Introduce page-level error checks

Signed-off-by: Christoph Auer <[email protected]>

docling/backend/abstract_backend.py

@@ -30,6 +30,10 @@ def get_page_image(
     def get_size(self) -> "PageSize":
         pass
 
+    @abstractmethod
+    def is_valid(self) -> bool:
+        pass
+
     @abstractmethod
     def unload(self):
         pass

docling/backend/docling_parse_backend.py

@@ -20,16 +20,18 @@ def __init__(
         self, parser: pdf_parser, document_hash: str, page_no: int, page_obj: PdfPage
     ):
         self._ppage = page_obj
-
         parsed_page = parser.parse_pdf_from_key_on_page(document_hash, page_no)
 
         self._dpage = None
-        self.broken_page = "pages" not in parsed_page
-        if not self.broken_page:
+        self.valid = "pages" in parsed_page
+        if self.valid:
             self._dpage = parsed_page["pages"][0]
 
+    def is_valid(self) -> bool:
+        return self.valid
+
     def get_text_in_rect(self, bbox: BoundingBox) -> str:
-        if self.broken_page:
+        if not self.valid:
             return ""
         # Find intersecting cells on the page
         text_piece = ""
@@ -65,7 +67,7 @@ def get_text_cells(self) -> Iterable[Cell]:
         cells = []
         cell_counter = 0
 
-        if self.broken_page:
+        if not self.valid:
             return cells
 
         page_size = self.get_size()

docling/backend/pypdfium2_backend.py

@@ -16,6 +16,10 @@ class PyPdfiumPageBackend(PdfPageBackend):
     def __init__(self, pdfium_doc: pdfium.PdfDocument, page_no: int):
         self._ppage: pdfium.PdfPage = pdfium_doc[page_no]
         self.text_page = None
+        self.valid = True
+
+    def is_valid(self) -> bool:
+        return self.valid
 
     def get_bitmap_rects(self, scale: int = 1) -> Iterable[BoundingBox]:
         AREA_THRESHOLD = 32 * 32