File ext + type detection improvements.

Author: vladd-bit

ocr_service/processor/converter.py

@@ -260,7 +260,7 @@ def _preprocess_doc(self, stream: bytes, file_name: str) -> bytes:
                             os.fsync(tmp_pdf_file.fileno())
                             pdf_stream = tmp_pdf_file.read()
 
-                            if b"%PDF-" not in pdf_stream[:64]:
+                            if not pdf_stream.startswith(b"%PDF-"):
                                 self.log.warning("invalid pdf header for file %s", pdf_file_path)
                                 pdf_stream = b""
                     else:

ocr_service/processor/processor.py

@@ -42,7 +42,7 @@ def _process(self, stream: bytes, file_name: str) -> tuple[str, dict]:
         """
 
         file_type = detect_file_type(stream)
-        file_name = normalise_file_name_with_ext(file_name, stream)
+        file_name = normalise_file_name_with_ext(file_name, stream, file_type)
         ctx = ProcessContext(stream=stream, file_name=file_name, file_type=file_type)
         ctx.metadata["content-type"] = self.converter.resolve_content_type(file_type)
 

ocr_service/tests/test_filename_handling.py

@@ -14,6 +14,14 @@ def test_unknown_binary_name_remains_extensionless(self):
         file_name = normalise_file_name_with_ext("request-id", b"\x00\x01\x02\x03")
         self.assertEqual(file_name, "request-id")
 
+    def test_detected_file_type_extension_is_used_when_name_has_no_suffix(self):
+        file_type = Mock()
+        file_type.extension = "docx"
+
+        file_name = normalise_file_name_with_ext("request-id", b"\x00\x01\x02\x03", file_type)
+
+        self.assertEqual(file_name, "request-id.docx")
+
     def test_processor_passes_extensionless_unknown_name_to_converter(self):
         processor = Processor()
         processor.converter = Mock()

ocr_service/utils/utils.py

@@ -335,7 +335,7 @@ def detect_file_type(stream: bytes) -> object | None:
     return file_type
 
 
-def normalise_file_name_with_ext(file_name: str, stream: bytes) -> str:
+def normalise_file_name_with_ext(file_name: str, stream: bytes, file_type: object | None = None) -> str:
     """Normalize filename and ensure an extension is present.
 
     LibreOffice relies on a reasonable filename with an extension to select
@@ -346,6 +346,7 @@ def normalise_file_name_with_ext(file_name: str, stream: bytes) -> str:
     Args:
         file_name: Original file name (may be empty or extension-less).
         stream: File content used for extension inference.
+        file_type: Optional previously detected file type descriptor.
 
     Returns:
         str: Normalized file name with an extension.
@@ -361,20 +362,25 @@ def normalise_file_name_with_ext(file_name: str, stream: bytes) -> str:
     if ext:
         return base + ext
 
-    # 2) let filetype guess it from content
+    # 2) prefer an already detected extension when available
+    detected_ext = getattr(file_type, "extension", None)
+    if detected_ext:
+        return f"{base}.{str(detected_ext)}"
+
+    # 3) let filetype guess it from content
     guessed_ext = filetype.guess_extension(stream)
     if guessed_ext:
         return f"{base}.{guessed_ext}"
 
-    # 3) fallbacks for texty formats our filetype may not catch
+    # 4) fallbacks for texty formats our filetype may not catch
     if is_file_type_html(stream):
         return base + ".html"
     if is_file_type_xml(stream):
         return base + ".xml"
     if is_file_type_rtf(stream):
         return base + ".rtf"
 
-    # 4) only tag as plain text when the content actually looks like text
+    # 5) only tag as plain text when the content actually looks like text
     if is_file_content_plain_text(stream):
         return base + ".txt"