This is a simple, thread-safe, count-based, progress logger.
proglog
hooks into your existing log
implementation (i.e. env_logger
) and will output a log message every unit
number of items it has seen.
There are two primary methods, record()
and record_with(...)
.
record()
simply increments the counter and will cause a log message to output when counter % unit == 0
.
record_with(Fn() -> impl Display)
takes a function that outputs anything implementing display which will be appended to the log message.
Add to your deps:
bash
cargo add proglog
Please see the rayon example.
```rust use proglog::ProgLogBuilder;
// Note a log
backend needs to be globally initialized first
envlogger::Builder::fromenv(envlogger::Env::default().defaultfilter_or("info")).init();
let logger = ProgLogBuilder::new().build(); for i in 0..10000 { logger.recordwith(|| format!("Logged item: {}", i)); }
// The logger will flush when it is dropped, writing a final progress message no mater the count. // Alternatively you can call .flush() or .flush_with(). ```
If unit
is too small, and your loop is too tight, this will output many log messages which will slow your program down in the same way any logging would slow a program down in a hot loop.
If unit
is sufficiently large, this should be safe to put in a hot loop as all it does increment update an atomic u64
.
If your loop is tight, unit
is small, and you are using rayon / updating from multiple threads your log messages may end up out of order.
There is no guaranteed ordering of the submission of the log message to the logger.
So thread A could hit the first unit
break, thread B could hit the second point at the same time, but thread B gets to submit its log message first.
Having sufficiently large unit
will mitigate this, but you should not be depending on the log output order here.
The tradeoff made is for speed of incrementing so this can be put in hot loops over guaranteed output ordering.
bash
cargo test