fix MRR

ispras · Aug 8, 2024 · 70c78fd · 70c78fd
1 parent 3334124
commit 70c78fd
Show file tree

Hide file tree

Showing 4 changed files with 12 additions and 4 deletions.
diff --git a/dedoc/readers/pdf_reader/pdf_auto_reader/txtlayer_detector.py b/dedoc/readers/pdf_reader/pdf_auto_reader/txtlayer_detector.py
@@ -31,7 +31,8 @@ def detect_txtlayer(self, path: str, parameters: dict) -> PdfTxtlayerParameters:
             lines = self.__get_lines_for_predict(path=path, parameters=parameters)
             if str(parameters.get("fast_textual_layer_detection", "false")).lower() == "true":
                 is_correct = any(line.line.strip() for line in lines)
-                first_page_correct = True if len([line for line in lines if line.metadata.page_id == 0]) > 0 else False
+                first_page_lines = [line for line in lines if line.metadata.page_id == 0]
+                first_page_correct = first_page_lines and any(line.line.strip() for line in first_page_lines)
             else:
                 is_correct = self.txtlayer_classifier.predict(lines)
                 first_page_correct = self.__is_first_page_correct(lines=lines, is_txt_layer_correct=is_correct)

diff --git a/docs/source/dedoc_api_usage/api.rst b/docs/source/dedoc_api_usage/api.rst
@@ -210,7 +210,7 @@ Api parameters description
               If the document doesn't have a textual layer (it is an image, scanned document), PDF document parsing works like with ``need_pdf_table_analysis=false``.
               It is highly recommended to use this option value for any PDF document parsing.
 
-    * - fast_auto
+    * - fast_textual_layer_detection
       - true, false
       - false
       - Enable fast textual layer detection. Works only when **auto** or **auto_tabby** is selected at **pdf_with_text_layer**.

diff --git a/docs/source/parameters/pdf_handling.rst b/docs/source/parameters/pdf_handling.rst
@@ -47,7 +47,7 @@ PDF and images handling
               If the document doesn't have a textual layer (it is an image, scanned document), :class:`dedoc.readers.PdfImageReader` will be used.
               It is highly recommended to use this option value for any PDF document parsing.
 
-    * - fast_auto
+    * - fast_textual_layer_detection
       - true, false
       - false
       - * :meth:`dedoc.readers.PdfAutoReader.read`

diff --git a/tests/api_tests/test_api_format_pdf_auto_text_layer.py b/tests/api_tests/test_api_format_pdf_auto_text_layer.py
@@ -86,8 +86,15 @@ def test_fast_textual_layer_detection(self) -> None:
         self.assertIn("Assume document has a correct textual layer", warnings)
         self.assertEqual(result["content"]["structure"]["subparagraphs"][5]["text"][:10], "This paper")
 
+        file_name = "tz_scan_1page.pdf"
         parameters = dict(pdf_with_text_layer="auto_tabby", fast_textual_layer_detection=True)
         result = self._send_request(file_name, parameters)
         warnings = result["warnings"]
+        self.assertIn("Assume document has incorrect textual layer", result["warnings"])
+
+        file_name = "mixed_pdf.pdf"
+        parameters = dict(pdf_with_text_layer="auto", fast_textual_layer_detection=True)
+        result = self._send_request(file_name, parameters)
+        warnings = result["warnings"]
         self.assertIn("Assume document has a correct textual layer", warnings)
-        self.assertEqual(result["content"]["structure"]["subparagraphs"][5]["text"][:10], "This paper")
+        self.assertIn("Assume the first page hasn't a textual layer", result["warnings"])