fix: preserve newlines in Table and TableChunk elements during PDF partitioning

The RE_MULTISPACE_INCLUDING_NEWLINES regex was being applied to all Text
elements, including Table and TableChunk. This incorrectly removed newline
characters that carry structural meaning in tables (row separation).

Fixes #3983

Author: eureka928

CHANGELOG.md

@@ -1,3 +1,8 @@
+## 0.18.33
+
+### Fixes
+- **Preserve newlines in Table/TableChunk elements during PDF partitioning**: Skip whitespace normalization for Table and TableChunk elements so newlines that carry structural meaning (row separation) are preserved (fixes #3983)
+
 ## 0.18.32
 
 ### Enhancements

unstructured/__version__.py

@@ -1 +1 @@
-__version__ = "0.18.32"  # pragma: no cover
+__version__ = "0.18.33"  # pragma: no cover

unstructured/partition/pdf.py

@@ -34,6 +34,8 @@
     Link,
     ListItem,
     PageBreak,
+    Table,
+    TableChunk,
     Text,
     Title,
 )
@@ -823,11 +825,13 @@ def _partition_pdf_or_image_local(
             out_elements.append(cast(Element, el))
         # NOTE(crag): this is probably always a Text object, but check for the sake of typing
         elif isinstance(el, Text):
-            el.text = re.sub(
-                RE_MULTISPACE_INCLUDING_NEWLINES,
-                " ",
-                el.text or "",
-            ).strip()
+            # Skip newline normalization for Table/TableChunk - newlines carry structural meaning
+            if not isinstance(el, (Table, TableChunk)):
+                el.text = re.sub(
+                    RE_MULTISPACE_INCLUDING_NEWLINES,
+                    " ",
+                    el.text or "",
+                ).strip()
             if el.text or isinstance(el, PageBreak):
                 out_elements.append(cast(Element, el))