Add convert_file_to_pdf helper using libreoffice (#670)

* Add doc_to_pdf helper using libreoffice

* Some cleanup

* Lint

* Move to utils package

* Add ci_only tests

* Revert "Add ci_only tests"

This reverts commit 14917d9.

* Use shutil to ensure libreoffice is available

.github/workflows/testing.yml

@@ -63,8 +63,8 @@ jobs:
         run: df
       - name: Update Apt
         run: sudo apt-get update
-      - name: Install poppler and tesseract
-        run: sudo apt-get install -y poppler-utils tesseract-ocr
+      - name: Install apt dependencies
+        run: sudo apt-get install -y poppler-utils tesseract-ocr libreoffice
       - name: DF-5
         run: df
       - name: Run tests

lib/sycamore/sycamore/tests/unit/utils/test_fileformat_tools.py

@@ -0,0 +1,27 @@
+import logging
+import os
+import shutil
+from io import BytesIO
+
+from pypdf import PdfReader
+
+import sycamore
+from sycamore.tests.config import TEST_DIR
+from sycamore.utils.fileformat_tools import binary_representation_to_pdf
+
+
+def test_binary_representation_to_pdf():
+    # Run this test locally only if libreoffice is installed
+    if shutil.which("libreoffice") is None:
+        assert "GITHUB_ACTIONS" not in os.environ
+        logging.warning("Skipping test ...; /usr/bin/libreoffice is not installed")
+        return
+    paths = str(TEST_DIR / "resources/data/docx/aryn_website_sample.docx")
+
+    context = sycamore.init()
+    doc = context.read.binary(paths, binary_format="docx").take(1)[0]
+    result = binary_representation_to_pdf(doc)
+
+    pdf_bytes = BytesIO(result.binary_representation)
+    reader = PdfReader(pdf_bytes)
+    assert len(reader.pages) == 2

lib/sycamore/sycamore/utils/fileformat_tools.py

@@ -0,0 +1,44 @@
+import subprocess
+from pathlib import Path
+from tempfile import NamedTemporaryFile
+
+from sycamore.data import Document
+
+
+def binary_representation_to_pdf(doc: Document) -> Document:
+    """
+    Utility to convert binary_representations into different file formats. Uses LibreOffice as the conversion engine.
+
+    Note: LibreOffice currently requires manual installation based on your platform.
+    """
+
+    def run_libreoffice(source_path, output_path):
+        subprocess.run(
+            [
+                "libreoffice",
+                "--headless",
+                "--convert-to",
+                "pdf",
+                source_path,
+                "--outdir",
+                output_path,
+            ]
+        )
+
+    assert doc.binary_representation is not None
+
+    extension = Path(doc.properties.get("path", "unknown")).suffix
+
+    with NamedTemporaryFile(suffix=f"{extension}") as temp_file:
+        temp_file.write(doc.binary_representation)
+        temp_file.flush()
+
+        temp_path = Path(temp_file.name)
+
+        run_libreoffice(temp_path, temp_path.parent)
+
+        with open(f"{temp_path.parent}/{temp_path.stem}.pdf", "rb") as processed_file:
+            doc.binary_representation = processed_file.read()
+            doc.properties["filetype"] = "application/pdf"
+
+    return doc