Test markitdown-ocr in CI against the sibling markitdown (#2378)

* Test markitdown-ocr in CI against the sibling markitdown
* Run the full matrix.

Author: afourney

.github/workflows/tests.yml

@@ -1,18 +1,45 @@
 name: tests
 on: [pull_request]
+permissions:
+  contents: read
 
 jobs:
   tests:
     runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
+
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
       - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
         with:
-          python-version: |
-            3.10
-            3.11
-            3.12
+          python-version: ${{ matrix.python-version }}
+
       - name: Install Hatch
         run: pipx install hatch
       - name: Run tests
         run: cd packages/markitdown; hatch test
+
+  ocr-tests:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
+
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      # Install markitdown from the local checkout rather than PyPI, so the
+      # plugin is tested against the sibling package it ships alongside.
+      - name: Install packages
+        run: pip install ./packages/markitdown ./packages/markitdown-ocr pytest
+
+      - name: Run tests
+        working-directory: packages/markitdown-ocr
+        run: pytest

packages/markitdown-ocr/tests/test_pdf_converter.py

@@ -144,17 +144,30 @@ def test_pdf_complex_layout(svc: MockOCRService) -> None:
 
 
 # ---------------------------------------------------------------------------
-# pdf_multipage.pdf — pdfplumber/pdfminer fail (EOF); PyMuPDF fallback used
+# pdf_multipage.pdf
 # ---------------------------------------------------------------------------
 
 
 def test_pdf_multipage(svc: MockOCRService) -> None:
-    # pdfplumber cannot open this file (Unexpected EOF), so _ocr_full_pages
-    # falls back to PyMuPDF for page rendering.  Each page becomes one OCR block.
     expected = (
-        f"## Page 1\n\n\n{_OCR_BLOCK}\n\n\n"
-        f"## Page 2\n\n\n{_OCR_BLOCK}\n\n\n"
-        f"## Page 3\n\n\n{_OCR_BLOCK}"
+        "## Page 1\n\n\n"
+        "Page 1 - Content before image\n\n"
+        "This is important text that appears BEFORE the image.\n\n\n\n"
+        "*[Image OCR]\nMOCK_OCR_TEXT_12345\n[End OCR]*\n\n\n"
+        "This text appears AFTER the image on page 1.\n\n"
+        "More content follows here.\n\n\n"
+        "## Page 2\n\n\n"
+        "Page 2 - Content with image at end\n\n"
+        "Main content of page 2 starts here.\n\n"
+        "This is paragraph 1.\n\n"
+        "This is paragraph 2.\n\n"
+        "Final paragraph before image.\n\n\n\n"
+        "*[Image OCR]\nMOCK_OCR_TEXT_12345\n[End OCR]*\n\n\n\n"
+        "## Page 3\n\n\n"
+        "Page 3 - Image at top\n\n\n\n"
+        "*[Image OCR]\nMOCK_OCR_TEXT_12345\n[End OCR]*\n\n\n"
+        "Content that follows the image.\n\n"
+        "This text is AFTER the image."
     )
     assert _convert("pdf_multipage.pdf", svc) == expected
 

packages/markitdown/src/markitdown/__main__.py

@@ -219,7 +219,7 @@ def main():
     elif args.use_cu:
         if args.cu_endpoint is None:
             _exit_with_error(
-                "Content Understanding Endpoint (--cu-endpoint) is required when using --use-cu."
+                "Content Understanding Endpoint (--cu-endpoint) is required when using --use-cu. "
                 "Pass --cu-endpoint or set MARKITDOWN_CU_ENDPOINT."
             )
 

packages/markitdown/src/markitdown/converter_utils/docx/pre_process.py

@@ -114,6 +114,11 @@ def _pre_process_strike(content: bytes) -> bytes:
     Returns:
         bytes: The processed content with "dstrike" elements renamed to "strike", encoded as bytes.
     """
+    # Double strikethrough is rare, and parsing/reserializing the XML is expensive
+    # on large documents, so skip the round-trip when there is nothing to rename.
+    if b"dstrike" not in content:
+        return content
+
     soup = BeautifulSoup(content.decode(), features="xml")
     for tag in soup.find_all("dstrike"):
         tag.name = "strike"