Collecting metrics at the start of update methods vs at the end

[paulyoung]

The README contains this example:

#[ic_cdk_macros::update(name = "doThis")]
pub async fn do_this() -> () {
    canistergeek_ic_rust::monitor::collect_metrics();
    canistergeek_ic_rust::logger::log_message(String::from("do_this"));
    // rest part of the your method...
}

Is there any reason not to do something like this instead?

#[ic_cdk_macros::update(name = "doThis")]
pub async fn do_this() -> () {
    canistergeek_ic_rust::logger::log_message(String::from("do_this"));
    // rest part of the your method...
    canistergeek_ic_rust::monitor::collect_metrics();
}

[alexeychirkov]

Hey! Of course, you can place and call log_message and collect_metrics in any order within your code!
Thanks!

[paulyoung]

Doesn't it make more sense to collect metrics after an operation is complete though?

[alexeychirkov]

I would say it's up to you.
If your function is atomic and one of the statements (in the middle of a function) causes a trap - state will be rolled back - so order does not matter in this case.
If function is nonAtomic: once again - it's up to you how you want to calculate update calls - before first await or in the very end (state will be rolled back to closest await...)

If you are talking about memory/heap memory metrics - I didn't actually test if memory values differ in the beginning and in the end of a method...

What do you think?

[paulyoung]

My intuition is that collecting memory at the end will provide a more accurate picture of the current state of things.

For example; I would expect that collecting metrics before adding/removing data won’t reflect those changes, but collecting them after would.