Investigate .expect potentially pulling in format code

[ChaoticTempest]

Related to #745 (comment)

Investigate whether the following .expect call is pulling formatting code which would increase the code size:

near_sdk::borsh::BorshSerialize::try_to_vec(&result).expect("Failed to serialize the return value using Borsh.")

[sn99]

Hi @frol I would like to be assigned this issue.

[sn99]

Hi @frol, I got busy over the weekend.

Investigate whether the following .expect call is pulling formatting code which would increase the code size:

Yes it does, internally .expect(...) calls unwrap_failed(..., ...) which in turn calls our beloved panic!(...), and from the docs:

When using panic!() you can specify a string payload that is built using [formatting syntax].

This does seem to increase code size as stated in the rustwasm book.

The book mentions to use something like:

#![allow(unused_variables)]
fn main() {
#[inline]
pub fn unwrap_abort<T>(o: Option<T>) -> T {
    use std::process;
    match o {
        Some(t) => t,
        None => process::abort(),
    }
}
}

Stating:

Ultimately, panics translate into aborts in wasm32-unknown-unknown anyways, so this gives you the same behavior but without the code bloat.

But this would likely make us lose the message in expect.

On side note i was able to decrease the wasm file size by just adding:

[profile.release]
opt-level = 's'
codegen-units = 1

I can make a pr if needed.

[frol]

The book mentions to use something like:

@sn99 Well, sure, we already have near_sdk::env::panic_str() for that, but we don't want to give up on the standard features. Thinking about it now I realized that expect requires formatting to Debug-format the error itself.

How big is the impact on the Wasm binary size afterall? My guess is that near-sdk-rs pulls it a ton of code that has expect in it, so adding a user code with .expect won't change the picture much. Also, consider using https://github.com/austinabell/nesdie in addition to near-sdk-rs to compare the sizes (e.g. with nesdie I was able to build a meaningful contract in Rust that is under 1kb)

On side note i was able to decrease the wasm file size by just adding:

These are recommended options when creating a new project: https://docs.near.org/sdk/rust/get-started#create-a-new-project, but since you mention it, I believe we missed to overcommunicate it enough. Where would you want to add it?

[TropicalDog17]

Hi @frol can I work on this issue? it seems the previous assignees was inactive for a while.

[frol]

@TropicalDog17 Sure. Reassigned it to you

[TropicalDog17]

Hi @frol, after replacing std expect with near-sdk's panic_str here and compile the fungible_token example, the wasm code size is reduced from 165.4KB to 165.1KB, which is 0.3KB (0.18%) reduction of code size. I'm not sure if it is significant or not, but actually remove the expect would decrease code size.

[frol]

@TropicalDog17 That is insignificant. Please, try replacing all .expect() calls in near-sdk-rs and in the code generated by near-sdk-macros, such as:

near-sdk-rs/near-sdk-macros/src/core_impl/code_generator/impl_item_method_info.rs

Lines 122 to 127 in 0f46b0d

	&::near_sdk::env::input().expect("Expected input since method has arguments.")
	).expect("Failed to deserialize input from JSON.")
	},
	SerializerType::Borsh => quote! {
	::near_sdk::borsh::BorshDeserialize::try_from_slice(
	&::near_sdk::env::input().expect("Expected input since method has arguments.")

near-sdk-rs/near-sdk-macros/src/core_impl/code_generator/attr_sig_info.rs

Line 313 in 0f46b0d

::near_sdk::serde_json::from_slice(&data).expect("Failed to deserialize callback using JSON")

[TropicalDog17]

@frol
After replacing essentially all .expect() calls(#1220) here is the result,

a decrease of 4kB, which is 2.42% reduction of code size compare to the original code in master branch.

[frol]

@TropicalDog17 I reviewed the PR, let's continue the discussion there