☂️ Python Coverage

current status: ✅

Overall Coverage

New Files

No new covered files...

Modified Files

updated for commit: b1cd003 by action🐍

Test Results (postgresql)

991 tests  +991   990 ✅ +990   2m 32s ⏱️ + 2m 32s
  1 suites +  1     1 💤 +  1 
  1 files   +  1     0 ❌ ±  0 

Results for commit b1cd003. ± Comparison against base commit fcfb74a.

♻️ This comment has been updated with latest results.

Test Results (mariadb)

991 tests  +1   990 ✅ +1   2m 45s ⏱️ -1s
  1 suites ±0     1 💤 ±0 
  1 files   ±0     0 ❌ ±0 

Results for commit b1cd003. ± Comparison against base commit fcfb74a.

♻️ This comment has been updated with latest results.

Greptile Summary

This PR replaces the untyped dict[str, Any] payload on the PUT /api/roms/{id}/props endpoint with a properly documented RomUserData Pydantic model and adopts model_dump(exclude_unset=True) so that omitted fields are never overwritten — directly addressing the reported data-clearing issue. The change makes the API self-documenting in OpenAPI and correctly handles the "send only what you want to change" contract.

Key changes:

• New RomUserData model with all fields optional and range constraints (ge/le) on rating, difficulty, and completion
• RomUserUpdatePayload.data updated from dict[str, Any] to RomUserData; update_last_played / remove_last_played gain OpenAPI descriptions
• Endpoint logic simplified to payload.data.model_dump(exclude_unset=True)
• New test test_update_rom_user_props_partial_update verifying that a single-field update leaves other fields intact

Issue to address:

• The rating, difficulty, and completion fields are typed as int | None, but the ge/le constraints are only enforced when the value is an integer — a None value passes Pydantic validation and reaches the DB as NULL on a NOT NULL column, causing a 500 error. These fields should be typed as plain int (with defaults matching the DB defaults) so that numeric range validation is always applied and None is rejected with a 422.

Confidence Score: 3/5

• The partial-update logic is sound but int | None fields bypass ge/le validation and risk 500-level DB errors if null is sent.
• The core exclude_unset=True mechanism is correct and the main feature (partial updates without data loss) works as intended. However, rating, difficulty, and completion being typed as int | None means None bypasses Pydantic's numeric range constraints and propagates to a NOT NULL database column, causing a runtime 500. This is a straightforward but real correctness gap that should be fixed before the endpoint is exposed in production API docs.
• backend/endpoints/roms/init.py — the int | None field types on rating, difficulty, and completion in RomUserData need attention to prevent None bypassing ge/le validation.

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant Client
    participant Endpoint as PUT /api/roms/{id}/props
    participant Pydantic as RomUserUpdatePayload
    participant DB as db_rom_handler

    Client->>Endpoint: JSON body (partial fields)
    Endpoint->>Pydantic: Parse into RomUserUpdatePayload
    Note over Pydantic: RomUserData validates each field<br/>(ge/le, enum, optional)
    Pydantic-->>Endpoint: Validated payload

    Endpoint->>Endpoint: payload.data.model_dump(exclude_unset=True)
    Note over Endpoint: Only explicitly provided fields<br/>end up in cleaned_data

    alt update_last_played = true
        Endpoint->>Endpoint: cleaned_data["last_played"] = now()
    else remove_last_played = true
        Endpoint->>Endpoint: cleaned_data["last_played"] = None
    end

    Endpoint->>DB: update_rom_user(id, cleaned_data)
    Note over DB: SQLAlchemy UPDATE with<br/>only the provided columns

    DB-->>Endpoint: Updated RomUser
    Endpoint-->>Client: RomUserSchema (200 OK)

_{Last reviewed commit: b1cd003}