feat: put pdfium call behind a threadlock (#4211)

[pdfium is not thread
safe](https://groups.google.com/g/pdfium/c/HeZSsM_KEUk?pli=1) so this PR
put it behind a thread lock for thread safety.

Author: badGarnet

CHANGELOG.md

@@ -1,3 +1,8 @@
+## 0.18.32
+
+### Enhancements
+- put `pdfium` calls behind a thread lock
+
 ## 0.18.31
 
 ### Enhancements
@@ -15,7 +20,7 @@
 - **Fix `coordinates=True` causing TypeError in hi_res PDF processing**: Filter out `coordinates` and `coordinate_system` from kwargs before passing to `add_element_metadata()` to prevent conflict with explicit parameters (fixes #4126)
 - **Preserve line breaks in code blocks during chunking**: `<pre>` elements now generate `CodeSnippet` elements instead of `Text`, and chunking preserves internal whitespace for code snippets. (fixes #4095)
 
-## 0.18.30 
+## 0.18.30
 
 ### Enhancements
 - Updated the Dockerfile to build from the chainguard base. Implemented updating and added base-packages that was done in the base-images repo to instead all be done here.

test_unstructured/partition/pdf_image/test_pdf_image_utils.py

@@ -62,6 +62,45 @@ def test_convert_pdf_to_image(file_mode, path_only):
             assert isinstance(images[0], PILImg.Image)
 
 
+@pytest.mark.parametrize("file_mode", ["filename", "rb"])
+@pytest.mark.parametrize("path_only", [True, False])
+def test_convert_pdf_to_image_twice(file_mode, path_only):
+    filename = example_doc_path("pdf/embedded-images.pdf")
+    with tempfile.TemporaryDirectory() as tmpdir:
+        if file_mode == "filename":
+            images = pdf_image_utils.convert_pdf_to_image(
+                filename=filename,
+                file=None,
+                output_folder=tmpdir,
+                path_only=path_only,
+            )
+            images = pdf_image_utils.convert_pdf_to_image(
+                filename=filename,
+                file=None,
+                output_folder=tmpdir,
+                path_only=path_only,
+            )
+        else:
+            with open(filename, "rb") as f:
+                images = pdf_image_utils.convert_pdf_to_image(
+                    filename="",
+                    file=f,
+                    output_folder=tmpdir,
+                    path_only=path_only,
+                )
+                images = pdf_image_utils.convert_pdf_to_image(
+                    filename="",
+                    file=f,
+                    output_folder=tmpdir,
+                    path_only=path_only,
+                )
+
+        if path_only:
+            assert isinstance(images[0], str)
+        else:
+            assert isinstance(images[0], PILImg.Image)
+
+
 def test_convert_pdf_to_image_raises_error():
     filename = example_doc_path("embedded-images.pdf")
     with pytest.raises(ValueError) as exc_info:

unstructured/__version__.py

@@ -1 +1 @@
-__version__ = "0.18.31"  # pragma: no cover
+__version__ = "0.18.32"  # pragma: no cover

unstructured/partition/pdf_image/pdf_image_utils.py

@@ -8,6 +8,7 @@
 from copy import deepcopy
 from io import BytesIO
 from pathlib import Path, PurePath
+from threading import Lock
 from typing import IO, TYPE_CHECKING, BinaryIO, Iterator, List, Optional, Tuple, Union, cast
 
 import cv2
@@ -29,6 +30,9 @@
     from unstructured.documents.elements import Element
 
 
+_pdfium_lock = Lock()
+
+
 def write_image(image: Union[Image.Image, np.ndarray], output_image_path: str):
     """
     Write an image to a specified file path, supporting both PIL Image and numpy ndarray formats.
@@ -69,42 +73,43 @@ def _render_pdf_pages(
     if path_only and not output_folder:
         raise ValueError("output_folder must be specified if path_only is True")
     exactly_one(filename=filename, file=file)
-    pdf = pdfium.PdfDocument(filename or file, password=password)
-    try:
-        images: dict[int, Image.Image] = {}
-        if dpi is None:
-            dpi = env_config.PDF_RENDER_DPI
-        scale = dpi / 72.0
-        for i, page in enumerate(pdf, start=1):
-            if first_page is not None and i < first_page:
-                continue
-            if last_page is not None and i > last_page:
-                break
-            bitmap = page.render(
-                scale=scale,
-                no_smoothtext=False,
-                no_smoothimage=False,
-                no_smoothpath=False,
-                optimize_mode="print",
-            )
-            try:
-                images[i] = bitmap.to_pil()
-            finally:
-                bitmap.close()
-        if not output_folder:
-            return list(images.values())
-        else:
-            # Save images to output_folder
-            filenames: list[str] = []
-            assert Path(output_folder).exists()
-            assert Path(output_folder).is_dir()
-            for i, image in images.items():
-                fn: str = os.path.join(str(output_folder), f"page_{i}.png")
-                image.save(fn, format="PNG", compress_level=1, optimize=False)
-                filenames.append(fn)
-            return filenames if path_only else list(images.values())
-    finally:
-        pdf.close()
+    with _pdfium_lock:
+        pdf = pdfium.PdfDocument(filename or file, password=password)
+        try:
+            images: dict[int, Image.Image] = {}
+            if dpi is None:
+                dpi = env_config.PDF_RENDER_DPI
+            scale = dpi / 72.0
+            for i, page in enumerate(pdf, start=1):
+                if first_page is not None and i < first_page:
+                    continue
+                if last_page is not None and i > last_page:
+                    break
+                bitmap = page.render(
+                    scale=scale,
+                    no_smoothtext=False,
+                    no_smoothimage=False,
+                    no_smoothpath=False,
+                    optimize_mode="print",
+                )
+                try:
+                    images[i] = bitmap.to_pil()
+                finally:
+                    bitmap.close()
+            if not output_folder:
+                return list(images.values())
+            else:
+                # Save images to output_folder
+                filenames: list[str] = []
+                assert Path(output_folder).exists()
+                assert Path(output_folder).is_dir()
+                for i, image in images.items():
+                    fn: str = os.path.join(str(output_folder), f"page_{i}.png")
+                    image.save(fn, format="PNG", compress_level=1, optimize=False)
+                    filenames.append(fn)
+                return filenames if path_only else list(images.values())
+        finally:
+            pdf.close()
 
 
 def convert_pdf_to_image(