y {
+					rgba.Set(x+i, y+j, color.RGBA{255, 0, 0, 255})
+				}
+			}
+		}
+	}
+
+	// 保存图像
+	outFile, err := os.Create(params.OutputPath)
+	if err != nil {
+		return fmt.Errorf("无法创建输出文件: %w", err)
+	}
+	defer outFile.Close()
+
+	// 编码为PNG并保存
+	if err := png.Encode(outFile, rgba); err != nil {
+		return fmt.Errorf("无法编码和保存图像: %w", err)
+	}
+
+	return nil
+}

uixt/driver_ext_test.go

@@ -3,9 +3,11 @@
 package uixt
 
 import (
+	"os"
 	"testing"
 	"time"
 
+	"github.com/httprunner/httprunner/v5/internal/builtin"
 	"github.com/httprunner/httprunner/v5/uixt/option"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -251,3 +253,33 @@ func TestDriverExt_Action_Offset(t *testing.T) {
 		option.WithTapRandomRect(true))
 	assert.Nil(t, err)
 }
+
+func TestSavePositionImg(t *testing.T) {
+	imageBase64, _, err := builtin.LoadImage("ai/testdata/popup_risk_warning.png")
+	require.NoError(t, err)
+
+	params := struct {
+		InputImgBase64 string
+		Rect           struct {
+			X float64
+			Y float64
+		}
+		OutputPath string
+	}{
+		InputImgBase64: imageBase64,
+		Rect: struct {
+			X float64
+			Y float64
+		}{
+			X: 500,
+			Y: 500,
+		},
+		OutputPath: "ai/testdata/output.png",
+	}
+
+	err = SavePositionImg(params)
+	require.NoError(t, err)
+
+	// cleanup
+	defer os.Remove(params.OutputPath)
+}