fix: keep every entry heading on listing pages (#774)

A blog index, archive or category page holds several <article> siblings.
BODY_XPATH's "(.//article)[1]" keeps the first entry only, so the page
looks under-extracted and escalates to justext, which emits paragraphs
and drops every heading.

Give the recall pass a listing expression that takes the container of
the sibling articles, so the retry returns all entries with their heads
and wins the escalation. Recall only: on an article page the same shape
is a related-posts strip, whose container would hide the body.

tests/unit_tests.py

@@ -2772,6 +2772,51 @@ def test_no_duplicate_content():
         assert out.count("Second synthetic paragraph") == 1
 
 
+def _listing_page(entries):
+    "Blog-index markup: sibling <article> teasers, each a link-only heading plus its lead paragraph."
+    # each teaser clears min_extracted_size on its own, so only the page-share
+    # test flags the layout as under-extracted and escalates to recall
+    articles = "".join(
+        f'<article><header class="entry-header"><h2 class="entry-title">'
+        f'<a href="/{i}">{title}</a></h2></header>'
+        f"<div class=\"entry-summary\"><p>{text}</p></div></article>"
+        for i, (title, text) in enumerate(entries)
+    )
+    return f'<html><body><div id="wrapper"><main>{articles}</main></div></body></html>'
+
+
+def test_listing_page_keeps_every_entry_heading():
+    "regression #774: an index page must not collapse to its first entry, losing every other heading."
+    entries = [
+        (
+            f"Entry number {i} about a self contained subject",
+            f"Teaser {i} carries enough prose to pass the extractor thresholds on its own, running "
+            f"over several clauses and a couple of sentences. It reads as a lead paragraph would, "
+            f"so nothing about this entry looks like boilerplate to the extractor, and the index "
+            f"page as a whole reads as a real body of text rather than a stub.",
+        )
+        for i in range(8)
+    ]
+    result = extract(_listing_page(entries), output_format="markdown", config=use_config()) or ""
+    for title, _ in entries:
+        assert f"## {title}" in result
+
+
+def test_article_page_ignores_related_posts_strip():
+    "the other side of #774: the same shape is a related-posts strip, which must not displace the body."
+    body = " ".join(
+        f"Sentence {i} of the actual article body, long enough that the extractor never escalates." for i in range(8)
+    )
+    related = "".join(f"<article><h2><a href='/r{i}'>Related teaser {i}</a></h2></article>" for i in range(3))
+    doc = (
+        f'<html><body><main><article class="post"><h1>The Real Headline</h1><p>{body}</p></article>'
+        f'<aside class="related"><div>{related}</div></aside></main></body></html>'
+    )
+    result = extract(doc, output_format="txt", config=use_config()) or ""
+    assert "Sentence 3 of the actual article body" in result
+    assert "Related teaser 1" not in result
+
+
 def test_short_document_keeps_structure():
     "regression #896: the baseline rescue must not flatten a valid short extraction it cannot improve on."
     doc = (

trafilatura/main_extractor.py

@@ -35,6 +35,7 @@ from .xpaths import (
     COMMENTS_DISCARD_XPATH,
     COMMENTS_XPATH,
     DISCARD_IMAGE_ELEMENTS,
+    LISTING_BODY_XPATH,
     OVERALL_DISCARD_XPATH,
     PRECISION_DISCARD_XPATH,
     TEASER_DISCARD_XPATH,
@@ -654,8 +655,10 @@ def _extract(tree: HtmlElement, options: Extractor) -> tuple[_Element, str, set[
     if options.links is True:
         potential_tags.add("ref")
     result_body = Element("body")
+    # kept out of the confident passes: see LISTING_BODY_XPATH
+    expressions = BODY_XPATH if options.focus != "recall" else LISTING_BODY_XPATH + BODY_XPATH
     # iterate
-    for expr in BODY_XPATH:
+    for expr in expressions:
         # select tree if the expression has been found
         subtree = next((s for s in expr(tree) if s is not None), None)
         if subtree is None:

trafilatura/xpaths.py

@@ -105,6 +105,13 @@ BODY_XPATH = [
         """
     ),
 ]
+
+# Listing layouts (blog index, archive, category) hold several <article> siblings in one container,
+# each a teaser with its own heading. BODY_XPATH's "(.//article)[1]" keeps the first entry only, so
+# the page escalates to justext, which flattens every heading into a paragraph (#774). Recall only:
+# on an article page the same shape is a related-posts strip, whose container would hide the body.
+LISTING_BODY_XPATH = [XPath("(.//article[preceding-sibling::article])[1]/..")]
+
 # starts-with(@id, "article") or
 # or starts-with(@id, "story") or contains(@class, "story")
 # starts-with(@class, "content ") or contains(@class, " content")