Only image hash on file hash collision

main.go

@@ -86,22 +86,62 @@ func main() {
 				decor.CurrentKibiByte("%5d"),
 			),
 		)
-		done := make(chan bool)
-		wg := sync.WaitGroup{}
+		// Stage 2: compute the file hash of every file found.
+		fileHashingDone := make(chan bool)
+		fileWG := sync.WaitGroup{}
 		for i := 0; i < runtime.GOMAXPROCS(0); i++ {
-			wg.Add(2)
-			go filesMap.ImageHashingWorker(&wg)
-			go filesMap.FileHashingWorker(&wg)
+			fileWG.Add(1)
+			go filesMap.FileHashingWorker(&fileWG)
 		}
-
-		go filesMap.HashedWorker(done)
+		go filesMap.fileHashMergeWorker(fileHashingDone)
 
 		countFiles = filesMap.WalkDirectories()
 
-		wg.Wait()
+		fileWG.Wait()
 		close(filesMap.FilesHashed)
+		<-fileHashingDone
+
+		// Stage 3: image hashes are only needed for files with no file-hash
+		// collision: either they were never file-hashed (unique size) or their
+		// file hash did not match any other file.
+		var imageCandidates []*imageEntry
+
+		// Files whose size is unique were never file-hashed, so they have no
+		// collision but may still be visually similar images.
+		for size, file := range filesMap.FilesBySize {
+			if file == "" {
+				continue
+			}
+			imageCandidates = append(imageCandidates, &imageEntry{path: file, size: size})
+		}
+
+		// Files that were file-hashed without colliding also need an image hash.
+		for _, files := range filesMap.FilesByHash {
+			if len(files) != 1 {
+				continue
+			}
+			imageCandidates = append(imageCandidates, &imageEntry{path: files[0], size: filesMap.fileSizes[files[0]]})
+		}
+
+		// Stage 4: compute the image hash of every candidate.
+		go func() {
+			for _, candidate := range imageCandidates {
+				filesMap.ImagesHashing <- candidate
+			}
+			close(filesMap.ImagesHashing)
+		}()
+
+		imageWG := sync.WaitGroup{}
+		imageHashingDone := make(chan bool)
+		for i := 0; i < runtime.GOMAXPROCS(0); i++ {
+			imageWG.Add(1)
+			go filesMap.ImageHashingWorker(&imageWG)
+		}
+		go filesMap.imageHashMergeWorker(imageHashingDone)
+
+		imageWG.Wait()
 		close(filesMap.ImagesHashed)
-		<-done
+		<-imageHashingDone
 	}
 
 	if *toFile != "" && *fromFile == "" {