Join citation-only lines onto the preceding text

lib/tickers.py

@@ -19,7 +19,9 @@
 PROTECTED_TAG_RE = re.compile(r"<(/?)(a|code|pre)\b", re.I)
 BOLD_TAG_RE = re.compile(r"<(/?)(b|strong)\b", re.I)
 URL_RE = re.compile(r'https?://[^\s<>"]+')
-RAW_ANCHOR_RE = re.compile(r'<a href="(https?://[^"]+)">\s*https?://[^<]*</a>')
+_CITE = r'<a href="[^"]*">\[\d+\]</a>'
+CITE_ONLY_RUN_RE = re.compile(rf"\s*\n\s*({_CITE}(?:\s+{_CITE})*)[ \t]*(?=\n|$)")  # lines holding only citations
+RAW_ANCHOR_RE =re.compile(r'<a href="(https?://[^"]+)">\s*https?://[^<]*</a>')
 # "Micron (MU)": up to four capitalised words (or a number) right before a parenthesised ticker.
 _WORD = r"[A-Z0-9][\w&'’-]*(?:\.[\w&'’-]+)*"
 NAME_TICKER_RE = re.compile(
@@ -94,7 +96,9 @@ def cite(m: re.Match) -> str:
         return f'<a href="{url}">[{n}]</a>{tail}'
 
     text = RAW_ANCHOR_RE.sub(lambda m: m.group(1), text)
-    return _rewrite(text, lambda seg, bold: URL_RE.sub(cite, seg))
+    text = _rewrite(text, lambda seg, bold: URL_RE.sub(cite, seg))
+    # Links the model put on their own lines join the text before them: "volume.[1][2]"
+    return CITE_ONLY_RUN_RE.sub(lambda m: re.sub(r"(?<=</a>)\s+", "", m.group(1)), text)
 
 
 def link_tickers(text: str) -> str:

tests/test_tickers.py

@@ -58,6 +58,14 @@ def test_raw_links_go_behind_citation_numbers():
     assert "quote/NVDA" not in out  # the ticker inside the URL is not linked
 
 
+def test_citations_on_their_own_lines_join_the_preceding_text():
+    a, b = '<a href="https://a.com/1">[1]</a>', '<a href="https://b.com/2">[2]</a>'
+    assert link_tickers("Thin volume.\n\nhttps://a.com/1\nhttps://b.com/2") == f"Thin volume.{a}{b}"
+    assert link_tickers("One.\n\nhttps://a.com/1\n\nTwo.\nhttps://b.com/2") == f"One.{a}\n\nTwo.{b}"
+    assert link_tickers("Said https://a.com/1 then more") == f"Said {a} then more"  # inline: untouched
+    assert link_tickers("Thin.\nhttps://a.com/1 and more") == f"Thin.\n{a} and more"  # not a citation-only line
+
+
 def test_link_labelled_with_its_own_url_is_numbered_but_named_links_and_code_are_not():
     assert link_tickers('<a href="https://y.com/p">https://y.com/p…</a> ok') == '<a href="https://y.com/p">[1]</a> ok'
     assert link_tickers('<a href="https://y.com/p">Reuters</a>') == '<a href="https://y.com/p">Reuters</a>'