fix: removing control characters (#66)

Tesseract is putting some control characters in out_text, this commit just delete all of them

* Style correction

* Version sync

Author: benjats07

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 0.2.10
+
+* Removed control characters from tesseract output
+
 ## 0.2.9
 
 * Removed multithreading from OCR (DocumentLayout.get_elements_from_layout)

test_unstructured_inference/inference/test_layout.py

@@ -327,3 +327,10 @@ def test_from_file_fixed_layout(fixed_layouts, called_method, not_called_method)
 def test_invalid_ocr_strategy_raises(mock_image):
     with pytest.raises(ValueError):
         layout.PageLayout(0, mock_image, MockLayout(), ocr_strategy="fake_strategy")
+
+
+@pytest.mark.parametrize(
+    ("text", "expected"), [("a\ts\x0cd\nfas\fd\rf\b", "asdfasdf"), ("\"'\\", "\"'\\")]
+)
+def test_remove_control_characters(text, expected):
+    assert layout.remove_control_characters(text) == expected

unstructured_inference/__version__.py

@@ -1 +1 @@
-__version__ = "0.2.9"  # pragma: no cover
+__version__ = "0.2.10"  # pragma: no cover

unstructured_inference/inference/layout.py

@@ -5,7 +5,7 @@
 import tempfile
 from tqdm import tqdm
 from typing import List, Optional, Tuple, Union, BinaryIO
-
+import unicodedata
 from layoutparser.io.pdf import load_pdf
 from layoutparser.elements.layout_elements import TextBlock
 from layoutparser.elements.layout import Layout
@@ -318,6 +318,7 @@ def interpret_text_block(
         out_text = ocr(text_block, image)
     else:
         out_text = "" if text_block.text is None else text_block.text
+    out_text = remove_control_characters(out_text)
     return out_text
 
 
@@ -329,3 +330,9 @@ def ocr(text_block: TextBlock, image: Image.Image) -> str:
     padded_block = text_block.pad(left=5, right=5, top=5, bottom=5)
     cropped_image = padded_block.crop_image(image_array)
     return tesseract.ocr_agent.detect(cropped_image)
+
+
+def remove_control_characters(text: str) -> str:
+    """Removes control characters from text."""
+    out_text = "".join(c for c in text if unicodedata.category(c)[0] != "C")
+    return out_text