Canbench : Debug and print

Developers Rust
1

Hi Everyone

I am encountering an issue where none of the printing statements seem to work within canbench bench functions. I have tried the following:

eprintln!("Error occurred: {:?}", error);
ic_cdk::println!("Starting register_user benchmark");
println!("Starting register_user benchmark");

However, none of these statements produce any output in the logs or the debug console. Could you please help me understand why this is happening and suggest a way to enable logging or debugging for bench functions in canbench?
here is a simple bench function

#[bench]
fn process_numbers_bench() {
    ic_cdk::println!("Benchmark started..");
    let count = 10_000;
    let result = process_numbers(count);
    ic_cdk::println!("Benchmark result: {:?}", result);
}
1 Like
2

canbench currently hides all output from the canister and only outputs its own report with the benchmarking results. Can you elaborate on your use-case?

1 Like
3

Actually, I think @ielashiā€™s information might be outdated. There was a PR some time ago that added an option to allow the output of canisters to be exposed if needed. I think if you use canbench 0.1.8 and above it should allow you to get the logs.

2 Likes
4

Is this documented somewhere? It not, would it make sense to add it to make it easy for people to find it? I wonder if exposing it as an explicit arg to canbench would be clearer than setting an environment variable.

1 Like
5

Yeah I think we should document it for easier access.

I think we just went with the easiest change back then. I donā€™t know if an explicit arg is that much better, as long as things are documented it should be fine. The only marginal benefit I find is that the help message of canbench could be self-documentation if we used an explicit arg.

6

Exactly, I see that as a big benefit, as then it becomes easier to discover such features.

7

I noticed that the GitHub PR contains code that hasnā€™t been documented for use in projects yet. Iā€™m currently in the middle of my project and need to integrate that function. Would it be possible to share the implementation of that specific function for now, so I can use it in my codebase?

8

Hey folks,
Iā€™m planning to add a ā€œDebuggingā€ section to the documentation for better clarity.

Also, what if we enable ic_cdk::println! output by default, and suppress it with canbench --less-verbose? What do you think?

1 Like
9

I would be in favor of not enabling it by default, since it would be make the output of the reports less readable (the println output would be inter-mixed with the canbench output). Iā€™m not still not clear on what the benefit of this output would be. Benchmarks are not tests that you debug. Am I missing something?

1 Like
10

I thought the purpose of canbench --less-verbose was to produce a very readable reportā€¦ Ok, so we add a --show-canister-output option to enable the output then?

11

On second thought, I agree with your original suggestion (I totally forgot we already have the less-verbose option), so enabling logs by default makes sense in that case.

1 Like
12

My personal opinion is that if we enable logs by default it could be quite noisy and usually when people run canbench theyā€™re interested in the benchmark results (canbench is first and foremost a benchmarking not a debugging tool), so it makes sense for the tool to output only that by default. I would go with the default we have now and add an option to allow showing more logs for anyone whoā€™s interested.

1 Like
13

Here are the PRs:

14

Hey Vicky,
Itā€™s implemented and released in canbench v0.1.11. Now, both ic_cdk::println!() and ic_cdk::eprintln!() will appear in the console when canbench is run with the --show-canister-output argument.

1 Like
15

hey @berestovskyy ,
Thank you for the update! I appreciate the implementation in canbench v0.1.11. The addition of ic_cdk::println!() and ic_cdk::eprintln!() to the console output with --show-canister-output is really helpful.

Apologies for the delayed responseā€”I had moved on to another task and saw your message today. Thanks again for addressing this!

1 Like