Escape HTML output

aws-samples · Aug 16, 2024 · a833acf · a833acf
2 parents 3fbe596 + 00023b1
commit a833acf
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 5 deletions.
diff --git a/textractor/entities/line.py b/textractor/entities/line.py
@@ -15,7 +15,8 @@
 from textractor.exceptions import InputError
 from textractor.entities.document_entity import DocumentEntity
 from textractor.visualizers.entitylist import EntityList
-
+from textractor.utils.html_utils import escape_text
+from textractor.data.text_linearization_config import TextLinearizationConfig
 
 class Line(DocumentEntity):
     """
@@ -64,13 +65,13 @@ def words(self):
         """
         return self._children
 
-    def get_text_and_words(self, config):
+    def get_text_and_words(self, config: TextLinearizationConfig = TextLinearizationConfig()):
         if not self.bbox:
             self.bbox = BoundingBox.enclosing_bbox(self.words)
         for w in self.words:
             w.line_id = self.id
             w.line_bbox = self.bbox
-        return self.text, self.words
+        return escape_text(self.text, config), self.words
 
     @property
     def page(self):

diff --git a/textractor/entities/word.py b/textractor/entities/word.py
@@ -8,7 +8,7 @@
 from textractor.data.text_linearization_config import TextLinearizationConfig
 from textractor.entities.bbox import BoundingBox
 from textractor.entities.document_entity import DocumentEntity
-
+from textractor.utils.html_utils import escape_text
 
 class Word(DocumentEntity):
     """
@@ -148,7 +148,7 @@ def words(self):
     def get_text_and_words(
         self, config: TextLinearizationConfig = TextLinearizationConfig()
     ):
-        return self.text, [self]
+        return escape_text(self.text, config), [self]
 
     def __repr__(self) -> str:
         """

diff --git a/textractor/utils/html_utils.py b/textractor/utils/html_utils.py
@@ -1,3 +1,4 @@
+import html
 from textractor.data.html_linearization_config import HTMLLinearizationConfig
 
 def add_id_to_html_tag(prefix, id, config):
@@ -9,3 +10,9 @@ def add_id_to_html_tag(prefix, id, config):
         return prefix[:-1] + f' id="{id}"' + prefix[-1]
     else:
         return prefix
+
+def escape_text(text, config):
+    if not isinstance(config, HTMLLinearizationConfig):
+        return text
+    else:
+        return html.escape(text)