[yunusarda-sudo]

Resolves #2108.

Context & Problem

In #2098, the FHIR collection reader was updated to refuse queried collection bundles as complete reads (verifying that no query is present across configured parameters, requested and effective URLs, and self-links).

However, as reported in #2108, the sibling reader in openadapt_flow/ensure_record.py (FhirReader.bundle()) still accepted any non-searchset bundle without a next link as complete even when query parameters were sent or present on the URL/self-link, allowing a capped or filtered read to count as a full read.

Solution

• openadapt_flow/fhir_read.py: Export check_collection_url (aliasing _check_collection_url for backwards compatibility).
• openadapt_flow/ensure_record.py (FhirReader.bundle):
  • Track initial_url and unpack headers from fetch.
  • When a bundle is not a searchset or history bundle and has no next link (if not nxt), apply the collection refusal rule:
    • Refuse if configured params are present.
    • Refuse if initial_url contains a query.
    • Refuse if the effective served URL (Content-Location or Location header) contains a query.
    • Refuse if any relation: "self" link contains a query.
  • Raise Unreadable(f"{path}: {exc}") on any violation.
• packages/flow/tests/test_ensure_record.py:
  • Add regression test test_bundle_refuses_parameterized_non_searchset_bundle covering configured params, query in path, query in self-link, query in effective URL (content-location), unqueried collection acceptance, and normal parameterized searchset pagination/coverage.

Verification

• PYTHONPATH=packages/flow pytest packages/flow/tests/test_ensure_record.py -q: 112/112 passed.
• PYTHONPATH=packages/flow pytest packages/flow/tests/test_page_snapshot_consistency.py -q: 82/82 passed.
• ruff check and ruff format --check: clean.