dangerzone/dangerzone-converter/scripts/pixels-to-pdf-unpriv

#!/usr/bin/env python3
import glob
import os
import sys
import subprocess


def print_flush(s=""):
    print(s)
    sys.stdout.flush()


def main():
    num_pages = len(glob.glob("/dangerzone/page-*.rgb"))
    print_flush(f"Document has {num_pages} pages")

    # Convert RGB files to PDF files
    for page in range(1, num_pages + 1):
        filename_base = f"/dangerzone/page-{page}"
        rgb_filename = f"{filename_base}.rgb"
        width_filename = f"{filename_base}.width"
        height_filename = f"{filename_base}.height"
        png_filename = f"/tmp/page-{page}.png"
        ocr_filename = f"/tmp/page-{page}"
        pdf_filename = f"/tmp/page-{page}.pdf"

        with open(width_filename) as f:
            width = f.read().strip()
        with open(height_filename) as f:
            height = f.read().strip()

        if os.environ.get("OCR") == "1":
            # OCR the document
            print_flush(f"Converting page {page} from pixels to searchable PDF")

            args = [
                "gm",
                "convert",
                "-size",
                f"{width}x{height}",
                "-depth",
                "8",
                f"rgb:{rgb_filename}",
                f"png:{png_filename}",
            ]
            try:
                p = subprocess.run(args, timeout=60)
            except subprocess.TimeoutExpired:
                print_flush(
                    "Error converting pixels to PNG, convert timed out after 60 seconds"
                )
                sys.exit(1)
            if p.returncode != 0:
                print_flush(f"Page {page} conversion failed: {p.stdout}")
                sys.exit(1)

            args = [
                "tesseract",
                png_filename,
                ocr_filename,
                "-l",
                os.environ.get("OCR_LANGUAGE"),
                "--dpi",
                "70",
                "pdf"
            ]
            try:
                p = subprocess.run(args, timeout=60)
            except subprocess.TimeoutExpired:
                print_flush(
                    "Error converting PNG to searchable PDF, tesseract timed out after 60 seconds"
                )
                sys.exit(1)
            if p.returncode != 0:
                print_flush(f"Page {page} conversion failed: {p.stdout}")
                sys.exit(1)

        else:
            # Don't OCR
            print_flush(f"Converting page {page} from pixels to PDF")

            args = [
                "gm",
                "convert",
                "-size",
                f"{width}x{height}",
                "-depth",
                "8",
                f"rgb:{rgb_filename}",
                f"pdf:{pdf_filename}",
            ]
            try:
                p = subprocess.run(args, timeout=60)
            except subprocess.TimeoutExpired:
                print_flush(
                    "Error converting RGB to PDF, convert timed out after 60 seconds"
                )
                sys.exit(1)
            if p.returncode != 0:
                print_flush(f"Page {page} conversion failed: {p.stdout}")
                sys.exit(1)

    print_flush()

    # Merge pages into a single PDF
    print_flush(f"Merging {num_pages} pages into a single PDF")
    args = ["pdfunite"]
    for page in range(1, num_pages + 1):
        args.append(f"/tmp/page-{page}.pdf")
    args.append(f"/tmp/safe-output.pdf")
    try:
        p = subprocess.run(args, timeout=60)
    except subprocess.TimeoutExpired:
        print_flush(
            "Error merging pages into a single PDF, pdfunite timed out after 60 seconds"
        )
        sys.exit(1)
    if p.returncode != 0:
        print_flush(f"Merge failed: {p.stdout}")
        sys.exit(1)

    # Compress
    print_flush("Compressing PDF")
    compress_timeout = num_pages * 3
    try:
        p = subprocess.run(
            ["ps2pdf", "/tmp/safe-output.pdf", "/tmp/safe-output-compressed.pdf"],
            timeout=compress_timeout,
        )
    except subprocess.TimeoutExpired:
        print_flush(f"Error compressing PDF, ps2pdf timed out after {compress_timeout} seconds")
        sys.exit(1)
    if p.returncode != 0:
        print_flush(f"Compression failed: {p.stdout}")
        sys.exit(1)


if __name__ == "__main__":
    main()