[AtharvaTilewale]

Documents that ForceField(), ForceField.parse_sources() and ForceField.parse_smirnoff_from_source() accept a string containing OFFXML content, in addition to file paths and file-like objects. As discussed in #2021, a string that does not resolve to a file name falls through to ParameterIOHandler.parse_string() and this has been de facto public API since the beginning, but all three sources/source docstrings described only files.

The parse_sources docstring also now says that a bare string is iterated character by character (iter(str) succeeds), so a single string source has to be wrapped in a list; ForceField(<string>) is unaffected because *sources already arrives as a tuple. This came out of checking each documented statement against the toolkit:

ForceField(<offxml string>)                              -> loads
ForceField().parse_sources([<offxml string>])            -> loads
ForceField().parse_sources(<offxml string>)              -> "Source '<' could not be read"
ForceField().parse_smirnoff_from_source(<offxml string>) -> {'SMIRNOFF': ...}
ForceField("openff-2.0.0.offxml")                        -> file-name lookup unchanged
ForceField("openff-2.0.0.offxml", <offxml string>)       -> mixed sources load

The __init__ entry is also reflowed into a bullet list, per the whitespace comment on the issue. ruff check passes (one rewrapped line was over the 119-character limit) and pytest openff/toolkit/_tests/test_forcefield.py -k "string or parse_sources or xml" gives 164 passed, 4 skipped. Docstring-only change, so no new tests. The release-history entry references this PR's number.

Fixes #2021

• Tag issue being addressed
• Add tests (docstring-only change)
• Update docstrings/documentation, if applicable
• Lint codebase
• Update changelog

See our AI Policy. Please indicate whether LLMs or other AI-powered tools (beyond simple IDE use cases) were used in this contribution:

• No
• Yes

[j-wags]

Thanks for opening this @AtharvaTilewale - The CI failures are expected (we use a secret that isn't accessible from forks). I'll review this shortly!

[j-wags]

Thanks @AtharvaTilewale , this looks good to me. I'll push one change and merge!