Unit-test the template / entity-ID extraction layer in util.py

[frenckatron]

Why This Matters

The regex stack in util.py:86-125 (ENTITY_ID_PATTERN, ENTITY_ID_TEMPLATE_PATTERNS) plus extract_entities_from_template_regex (line 499), split_comma_separated_entity_ids (line 628), and extract_template_strings_from_config (line 651) are the engine behind every unknown_*_references repair in ectoplasms/automation/repairs/, ectoplasms/script/repairs/, ectoplasms/lovelace/repairs/, etc. They are pure(-ish) functions with rich edge cases (comma-separated lists, nested templates, states.domain.entity direct access, expand(...) filters, IGNORED_ENTITY_DOMAINS prefix matching). Pure functions = highest test ROI: no hass fixture overhead, parametrized tables cover dozens of cases in milliseconds, and any future regex tweak (a recurring source of false-positive repairs — see commit 29fd59d) gets immediate feedback.

Approach

Add tests/test_util.py with three parametrized test classes: (a) test_extract_entities_from_template_regex over a table of ~20 templates including {{ states('light.kitchen') }}, {{ state_attr('sensor.x', 'foo') }}, {{ expand('group.lights') }}, states.binary_sensor.door.state, multi-entity strings, malformed templates, and false-positive bait like service names. (b) test_split_comma_separated_entity_ids covering empty/None/single/multi/whitespace edge cases. (c) test_extract_template_strings_from_config over dicts, lists, nested structures, and non-string values. For functions that take hass (e.g. async_filter_known_entity_ids), use a minimal hass fixture and seed er.async_get(hass) with known entities.

Acceptance Criteria

• >=20 parametrized cases for extract_entities_from_template_regex, including known-false-positive bait
• split_comma_separated_entity_ids covers all branches in the function
• IGNORED_ENTITY_DOMAINS filtering is asserted (no scene.* or group.* leaks)
• At least one test reproduces the bug fixed by commit 29fd59d ("Find unknown entities in heading badge") to prevent regression

Risks & Caveats

These regexes are deliberately permissive to catch unknown entities — be careful not to write tests that lock down behavior the maintainer actually wants flexible. Pair with the maintainer on the table of cases. The KNOWN_DOMAINS list (util.py:84) is a hard-coded enum; tests should treat it as a fixture so domain additions don't silently break.

Scores

• Impact: ██████░░░░ 6/10
• Difficulty: ███░░░░░░░ 3/10
• Short-Term ROI: █████████░ 9/10
• Long-Term Value: ███████░░░ 7/10

Priority

Prototype First

Dependencies

#1257