model: escape single quotes in Edm.String literals

[lefranz]

Hi — thanks for maintaining pyodata.

Problem

EdmStringTypTraits.to_literal() wraps the value in single quotes but does not escape quotes inside the value:

>>> from pyodata.v2.model import EdmStringTypTraits
>>> EdmStringTypTraits().to_literal("délit d'initié")
"'délit d'initié'"          # the literal ends at the apostrophe

An embedded single quote in an OData string literal has to be doubled (''), so this produces a query the service refuses. Any keyword-argument or callable filter over a value containing an apostrophe is affected — which in practice means a lot of French, Italian and Irish text (l'Etat, dell', O'Brien).

Reproduction against a live service

Filtering the Swiss parliament's public OData v2 endpoint (ws.parlament.ch, via swissparlpy) on Title eq "délit d'initié":

GET /odata.svc/Business?$filter=Language eq 'FR' and Title eq 'délit d'initié'
→ HTTP 400
  {"error":{"message":{"value":"Syntax error at position 45 in
   'Language eq 'FR' and Title eq 'délit d'initié''."}}}

Position 45 is the apostrophe in d'initié. With this patch the identical call returns HTTP 200 and the expected row.

Change

• to_literal: double any ' in the value.
• from_literal: collapse '' back to ', so the two directions round-trip. This felt in scope since model: fix stripping single-quotes from string values #309 just tightened from_literal, but I am happy to drop it and keep this PR to the wire-format bug alone if you would rather treat the inbound direction separately.

Notes on decisions

• to_literal now calls str(value) explicitly, because the %s interpolation that used to stringify implicitly is replaced by a .replace() call. Accepted types are unchanged.
• I did not touch pyodata/v4/, having only verified the v2 path.
• Escaping is applied in the traits rather than at the filter-construction sites, so it covers every caller that goes through Edm.String literals.

Tests

Added to test_traits in tests/test_model_v2.py, next to the from_literal cases from #309: to_literal doubling, and a round-trip loop over ["O'Brien", "l'Etat", "a''b", "'wrapped'", 'no quotes', ''].

Full suite passes: 295 passed. pylint/flake8 from dev-requirements.txt could not run in my environment (Python 3.12 — the pinned versions fail on the removed pkg_resources and on the importlib.metadata entry-points API); this is independent of the change, and no added line exceeds 120 columns.

CHANGELOG updated under [Unreleased] → Fixed.

[cla-assistant]

All committers have signed the CLA.

[filak-sap]

Well done! Thank you very much for your contribution. Excellent summary helping us to save time re-learning the odata stuff.