fix: return appropriate error for failing pdf, before sending the request

Author: mpolomdeepsense

src/unstructured_client/_hooks/custom/pdf_utils.py

@@ -5,7 +5,7 @@
 from typing import cast, Optional, BinaryIO, Union
 
 from pypdf import PdfReader
-from pypdf.errors import PdfReadError
+from pypdf.errors import FileNotDecryptedError, PdfReadError
 
 from unstructured_client._hooks.custom.common import UNSTRUCTURED_CLIENT_LOGGER_NAME
 
@@ -16,6 +16,15 @@
 pdf_logger = logging.getLogger("pypdf")
 pdf_logger.setLevel(logging.ERROR)
 
+
+class PDFValidationError(Exception):
+    """Base exception for PDF validation errors."""
+
+    def __init__(self, message: str):
+        self.message = message
+        super().__init__(self.message)
+
+
 def read_pdf(pdf_file: Union[BinaryIO, bytes]) -> Optional[PdfReader]:
     """Reads the given PDF file.
 
@@ -33,3 +42,34 @@ def read_pdf(pdf_file: Union[BinaryIO, bytes]) -> Optional[PdfReader]:
         return PdfReader(pdf_file, strict=False)
     except (PdfReadError, UnicodeDecodeError):
         return None
+
+
+def check_pdf(pdf: PdfReader) -> PdfReader:
+    """
+    Check if PDF is:
+    - Encrypted
+    - Has corrupted pages
+    - Has corrupted root object
+
+    Throws:
+    - RequestError if file is encrypted or corrupted
+    """
+    try:
+        # This will raise if the file is encrypted
+        pdf.metadata
+
+        # This will raise if the file's root object is corrupted
+        pdf.root_object
+
+        # This will raise if the file's pages are corrupted
+        list(pdf.pages)
+
+        return pdf
+    except FileNotDecryptedError as e:
+        raise PDFValidationError(
+            "File is encrypted. Please decrypt it with password.",
+        ) from e
+    except PdfReadError as e:
+        raise PDFValidationError(
+            f"File does not appear to be a valid PDF. Error: {e}",
+        ) from e

src/unstructured_client/_hooks/custom/split_pdf_hook.py

@@ -15,6 +15,7 @@
 
 import aiofiles
 import httpx
+from httpx import RequestError
 import nest_asyncio  # type: ignore
 from httpx import AsyncClient
 from pypdf import PdfReader, PdfWriter
@@ -303,6 +304,14 @@ def before_request(
         if pdf is None:
             return request
 
+        try:
+            pdf = pdf_utils.check_pdf(pdf)
+        except pdf_utils.PDFValidationError as e:
+            raise RequestError(
+                message=e.message,
+                request=request,
+            ) from e
+
         starting_page_number = form_utils.get_starting_page_number(
             form_data,
             key=PARTITION_FORM_STARTING_PAGE_NUMBER_KEY,