Merge pull request #99 from fcasalen/feature/ordered_list_item_with_value_attr

AlbertWeichselbraun · web-flow · commit b2dbccf2ba39 · 2025-10-22T20:28:38.000+02:00
chg: handling list items with value attribute
diff --git a/src/inscriptis/model/tag/list_tag.py b/src/inscriptis/model/tag/list_tag.py
@@ -1,4 +1,5 @@
 """Handle the <li>, <ol>, <ul> tags."""
+from typing import Dict
 
 from inscriptis.model.html_document_state import HtmlDocumentState
 
@@ -11,9 +12,16 @@ def get_bullet(state: HtmlDocumentState) -> str:
     return UL_COUNTER[len(state.li_counter) % UL_COUNTER_LEN]
 
 
-def li_start_handler(state: HtmlDocumentState, _: dict) -> None:
+def li_start_handler(state: HtmlDocumentState, tag: Dict) -> None:
     """Handle the <li> tag."""
     bullet = state.li_counter[-1] if state.li_counter else "* "
+    if tag.get('value'):
+        if tag.get('value').isdigit():
+            bullet = int(tag.get('value'))
+            if not state.li_counter:
+                state.li_counter.append(bullet)
+        else:
+            bullet = tag.get('value')
     if isinstance(bullet, int):
         state.li_counter[-1] += 1
         state.tags[-1].list_bullet = f"{bullet}. "
@@ -23,7 +31,7 @@ def li_start_handler(state: HtmlDocumentState, _: dict) -> None:
     state.tags[-1].write("")
 
 
-def ul_start_handler(state: HtmlDocumentState, _: dict) -> None:
+def ul_start_handler(state: HtmlDocumentState, _: Dict) -> None:
     """Handle the <ul> tag."""
     state.li_counter.append(get_bullet(state))
 
@@ -33,7 +41,7 @@ def ul_end_handler(state: HtmlDocumentState) -> None:
     state.li_counter.pop()
 
 
-def ol_start_handler(state: HtmlDocumentState, _: dict) -> None:
+def ol_start_handler(state: HtmlDocumentState, _: Dict) -> None:
     """Handle the <ol> tag."""
     state.li_counter.append(1)
 
diff --git a/tests/test_list_div.py b/tests/test_list_div.py
@@ -27,3 +27,15 @@ def test_divs():
 
     html = "<body>Thomas <ul><li> a  <div>Anton</div>Maria</ul></body>"
     assert get_text(html, config) == "Thomas\n  * a\n    Anton\n    Maria"
+
+    html = "<body>Thomas <ol><li> a  <div>Anton</div>Maria</ol></body>"
+    assert get_text(html, config) == "Thomas\n 1. a\n    Anton\n    Maria"
+
+    html = """<body>Thomas <ol><li value="2"> a  <div>Anton</div>Maria</ol></body>"""
+    assert get_text(html, config) == "Thomas\n 2. a\n    Anton\n    Maria"
+
+    html = """<body>Thomas <ol><li value="2"> a  <ol><li><div>Anton</div></li></ol>Maria</ol></body>"""
+    assert get_text(html, config) == "Thomas\n 2. a\n     1. Anton\n    Maria"
+
+    html = """<body>Thomas <ol><li value="2"> a  <ol><li value="10"><div>Anton</div></li></ol>Maria</ol></body>"""
+    assert get_text(html, config) == "Thomas\n 2. a\n    10. Anton\n    Maria"
