[WIP] Add --badData argument to find jellyfin media entries with bad data

[Copilot]

• Inspect repository structure and current CLI/data flow for path-based deletion
• Run baseline validation (ruff check . and pytest) to confirm starting state
• Add --badData CLI mode to target malformed Jellyfin metadata instead of path prefixes
• Implement DB query/support code to detect bad entries (missing season/episode numbering or missing versions)
• Update orchestration output so dry-run clearly previews what would be deleted in --badData mode
• Add/adjust focused tests for new CLI + DB/core behavior
• Run targeted tests for changed areas, then full lint/tests
• Run final validation and report results

[Copilot]

Pull request overview

Adds a new CLI mode (--badData) to identify and delete Jellyfin library entries that appear malformed (e.g., missing season/episode numbering or missing media versions), extending the cached SQLite schema to store additional metadata needed for detection.

Changes:

• Add --badData/--bad-data flag and route target selection to a new “bad metadata” DB query.
• Extend SQLite items schema + upsert logic to persist IndexNumber, ParentIndexNumber, and a media_source_count.
• Update README and add focused tests for the new CLI flag and DB behavior.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tests/test_db.py	Adds coverage for persisting new metadata fields and for get_bad_data_targets selection logic.
tests/test_cli.py	Verifies --badData parsing and default flag values.
README.md	Documents the new --badData mode and updates the “How it works” flow.
jellyfin_cleanup/database.py	Adds schema migration for new columns, persists metadata during upsert, and implements get_bad_data_targets.
jellyfin_cleanup/core.py	Adds --badData execution path and updated preview output.
jellyfin_cleanup/client.py	Expands /Items scrape fields to include metadata needed for --badData.
jellyfin_cleanup/init.py	Exports get_bad_data_targets at the package level.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

get_bad_data_targets treats NULL metadata as definitively “bad”. After migrating an existing cache DB (created before these columns existed), index_number, parent_index_number, and media_source_count will be NULL for most/all rows, so --badData can match nearly everything (especially due to COALESCE(media_source_count, 0) = 0). This is risky because it can drive unintended mass deletions when users choose to reuse cached data. Consider introducing a schema/user_version and forcing a rescrape (or warning+exit) when cached rows don’t have these fields populated, and/or adjusting the predicate so “unknown/not-scraped” isn’t treated as “bad” (e.g., don’t COALESCE NULL to 0 for versions).

Suggested change

	AND (parent_index_number IS NULL OR index_number IS NULL)
	THEN 'missing season or episode number'
	WHEN type='Season'
	AND index_number IS NULL
	THEN 'missing season number'
	WHEN type IN ('Episode', 'Movie', 'Video')
	AND COALESCE(media_source_count, 0) = 0
	THEN 'no media versions'
	END AS bad_reason
	FROM items
	WHERE delete_status IN ('pending', 'failed')
	AND (
	(type='Episode' AND (parent_index_number IS NULL OR index_number IS NULL))
	OR (type='Season' AND index_number IS NULL)
	OR (type IN ('Episode', 'Movie', 'Video') AND COALESCE(media_source_count, 0) = 0)
	AND (
	(parent_index_number IS NULL AND index_number IS NOT NULL)
	OR (parent_index_number IS NOT NULL AND index_number IS NULL)
	)
	THEN 'missing season or episode number'
	WHEN type IN ('Episode', 'Movie', 'Video')
	AND media_source_count = 0
	THEN 'no media versions'
	END AS bad_reason
	FROM items
	WHERE delete_status IN ('pending', 'failed')
	AND (
	(
	type='Episode'
	AND (
	(parent_index_number IS NULL AND index_number IS NOT NULL)
	OR (parent_index_number IS NOT NULL AND index_number IS NULL)
	)
	)
	OR (type IN ('Episode', 'Movie', 'Video') AND media_source_count = 0)

[Copilot]

The --badData help text doesn’t mention that it ignores any provided PATH/--target-path filters (the program does this in core). Consider updating the flag description to explicitly state that path filters are ignored in this mode so users don’t think both filters apply.

Suggested change

	help="Target entries with bad metadata (missing season/episode data or no versions).",
	help="Target entries with bad metadata (missing season/episode data or no versions). "
	"Ignores any positional PATH values and --target-path filters in this mode.",

[Copilot]

In --badData mode the target selection depends on additional scraped metadata fields; if the user opts to reuse cached data (prompt “Re-scrape?” → no, or --no-rescrape) and the cache predates these fields, the selection can be misleading/over-inclusive. Consider detecting whether the cache includes non-NULL values for the new metadata columns and forcing a re-scrape (or warning+exit) when it doesn’t.