Crates.io | only_every |
lib.rs | only_every |
version | 0.1.0 |
source | src |
created_at | 2022-02-12 22:43:33.993943 |
updated_at | 2022-02-12 22:43:33.993943 |
description | A simple rate-limiter macro: only_every!(Duration::from_millis(200), expensive_expression) |
homepage | |
repository | https://github.com/ahicks92/only_every |
max_upload_size | |
id | 531524 |
size | 11,890 |
There are lots of rate-limiting crates that do lots of things, but sometimes you just want to evaluate an expression once every whatever interval, for example rate-limited logging. This crate exposes a macro:
only_every!(Duration::from_millis(100), expensive_thing)
This macro will evaluate the given expression at most every duration rounded up to the next ms. The limiter is global: if you use 20 threads, it's still only going to happen once every duration. The expression is not inside a closure and is consequently somewhat more transparent to the borrow checker (This probably doesn't matter to you; you can read it as "there aren't edge cases").
Enable the quanta
feature for faster time-keeping at the cost of using the
quanta crate, which requires a calibration period when executing the first
rate-limited expression and an O(1) heap allocation. If you use quanta for
other things, whoever gets there first handles the calibration.
If you need a bit more, e.g. storing these in structs instead of using the
macro, there is also a OnlyEvery
type. I suggest
something like governor if you need
more than "execute this once every x".
For completeness, internally we hold times in an i64 as ms since the process
started. Behavior is undefined if your process runs for long enough that pt + interval > i64::MAX
where pt
is the uptime of the process and units are in
ms. In other words, let me know if you have a billion years of continuous
uptime and I'll fix it for you.