Fixed problematic date parsing (#888)

We had some trouble with datetime strings coming back from server not parsing correctly. I added a bit more flexible parsing.

Author: whitead

paperqa/types.py

@@ -30,6 +30,7 @@
     encode_id,
     format_bibtex,
     get_citenames,
+    maybe_get_date,
 )
 from paperqa.version import __version__ as pqa_version
 
@@ -585,8 +586,8 @@ def populate_bibtex_key_citation(  # noqa: PLR0912
                 "pages": data.get("pages"),
                 "month": (
                     None
-                    if not data.get("publication_date")
-                    else data["publication_date"].strftime("%b")
+                    if not (maybe_date := maybe_get_date(data.get("publication_date")))
+                    else maybe_date.strftime("%b")
                 ),
                 "doi": data.get("doi"),
                 "url": data.get("doi_url"),

paperqa/utils.py

@@ -557,3 +557,25 @@ def citation_to_docname(citation: str) -> str:
     if match is not None:
         year = match.group(1)
     return f"{author}{year}"
+
+
+def maybe_get_date(date: str | datetime | None) -> datetime | None:
+    if not date:
+        return None
+    if isinstance(date, str):
+        # Try common date formats in sequence
+        formats = [
+            "%Y-%m-%dT%H:%M:%S%z",  # ISO with timezone: 2023-01-31T14:30:00+0000
+            "%Y-%m-%d %H:%M:%S",  # ISO with time: 2023-01-31 14:30:00
+            "%B %d, %Y",  # Full month day, year: January 31, 2023
+            "%b %d, %Y",  # Month day, year: Jan 31, 2023
+            "%Y-%m-%d",  # ISO format: 2023-01-31
+        ]
+
+        for fmt in formats:
+            try:
+                return datetime.strptime(date, fmt)
+            except ValueError:
+                continue
+        return None
+    return date

tests/test_paperqa.py

@@ -50,6 +50,7 @@
 from paperqa.utils import (
     extract_score,
     get_citenames,
+    maybe_get_date,
     maybe_is_html,
     maybe_is_text,
     name_in_text,
@@ -1644,3 +1645,11 @@ def test_fallback_non_json(self, input_text: str) -> None:
     )
     def test_llm_parse_json_with_escaped_characters(self, input_text, expected_output):
         assert llm_parse_json(input_text) == expected_output
+
+
+def test_maybe_get_date():
+    assert maybe_get_date("2023-01-01") == datetime(2023, 1, 1)
+    assert maybe_get_date("2023-01-31 14:30:00") == datetime(2023, 1, 31, 14, 30)
+    assert maybe_get_date(datetime(2023, 1, 1)) == datetime(2023, 1, 1)
+    assert maybe_get_date("foo") is None
+    assert maybe_get_date("") is None