fix: Python 3.11 date.fromisoformat() allows extra formats

Python 3.11 and later allow additional ISO8601 formats in `datetime` module
ISO8601 parsing. These formats are not RFC3339 section 5.6 compliant.

Especially `datetime.date.fromisoformat()` now allows strings like:
 * `20230328` (2023-03-28)
 * `2022W527` (2023-01-01)
 * `2023-W01` (2023-01-02)
 * `2023-W13-2` (2023-03-28)

Fix by doing a regular expression check before passing the value to `datetime`
module. This made the original `.isascii()` check unnecessary.

See:
 * https://docs.python.org/3/whatsnew/3.11.html#datetime
 * https://github.com/python/cpython/commit/1303f8c927
 * https://docs.python.org/3.11/library/datetime.html#datetime.date.fromisoformat
 * https://www.rfc-editor.org/rfc/rfc3339#section-5.6

Tests covering the invalid values to be sent to json-schema-org/JSON-Schema-Test-Suite

Fixes #1056.

1 files changed, 6 insertions, 1 deletions

diff --git a/jsonschema/_format.py b/jsonschema/_format.py
index d274c5f..2231f5e 100644
--- a/jsonschema/_format.py
+++ b/jsonschema/_format.py
@@ -17,6 +17,8 @@ _RaisesType = typing.Union[
     typing.Type[Exception], typing.Tuple[typing.Type[Exception], ...],
 ]
 
+_RE_DATE = re.compile(r"^\d{4}-\d{2}-\d{2}$", re.ASCII)
+
 
 class FormatChecker:
     """
@@ -396,7 +398,10 @@ def is_regex(instance: object) -> bool:
 def is_date(instance: object) -> bool:
     if not isinstance(instance, str):
         return True
-    return bool(instance.isascii() and datetime.date.fromisoformat(instance))
+    return bool(
+        _RE_DATE.fullmatch(instance)
+        and datetime.date.fromisoformat(instance)
+    )
 
 
 @_checks_drafts(draft3="time", raises=ValueError)