rescribe: handle PDF errors much more gracefully

2 files changed, 46 insertions, 1 deletions

diff --git a/cmd/rescribe/gui.go b/cmd/rescribe/gui.go
index fb0450a..7ae5465 100644
--- a/cmd/rescribe/gui.go
+++ b/cmd/rescribe/gui.go
@@ -298,7 +298,38 @@ func startGui(log log.Logger, cmd string, training string, tessdir string) error
 			}
 		}()
 
-		err = startProcess(log, cmd, dir.Text, filepath.Base(dir.Text), trainingOpts.Selected, dir.Text, tessdir)
+		bookdir := dir.Text
+		savedir := dir.Text
+		bookname := filepath.Base(dir.Text)
+
+		f, err := os.Stat(bookdir)
+		if err != nil {
+			// TODO: surface error and cancel process better
+			fmt.Fprintf(os.Stderr, "Error opening file as PDF: %v\n", err)
+			return
+		}
+
+		if strings.HasSuffix(dir.Text, ".pdf") && !f.IsDir() {
+			bookdir, err = extractPdfImgs(bookdir)
+			if err != nil {
+				// TODO: surface error and cancel process better
+				fmt.Fprintf(os.Stderr, "Error opening file as PDF: %v\n", err)
+				return
+			}
+
+			// happens if extractPdfImgs recovers from a PDF panic,
+			// which will occur if we encounter an image we can't decode
+			if bookdir == "" {
+				// TODO: surface error and cancel process better
+				fmt.Fprintf(os.Stderr, "Error opening PDF\nThe format of this PDF is not supported, extract the images manually into a folder first.\n")
+				return
+			}
+
+			savedir = strings.TrimSuffix(savedir, ".pdf")
+			bookname = strings.TrimSuffix(bookname, ".pdf")
+		}
+
+		err = startProcess(log, cmd, bookdir, bookname, trainingOpts.Selected, savedir, tessdir)
 		if err != nil {
 			// add a newline before this printing as another message from stdout
 			// or stderr may well be half way through printing
diff --git a/cmd/rescribe/main.go b/cmd/rescribe/main.go
index 72a03d3..019d038 100644
--- a/cmd/rescribe/main.go
+++ b/cmd/rescribe/main.go
@@ -273,6 +273,11 @@ These training files are included in rescribe, and are always available:
 		if err != nil {
 			log.Fatalln("Error opening file as PDF:", err)
 		}
+		// if this occurs then extractPdfImgs() will have recovered from
+		// a panic in the pdf package
+		if bookdir == "" {
+			log.Fatalln("Error opening file as PDF: image type not supported, you will need to extract images manually.")
+		}
 
 		bookname = strings.TrimSuffix(bookname, ".pdf")
 
@@ -299,6 +304,15 @@ These training files are included in rescribe, and are always available:
 // extractPdfImgs extracts all images embedded in a PDF to a
 // temporary directory, which is returned on success.
 func extractPdfImgs(path string) (string, error) {
+	defer func() {
+		// unfortunately the pdf library will panic if it sees an encoding
+		// it can't decode, so recover from that and give a warning
+		r := recover()
+		if r != nil {
+			fmt.Fprintf(os.Stderr, "Warning: Error extracting from PDF: %v\n", r)
+		}
+	}()
+
 	p, err := pdf.Open(path)
 	if err != nil {
 		return "", err