# TODO
## StandardConfig refactoring (DONE)
All config structs now embed `StandardConfig` via `#[serde(flatten)]`.
## Cache cleanup
- Remove old DB cache code: `CacheEntry`, `OutputEntry`, `get_entry`, `has_cache_entry`, `get_cached_input_checksum`, `CACHE_TABLE`. These are legacy from the pre-descriptor system. `has_cache_entry` (used in status display to distinguish "stale" vs "new") should use the descriptor system instead. ~80 lines of dead code.
- Remove `cache_key()` method from `Product`. Only used by `has_cache_entry` and `remove_stale`. Once `has_cache_entry` is migrated to descriptors, it may become fully unused.
- Split db.redb: the configs table (`CONFIGS_TABLE`) is still in the same DB as the now-unused cache table. Give configs its own file (`configs.redb`), then delete `db.redb` entirely.
## Cache correctness
- Implement `output_depends_on_input_name` flag. Documented in `docs/src/cache.md` but not implemented. Needed for processors that embed the input filename in their output (e.g., a `// Generated from foo.c` header). Without it, renaming such a file would produce a cache hit with stale content.
- Write a test for identical content processed by different processors. Verify two different processors processing the same file get separate cache entries (the processor name is in the descriptor key).
## Code consolidation
- Inline single-use `names` constants. 20+ constants in `processors::names` are used in exactly one place each (their processor's `new()` call). Inline them as string literals.
- Clean `processor_configs.rs`. Still 2,100+ lines. Check for:
- ClangTidyConfig is nearly identical to StandardConfig — could it become a type alias?
- Unused `default_*` helper functions left over from cppcheck removal.
- Other config structs that are structurally identical to StandardConfig.
## Documentation
- Add `docs/src/processors/creator.md` — per-processor documentation for the Creator processor, like the other processor docs.
## Housekeeping
- Remove the `tar` lockfile entries. The crate was added and removed, but `Cargo.lock` may still reference it.
- Commit everything. There is a large amount of uncommitted work spanning:
- HasScanConfig trait elimination
- SimpleGenerator (14 generators collapsed to data-driven)
- Creator processor (new processor type with multi-dir caching)
- Cache redesign (descriptor-based, content-addressed keys, no DB for cache data)
- Checksum cache centralization (moved mtime logic to `checksum.rs` with own DB)
- MassGenerator → Creator type rename
- `ProcessorType` enum with strum iteration
- `processors types` CLI command
- `--no-mtime-cache` CLI flag
- Mandatory `supports_batch` on all processors
- Checker consolidation (5 checkers → SimpleChecker entries)
- Removed unused `dirs` crate
- New documentation: cache.md, checksum-cache.md, processor-types.md