refactor: simplify APIs, remove GetUIResultMap

hrp/pkg/uixt/ext.go

@@ -54,11 +54,13 @@ type ScreenResult struct {
 	bufSource *bytes.Buffer // raw image buffer bytes
 	imagePath string        // image file path
 
-	Texts     OCRTexts   `json:"texts"`                // dumped raw OCRTexts
+	UploadedURL string      `json:"uploaded_url"`         // uploaded image url
-	Tags      []string   `json:"tags"`                 // tags for image, e.g. ["feed", "ad", "live"]
+	Texts       OCRTexts    `json:"texts"`                // dumped raw OCRTexts
-	VideoType string     `json:"video_type,omitempty"` // video type: feed, live-preview or live
+	Icons       UIResultMap `json:"icons"`                // CV 识别的图标
-	Feed      *FeedVideo `json:"feed,omitempty"`
+	Tags        []string    `json:"tags"`                 // tags for image, e.g. ["feed", "ad", "live"]
-	Live      *LiveRoom  `json:"live,omitempty"`
+	VideoType   string      `json:"video_type,omitempty"` // video type: feed, live-preview or live
+	Feed        *FeedVideo  `json:"feed,omitempty"`
+	Live        *LiveRoom   `json:"live,omitempty"`
 
 	SwipeStartTime  int64 `json:"swipe_start_time"`  // 滑动开始时间戳
 	SwipeFinishTime int64 `json:"swipe_finish_time"` // 滑动结束时间戳
@@ -72,8 +74,7 @@ type ScreenResult struct {
 
 type cacheStepData struct {
 	// cache step screenshot paths
-	screenShots     []string
+	screenShots []string
-	screenShotsUrls map[string]string // map screenshot file path to uploaded url
 	// cache step screenshot ocr results, key is image path, value is ScreenResult
 	screenResults map[string]*ScreenResult
 	// cache feed/live video stat
@@ -82,7 +83,6 @@ type cacheStepData struct {
 
 func (d *cacheStepData) reset() {
 	d.screenShots = make([]string, 0)
-	d.screenShotsUrls = make(map[string]string)
 	d.screenResults = make(map[string]*ScreenResult)
 	d.videoCrawler = nil
 }
@@ -222,7 +222,15 @@ func (dExt *DriverExt) GetStepCacheData() map[string]interface{} {
 	cacheData := make(map[string]interface{})
 	cacheData["video_stat"] = dExt.cacheStepData.videoCrawler
 	cacheData["screenshots"] = dExt.cacheStepData.screenShots
-	cacheData["screenshots_urls"] = dExt.cacheStepData.screenShotsUrls
+
+	screenShotsUrls := make(map[string]string)
+	for imagePath, screenResult := range dExt.cacheStepData.screenResults {
+		if screenResult.UploadedURL == "" {
+			continue
+		}
+		screenShotsUrls[imagePath] = screenResult.UploadedURL
+	}
+	cacheData["screenshots_urls"] = screenShotsUrls
 
 	screenSize, err := dExt.Driver.WindowSize()
 	if err != nil {

hrp/pkg/uixt/service_vedem.go

@@ -353,7 +353,6 @@ func (dExt *DriverExt) GetScreenResult(options ...ActionOption) (screenResult *S
 		ScreenshotTakeElapsed: screenshotTakeElapsed,
 	}
 
-	var imageUrl string
 	imageResult, err := dExt.ImageService.GetImage(bufSource, options...)
 	if err != nil {
 		log.Error().Err(err).Msg("GetImage from ImageService failed")
@@ -362,11 +361,8 @@ func (dExt *DriverExt) GetScreenResult(options ...ActionOption) (screenResult *S
 	if imageResult != nil {
 		screenResult.ScreenshotCVElapsed = time.Since(startTime).Milliseconds() - screenshotTakeElapsed
 		screenResult.Texts = imageResult.OCRResult.ToOCRTexts()
-
+		screenResult.UploadedURL = imageResult.URL
-		imageUrl = imageResult.URL
+		screenResult.Icons = imageResult.UIResult
-		if imageUrl != "" {
-			dExt.cacheStepData.screenShotsUrls[imagePath] = imageUrl
-		}
 
 		if imageResult.LiveType != "" && imageResult.LiveType != "NoLive" {
 			screenResult.Live = &LiveRoom{
@@ -379,7 +375,7 @@ func (dExt *DriverExt) GetScreenResult(options ...ActionOption) (screenResult *S
 
 	log.Debug().
 		Str("imagePath", imagePath).
-		Str("imageUrl", imageUrl).
+		Str("imageUrl", screenResult.UploadedURL).
 		Int64("screenshot_take_elapsed(ms)", screenResult.ScreenshotTakeElapsed).
 		Int64("screenshot_cv_elapsed(ms)", screenResult.ScreenshotCVElapsed).
 		Msg("log screenshot")
@@ -506,36 +502,13 @@ func (u UIResults) GetUIResult(options ...ActionOption) (UIResult, error) {
 	return uiResults[idx], nil
 }
 
-func (dExt *DriverExt) GetUIResultMap(uiTypes []string) (uiResultMap UIResultMap, err error) {
-	screenResult, err := dExt.GetScreenResult()
-	if err != nil {
-		return nil, err
-	}
-	bufSource := screenResult.bufSource
-	imagePath := screenResult.imagePath
-
-	imageResult, err := dExt.ImageService.GetImage(bufSource,
-		WithScreenShotUITypes(uiTypes...)) // turn on UI type detection
-	if err != nil {
-		log.Error().Err(err).Msg("GetImage from ImageService failed")
-		return
-	}
-
-	imageUrl := imageResult.URL
-	if imageUrl != "" {
-		dExt.cacheStepData.screenShotsUrls[imagePath] = imageUrl
-		log.Debug().Str("imagePath", imagePath).Str("imageUrl", imageUrl).Msg("log screenshot")
-	}
-	uiResultMap = imageResult.UIResult
-	return
-}
-
 func (dExt *DriverExt) FindUIResult(uiTypes []string, options ...ActionOption) (point PointF, err error) {
-	uiResultMap, err := dExt.GetUIResultMap(uiTypes)
+	screenResult, err := dExt.GetScreenResult(WithScreenShotUITypes(uiTypes...))
 	if err != nil {
 		return
 	}
-	uiResults, err := uiResultMap.FilterUIResults(uiTypes)
+
+	uiResults, err := screenResult.Icons.FilterUIResults(uiTypes)
 	if err != nil {
 		return
 	}