Description

This PR extends the timestamp conversion fixes in #262 so the SDK accepts payloads without requiring callers to normalize them:

• dict/object timestamps may use the nanos key/attribute (protobuf-style) in addition to nanoseconds;
• string timestamps whose fractional part has no timezone suffix (e.g. "2025-11-06T12:00:00.000") no longer cause ValueError in get_precision_timestamp. The same integer-based normalization (divmod, timedelta) and review feedback from PR fix(util): improve the robustness of timestamp conversion function #262 are preserved.

Related Issues

Fixes #260 (fully; PR #262 fixed the dict/object path for nanoseconds only; this adds nanos and fixes the string path when Z/z is missing).

Changes Made

• timestamp_conversion (util.py):
  • Object branch: Resolve sub-second value via getattr(timestamp, "nanoseconds", getattr(timestamp, "nanos", None)) so both .nanoseconds and .nanos are accepted.
  • Dict branch: Resolve via timestamp.get("nanoseconds", timestamp.get("nanos")) so both "nanoseconds" and "nanos" keys are accepted (e.g. protobuf/Firebase backend payloads).
• get_precision_timestamp (util.py): Only split the fraction on "Z" or "z" when present; otherwise use the fraction as-is. Prevents ValueError: not enough values to unpack when the Cloud Event time (or other string) has no trailing timezone (e.g. "2025-11-06T12:00:00.000").
• Tests (test_util.py):
  • test_timestamp_conversion_dict_nanos_equivalent_to_nanoseconds: dict with "nanos" produces the same result as "nanoseconds".
  • test_timestamp_conversion_object_nanos: object with .seconds and .nanos (no .nanoseconds) is accepted and matches dict result.
  • test_get_precision_timestamp_without_timezone_suffix: get_precision_timestamp("2025-11-06T12:00:00.000") does not raise and returns MICROSECONDS.

Testing

Tests Run

• Unit tests
  • pytest tests/test_util.py -v -k "timestamp_conversion or get_precision"
• Integration tests
• Manual testing
  • Built the package and installed it into the issue-260 MRE. Started Firebase Functions emulator, sent POST with payloads/stability_digest_ce.json (payload uses digestDate with nanos and includes top-level time). Result: handler ran successfully, log showed Stability digest: app_id=1:123456789:android:abcdef, curl returned OK. No AttributeError or ValueError.

Additional Notes

• Enhancement over PR fix(util): improve the robustness of timestamp conversion function #262: PR fix(util): improve the robustness of timestamp conversion function #262 added handling for dict/object timestamps with "seconds" and "nanoseconds" and kept the existing string path. This PR; widens accepted keys/attributes to include nanos, and (2) fixes the string path in get_precision_timestamp when the fraction has no Z/z. Same normalization logic and alignment with review (no fromtimestamp, integer math, docstring/parameter names) are unchanged.
• What was happening before: With PR fix(util): improve the robustness of timestamp conversion function #262, payloads where the backend sends digestDate as {"seconds": "...", "nanos": 0} still failed (dict branch required "nanoseconds"). After accommodating that, the event’s top-level time field could be a string without a trailing Z/z; get_precision_timestamp then did s_fraction.split("z") and unpacked to two values, causing ValueError: not enough values to unpack (expected 2, got 1). This PR addresses both cases so the SDK accepts varied real-world payloads.