Bump Beautiful Soup (#1167)

* bump

* check lint

* fix action

* change action

* update

* assert Tag

* fix table and partition

* revert workflow changes

lib/sycamore/pyproject.toml

@@ -22,7 +22,7 @@ ray = { extras = ["default"], version = "^2.36.0" }
 pyarrow = "^14.0.2"
 numpy = "<2.0.0"
 openai = "^1.60.2"
-beautifulsoup4 = "^4.12.2"
+beautifulsoup4 = "^4.13.1"
 amazon-textract-textractor = "^1.3.2"
 boto3 = "^1.28.70"
 boto3-stubs = {extras = ["essential"], version = "^1.35.12"}

lib/sycamore/sycamore/data/table.py

@@ -202,7 +202,7 @@ def from_html(cls, html_str: Optional[str] = None, html_tag: Optional[Tag] = Non
 
         if (html_str is not None and html_tag is not None) or (html_str is None and html_tag is None):
             raise ValueError("Exactly one of html_str and html_tag must be specified.")
-
+        root: Union[Tag, BeautifulSoup]
         if html_str is not None:
             html_str = html_str.strip()
             if not html_str.startswith("<table") or not html_str.endswith("</table>"):
@@ -222,9 +222,10 @@ def from_html(cls, html_str: Optional[str] = None, html_tag: Optional[Tag] = Non
 
         cells = []
         caption = None
-
+        assert isinstance(root, Tag), "Expected root to be a Tag"
         # Traverse the tree of elements in a pre-order traversal.
         for tag in root.find_all(recursive=True):
+            assert isinstance(tag, Tag), "Expected root to be a Tag"
             if tag.name == "tr":
                 cur_row += 1  # TODO: Should this be based on rowspan?
                 cur_col = 0
@@ -234,9 +235,12 @@ def from_html(cls, html_str: Optional[str] = None, html_tag: Optional[Tag] = Non
                 # they have a thead.
                 if cur_row < 0:
                     cur_row += 1
-
-                rowspan = int(tag.attrs.get("rowspan", "1"))
-                colspan = int(tag.attrs.get("colspan", "1"))
+                if rowspan_str := tag.attrs.get("rowspan", "1"):
+                    assert isinstance(rowspan_str, str)  # For mypy
+                    rowspan = int(rowspan_str)
+                if colspan_str := tag.attrs.get("colspan", "1"):
+                    assert isinstance(colspan_str, str)  # For mypy
+                    colspan = int(colspan_str)
 
                 content = tag.get_text()
 

lib/sycamore/sycamore/transforms/partition.py

@@ -2,7 +2,7 @@
 import io
 from typing import Any, Literal, Optional, Union
 
-from bs4 import BeautifulSoup
+from bs4 import BeautifulSoup, Tag
 
 from sycamore.functions import TextOverlapChunker, Chunker
 from sycamore.functions import CharacterTokenizer, Tokenizer
@@ -312,6 +312,7 @@ def partition(self, document: Document) -> Document:
         if self._extract_tables:
             for table in soup.find_all("table"):
                 # ignore nested tables
+                assert isinstance(table, Tag)
                 if len(table.find_all("table")) > 0:
                     continue