[joelhawksley]

What are you trying to accomplish?

Reduce repeated filesystem work during cold cache digest computation.

Measurement	Before	After
Cold digest	97.6 ms	16.4 ms
Cold digest / partial digest ratio	192.819	31.670
ViewComponent boot work	654 ms	645 ms

Three paired runs, 15 passes per run, Ruby 4.0.7 / Rails 8.1.4. The cold-digest ratio improves 83.6%; rendering ratios stay within about 1%. The small boot difference is within noise.

What approach did you choose and why?

Reuse the existing boot-time directory and source-file caches while building a digest tree. Make nested cache scopes preserve the outer cache, and memoize Ruby-file discovery and file hashes within one digest computation. Nothing persists across computations or reloads.

Anything you want to highlight for special attention from reviewers?

Cold-digest allocations increase from 69,544 to 77,783 (about 12%) in exchange for fewer filesystem operations. Regression coverage checks nested cache cleanup, shared scans and hashes, unchanged digest values, and invalidation.

Benchmark reproduction

Uses the unchanged performance/measure.rb and performance/compare.rb scripts on this branch, with a separate checkout of the base revision:

for round in 1 2 3; do
  VIEW_COMPONENT_LIB=/path/to/base/lib bundle exec ruby performance/measure.rb "tmp/performance/base/$round.json"
  bundle exec ruby performance/measure.rb "tmp/performance/head/$round.json"
done
bundle exec ruby performance/compare.rb tmp/performance/base tmp/performance/head

The comparison pools individual scenario/reference ratios across processes; the absolute times above are medians of the three run medians.