FROSADO / onecloudriver / 32298087461

81%

Build Type

push

github

Committed by web-flow

Commit Message

feat(fs): configurable metadata pre-warm depth after mount (#107)

* feat(fs): configurable metadata pre-warm depth after mount (--pre-warm-depth)

Implements Issue #71: Add configurable metadata pre-warming (BFS traversal)
to prefetch folder structures up to N levels deep after mount, improving
perceived performance on first access.

Changes:
- Add PreWarmDepth to MountConfig (default 2, range 0-10)
- Add PreWarmDepth to AccountPersistedConfig (persisted in account.json)
- Add CLI flag --pre-warm-depth with validation
- Implement preWarm() BFS algorithm with context timeout (30s)
- Integrate preWarm() async call in Mount() lifecycle
- Add 11 unit + integration tests with full coverage
- Add CLI tests for persistence and defaults
- Update MANUAL.md and MANUAL.es.md documentation
- All CI checks passing: tests, lint, race detector

Closes #71

* fix(fs): honor pre-warm-depth 0 and traverse by item ID

The initial issue #71 implementation had two correctness bugs and an
off-by-one:

- The BFS enqueued the display name (the map key returned by GetChildren)
  instead of the real item ID, so only the root was ever pre-warmed.
- --pre-warm-depth 0 did not disable pre-warming: the CLI and
  DefaultMountConfig both used a >0 guard, and omitempty dropped the
  zero from the persisted JSON. PreWarmDepth is now a pointer so an
  explicit 0 is distinguished from "unset" and honored end-to-end.
- Depth now matches the documented semantics (root = level 1).

Also avoid dirtying the whole pre-warmed tree: preWarm now uses
PrefetchChildren (GetChildren without dirty marking) so the first
SerializeDirty after a bulk warm-up doesn't rewrite the entire subtree,
and the pre-warm goroutine is skipped entirely when depth is 0.

Generated with Codebuff 🤖
Co-Authored-By: Codebuff <noreply@codebuff.com>

---------

Co-authored-by: Codebuff <noreply@codebuff.com>

Coverage Stats

78 of 106 new or added lines in 3 files covered. (73.58%)

4 existing lines in 2 files now uncovered.

4567 of 5875 relevant lines covered (77.74%)

3364.44 hits per line