Run self-contained test files in parallel with Nu’s par-each

In my LeetCode grind each solution file is self-contained: the function, its unittest cases, and an if __name__ == "__main__" block that runs the tests — so uv run solution.py runs that file’s suite. With a folder full of them, I ran them one at a time in a loop:

for $file in (glob *.py) { uv run $file }

That works, but every iteration waits for the previous one. par-each is the parallel variant of each — it runs the closure across threads and auto-detects how many cores your machine has:

glob *.py | par-each { |file| uv run $file } | ignore

On five files the speedup was nearly 2×:

Why it’s faster

for is strictly sequential: process 1 finishes before process 2 starts, so the total time is the sum of all the individual runs. par-each fans the closures out over your CPU cores, so the five independent processes run at the same time and the wall time collapses toward the slowest single run. The | ignore at the end discards the returned results — each uv run still prints its test output to stdout directly.

Gotchas

• Order is not preserved. par-each hands work out in parallel, so output comes back in completion order, not file order. Keep each when sequence matters.
• Only for independent work. These files share nothing — no ports, no shared state — so parallel is safe. If the scripts contended for one resource, the speedup would turn into flakiness.
• Cap the fan-out. par-each defaults to one thread per core; use par-each --threads N to limit concurrency.
• The win is process overlap, not parallelism of a single computation. The speedup comes from overlapping interpreter startup and test runs across files.

Related

• nushell-splat-docker-stop — another nushell pipeline trick, batching instead of looping