fix(node): include user keys on fetch to avoid spurious changed status

[BeArchiTek]

Summary

The node module can report changed: true on idempotent re-runs of a task whose data includes a relationship. Re-applying the exact same payload should be a no-op but currently flips status to changed.

Root cause

_get_object calls fetch_single_node without an include list, so cardinality-MANY non-attribute/non-parent relationships are not requested in the GraphQL query and stay uninitialized on the returned node. When _update_object builds the diff:

• serialized_before = node._generate_input_data() — skips uninitialized relationships (SDK: if not rel.initialized: continue).
• setattr(node, rel_name, …) initializes the relationship.
• serialized_after = node._generate_input_data() — now includes the relationship.

dict_hash(serialized_before) != dict_hash(serialized_after) even when the underlying value matches → false changed: true.

This regressed when the original include-on-fetch fix (commit 6f1315e) was removed during the diff rework in 0142ad6.

Fix

In _get_object, derive include from the user-provided data keys (excluding id / hfid which are not schema fields) and pass it to fetch_single_node. Empty list collapses to None so the state: absent path and identifier-only lookups keep their default fetch behavior.

include = [key for key in data if key not in ("id", "hfid")] or None
node = self.client.fetch_single_node(
    kind=kind, id=node_id, hfid=node_hfid, include=include, raise_when_missing=False
)

Test plan

• Unit tests pass: pytest tests/unit/plugins/module_utils/test_node.py → 23/23
• Integration cycle playbook against sandbox.infrahub.app:
  • Create tag1 → changed
  • Update tag1 → changed
  • Existing tag1 → ok (not changed) ← the idempotency assertion this PR restores
  • Delete tag1 → changed
  • Delete tag1 (already absent) → ok
• Re-run the cycle playbook a second time to confirm deterministic idempotency on relationship-bearing kinds.

---

Summary by cubic

Fixes false changed status in the node module by fetching nodes with the user’s fields included. Adds regression tests to prevent this idempotency issue from returning.

• Bug Fixes
  • Derive include from user data keys (excluding id and hfid) and pass to fetch_single_node.
  • Ensure relationships are initialized in the "before" snapshot to avoid spurious diffs.
  • Fall back to default fetch when only identifiers are provided, preserving state: absent and lookup behavior.
  • Add unit tests asserting include contains user keys and defaults to None for identifier-only data.

^{Written for commit 0f977e1. Summary will update on new commits.}

[cloudflare-workers-and-pages]

Deploying infrahub-ansible with    Cloudflare Pages

Latest commit:	0f977e1
Status:	✅  Deploy successful!
Preview URL:	https://30d08f5c.infrahub-ansible.pages.dev
Branch Preview URL:	https://fix-diff-include-relationshi.infrahub-ansible.pages.dev

View logs

[cubic-dev-ai]

No issues found across 1 file

_{Re-trigger cubic}