How is GNU `yes` so fast?

Re:How is GNU `yes` so fast? by metaphorm on Tuesday June 13, 2017 @10:55AM (#14545194)

I thought this was a fascinating read but it left a serious question lingering in my mind, which is a little out-of-scope for the article, but I hope someone here can address.

Why did the GNU developers go to such lengths to optimize the yes program? It's a tiny, simple shell utility that is mostly used for allowing developers to lazily "y" there way through confirm prompts thrown out by other shell scripts.

is this a case of optimization "horniness" (for lack of a better word) taken to its most absurd extreme, or is there some use case where making the yes program very fast is actually important?

[ Parent | on HN ]

Re:How is GNU `yes` so fast? by kevingranade on Tuesday June 13, 2017 @07:51PM (#14549690)

The stated use case for the perf improvement was "yes(1) may be used to generate repeating patterns of text for test inputs etc., so adjust to be more efficient."

Source: https://github.com/coreutils/coreutils/commit/35217221c211f3...

I've personally used it for generating repeating text and filling disks in system testing, so I appreciate it being faster at those tasks. I also sometimes use it as a signal generator for a hacky load generator, like so:

  yes | xargs -L 1 -P NUM_PROCESSES -I {} curl SOME_TARGET_URL > /dev/null

This doesn't benefit from being faster per se, but I appreciate it using less CPU since I want to give curl as much system resources as possible.

[ Parent | on HN ]