User description

Motivation

• Prevent malformed dailyInfo payloads from causing day-offset reindexing or accidental entity churn by aligning config-flow validation with runtime parsing rules.
• Ensure the coordinator treats non-list or mixed/invalid dailyInfo payloads as malformed and either preserves prior data or fails cleanly on first refresh.
• Avoid creating D+1/D+2 sensors when effective forecast_days options do not permit them and reduce accidental unique-id collisions for nearby coordinates.
• Improve test isolation to avoid module-level stub cross-contamination between tests.

Description

• Config flow now treats an HTTP-200 response as invalid unless dailyInfo is a non-empty list of dict objects and logs/wraps the error via the existing cannot_connect placeholder (custom_components/pollenlevels/config_flow.py).
• Coordinator parsing now treats dailyInfo as invalid when it is not a list or when any list item is not a dict, preserves the last successful dataset when present, and raises UpdateFailed on first refresh when no prior data exists (custom_components/pollenlevels/coordinator.py).
• Sensor setup uses a validated local data snapshot for iteration and skips creating *_d1/*_d2 sensors when runtime flags and forecast_days disallow them, and proactively cleans up stale per-day entities (custom_components/pollenlevels/sensor.py).
• Integration setup normalizes stored forecast sensor mode via normalize_sensor_mode and gates create_d1/create_d2 by the effective forecast_days, and config-flow coordinate unique-id precision was increased to 6 decimals while rejecting whitespace-only API keys (custom_components/pollenlevels/__init__.py and config_flow.py).
• Tests updated and added: new config-flow regression tests for non-list and non-dict dailyInfo shapes; coordinator tests for invalid/mixed dailyInfo behavior and preservation of last data; sensor test ensuring disabled D+ sensors are not recreated; and setup tests were switched to scoped monkeypatch usage to avoid persistent module reassignment (tests/*).
• CHANGELOG.md updated under the existing 1.9.3 section to reflect these hardening and test-isolation changes (no version bump).

Testing

• Ran import-order fix and import checks with ruff check --fix --select I and ruff check on the modified files, and they passed.
• Formatted changed files with black and applied the reformat to the targeted files successfully.
• Executed the full test suite with pytest -q and all tests passed (138 passed).

Codex Task

PR Type

Enhancement, Bug fix, Tests

Description

• Hardened dailyInfo validation in config flow and coordinator to reject non-list/non-dict payloads and preserve last successful data

• Gated D+1/D+2 sensor creation by effective forecast_days to prevent disabled sensors from being recreated

• Increased unique-id precision from 4 to 6 decimals to reduce collisions for nearby coordinates

• Normalized stored forecast sensor mode values during setup and improved test isolation with scoped monkeypatching

Diagram Walkthrough

flowchart LR
  A["Config Flow Validation"] -->|"Validate dailyInfo<br/>is non-empty list of dicts"| B["Reject Invalid Payloads"]
  C["Coordinator Parsing"] -->|"Check dailyInfo type<br/>and item types"| D["Preserve Last Data<br/>or Raise UpdateFailed"]
  E["Setup Entry"] -->|"Normalize mode +<br/>Gate by forecast_days"| F["Create D+1/D+2 Flags"]
  F -->|"Pass flags to sensor setup"| G["Skip Disabled Sensors"]
  H["Unique ID Generation"] -->|"6 decimal precision"| I["Reduce Collisions"]