Merge pull request #42 from eSlider/feat/docs-optimize-gs

feat(docs): Ghostscript PDF optimize for OO

cmd/oo/docs.go

@@ -29,6 +29,7 @@ OCR a scan locally:       oo docs ocr scan.pdf --md out.md
 Structured OCR (hOCR→MD): oo docs hocr scan.jpg --md out.md --yaml out.yml`,
 	}
 	cmd.AddCommand(docsConvertCmd())
+	cmd.AddCommand(docsOptimizeCmd())
 	cmd.AddCommand(docsOCRCmd())
 	cmd.AddCommand(docsHOCRCmd())
 	cmd.AddCommand(docsAsMDCmd())
@@ -49,6 +50,7 @@ func docsToolsCmd() *cobra.Command {
 				"ocrmypdf":    strOrNil(t.OCRMyPDF),
 				"pdftotext":   strOrNil(t.PDFToText),
 				"tesseract":   strOrNil(t.Tesseract),
+				"ghostscript": strOrNil(t.Ghostscript),
 			})
 			return nil
 		},
@@ -96,6 +98,43 @@ func docsConvertCmd() *cobra.Command {
 	return cmd
 }
 
+func docsOptimizeCmd() *cobra.Command {
+	var out string
+	cmd := &cobra.Command{
+		Use:   "optimize PDF_PATH",
+		Short: "Rewrite PDF via Ghostscript (PostScript pdfwrite) for OO-friendly size/text",
+		Long: `Use for InDesign/iText PDFs with a good text layer — avoids ocrmypdf invisible
+text overlays that break OnlyOffice DocEditor. Skips OCR; rewrites via gs pdfwrite.`,
+		Args: cobra.ExactArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			in := args[0]
+			t := docpipe.LookPath()
+			if out == "" {
+				base := strings.TrimSuffix(filepath.Base(in), filepath.Ext(in))
+				out = filepath.Join(filepath.Dir(in), base+".optimized.pdf")
+			}
+			if err := docpipe.EnsureDir(out); err != nil {
+				return err
+			}
+			chars, _ := t.PDFTextLayerChars(in)
+			if err := t.OptimizePDF(in, out); err != nil {
+				return err
+			}
+			outChars, _ := t.PDFTextLayerChars(out)
+			printObject(map[string]any{
+				"in":              in,
+				"pdf":             out,
+				"text_chars_in":   chars,
+				"text_chars_out":  outChars,
+				"note":            "native text layer preserved; no OCR overlay",
+			})
+			return nil
+		},
+	}
+	cmd.Flags().StringVar(&out, "out", "", "output PDF (default: <name>.optimized.pdf)")
+	return cmd
+}
+
 func docsOCRCmd() *cobra.Command {
 	var out, mdOut, lang string
 	var force bool
@@ -140,7 +179,7 @@ func docsOCRCmd() *cobra.Command {
 	cmd.Flags().StringVar(&mdOut, "md", "", "write Markdown extraction to this path")
 	cmd.Flags().BoolVar(&writeMD, "markdown", false, "also write sibling .md next to OCR PDF")
 	cmd.Flags().StringVar(&lang, "lang", "eng", "OCR language(s) for tesseract/ocrmypdf")
-	cmd.Flags().BoolVar(&force, "force", true, "force OCR even if a text layer exists")
+	cmd.Flags().BoolVar(&force, "force", false, "force OCR even if a text layer exists (default: skip when pdftotext finds enough text)")
 	return cmd
 }
 

internal/docpipe/docpipe.go

@@ -6,6 +6,7 @@
 //   - ocrmypdf   — OCR into a searchable PDF
 //   - pdftotext  — extract text layer
 //   - tesseract  — OCR single images when ocrmypdf is unsuitable
+//   - ghostscript (gs) — PDF rewrite/optimize via PostScript (pdfwrite)
 package docpipe
 
 import (
@@ -26,6 +27,7 @@ type Tools struct {
 	OCRMyPDF    string
 	PDFToText   string
 	Tesseract   string
+	Ghostscript string
 }
 
 // LookPath resolves converter binaries (empty string if missing).
@@ -43,6 +45,7 @@ func LookPath() Tools {
 		OCRMyPDF:    find("ocrmypdf"),
 		PDFToText:   find("pdftotext"),
 		Tesseract:   find("tesseract"),
+		Ghostscript: find("gs", "ghostscript"),
 	}
 }
 
@@ -170,6 +173,39 @@ func (t Tools) DOCXToMD(docxPath, mdPath string) error {
 	return t.ConvertFile(docxPath, mdPath)
 }
 
+// OptimizePDF rewrites a PDF through Ghostscript (PostScript pdfwrite).
+// Preserves native text layers; strips broken OCR overlays; shrinks for OO preview.
+// Use instead of ocrmypdf when pdftotext already extracts enough text.
+func (t Tools) OptimizePDF(inPath, outPath string) error {
+	if t.Ghostscript == "" {
+		return fmt.Errorf("ghostscript (gs) not found on PATH")
+	}
+	if outPath == "" {
+		return fmt.Errorf("output PDF path required")
+	}
+	args := []string{
+		"-sDEVICE=pdfwrite",
+		"-dCompatibilityLevel=1.5",
+		"-dNOPAUSE", "-dQUIET", "-dBATCH",
+		"-dPDFSETTINGS=/ebook",
+		"-dEmbedAllFonts=true",
+		"-dSubsetFonts=true",
+		"-dCompressFonts=true",
+		"-dCompressPages=true",
+		"-dDetectDuplicateImages=true",
+		"-dAutoRotatePages=/None",
+		"-sOutputFile=" + outPath,
+		inPath,
+	}
+	cmd := exec.Command(t.Ghostscript, args...)
+	var stderr bytes.Buffer
+	cmd.Stderr = &stderr
+	if err := cmd.Run(); err != nil {
+		return fmt.Errorf("ghostscript pdfwrite: %w (%s)", err, strings.TrimSpace(stderr.String()))
+	}
+	return nil
+}
+
 // PDFTextLayerChars returns approximate extracted character count (0 if unavailable).
 func (t Tools) PDFTextLayerChars(pdfPath string) (int, error) {
 	if t.PDFToText == "" {

internal/docpipe/docpipe_test.go

@@ -128,6 +128,33 @@ func TestMDDocxRoundTrip(t *testing.T) {
 	}
 }
 
+func TestOptimizePDF(t *testing.T) {
+	tools := LookPath()
+	if tools.Ghostscript == "" || tools.PDFToText == "" {
+		t.Skip("ghostscript/pdftotext not installed")
+	}
+	in := "/tmp/ccgg-original.pdf"
+	if _, err := os.Stat(in); err != nil {
+		t.Skip("local fixture not present")
+	}
+	dir := t.TempDir()
+	out := filepath.Join(dir, "out.pdf")
+	charsIn, err := tools.PDFTextLayerChars(in)
+	if err != nil || charsIn < 1000 {
+		t.Skip("fixture has no text layer")
+	}
+	if err := tools.OptimizePDF(in, out); err != nil {
+		t.Fatal(err)
+	}
+	charsOut, err := tools.PDFTextLayerChars(out)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if charsOut < charsIn/2 {
+		t.Fatalf("text layer lost: in=%d out=%d", charsIn, charsOut)
+	}
+}
+
 func TestToMarkdown_PlainMD(t *testing.T) {
 	tools := LookPath()
 	dir := t.TempDir()