Performance

For controlled, versioned before/after measurements (including what didn’t improve and how each run was measured), see the benchmark log. The numbers below are a quick reference.

Measured on Kubernetes (31,412 files) at 636d292, on an Apple M4 with a warm filesystem cache (medians from the controlled run in the benchmark log):

Command	Time	Peak memory
sprout	0.23 s	31 MB
sprout --json	0.33 s	29 MB
sprout --entry	0.57 s	302 MB
sprout --ai	0.72 s	310 MB

Why it’s fast

• Only the needed metadata. A plain tree never reads file sizes or times; they’re read only for --size, --sort, --changed-within, --json, --stats, --ai and --entry.
• Parallel parsing. Source files for the code graph are parsed across all CPU cores.
• Tests and vendored code are skipped. They never count toward the ranking, so they’re never read.
• Buffered output. Printing a 30,000-line tree is one stream, not 30,000 writes.

Measure it yourself

git clone --depth 1 https://github.com/kubernetes/kubernetes /tmp/k8s
cd sprout
SPROUT_BENCH_DIR=/tmp/k8s go test -run '^$' -bench .