[VitorCoelho]

What

Edm.DateTime and Edm.DateTimeOffset literal parsing now normalizes the fractional seconds to exactly 6 digits (pad or truncate) through one small helper, normalize_fractional_seconds, before calling datetime.fromisoformat.

Why

#303 switched to fromisoformat and padded the fraction, because Python 3.10 (still supported and in the CI matrix) accepts only 3 or 6 digits. Two cases were still broken on 3.10:

• Edm.DateTimeOffset took the first 6 characters after the dot as the fraction, so a shorter fraction swallowed part of the offset: datetimeoffset'1976-11-23T03:33:06.5+01:00' and datetimeoffset'...06.12Z' raised PyODataModelError.
• 7-digit fractions (fffffff, allowed by OData V2 and mentioned in the Edm.DateTime docstring) were padded but never truncated, so both types failed.

Python 3.11+ accepts these natively and truncates to microseconds; the helper gives 3.10 the same result.

How it was tested

• New cases in tests/test_model_v2.py: 7-digit Edm.DateTime; Edm.DateTimeOffset with 1, 2 and 7 digits plus offset/Z.
• Without the fix, on Python 3.10: 4 of the new cases fail. With the fix: full suite passes on Python 3.10 and 3.13 (301 passed).
• CHANGELOG entry added under Unreleased / Fixed.

[phanak-sap]

Hi @VitorCoelho,
thx for suggestion, this feels like temporary helper refactoring for 3.10 python only, than necessary bugfix. While the code is correct, I would close this PR anyway from following reasons:

• while we are intentionally bit lagging behind Python versions in support, in pyodata 1.13 or definitely in 1.14 (if some feature would enforce creating own MINOR version) the minimum Python version will be 3.11. It is already quite after EOL and even 3.11 is quite long out.
• with that, all things mentioned in fromisotime cleanup when 3.11 is Python miniumum #304 will be done as well and no helper should be IMHO needed anymore. "remove fractional-seconds padding workaround" is second item there.
• the DateTime split is the one of most "in the face" change between V2 and V4 odata. I would use the "3.11 is minimal + all refactoring cleaned" as a basis for starting of V4 support. At the moment it lives in experimental branch, refer to https://github.com/SAP/python-pyodata/wiki/Road-to-OData-V4 for detail. Since you are so eager to help, I will enjoy any testing effort (at the moment I tried that code against Nothwind sample service - Batch and FunctionImports/Batches are not working at all, but hey, at least it parses metadata and most of the /docs examples seems to work)

[VitorCoelho]

Thanks for the context. That makes sense: with 3.11 as the minimum the workaround goes away anyway, so there's no point patching it now.

[phanak-sap]

Thanks for the context. That makes sense: with 3.11 as the minimum the workaround goes away anyway, so there's no point patching it now.

OK, closing on mutual agreement.