[Dev Tools] @matterlabs/hardhat-zksync-node slow performance in hardhat tests

[mmv08]

Team or Project

Safe

ZK chain

Era

Environment

Mainnet

Select the Dev Tool you are using

Hardhat Plugins

Provide the version of the tool (if applicable)

"@matterlabs/hardhat-zksync-node": "^1.5.1",

Provide a brief description of the functionality you're trying to implement and the issue you are running into.

First of all, working with the node plugin has been a nightmare. Breaking changes were introduced under a minor version without warning, forcing us to spend significant time just getting our test setup to work. But even that pales in comparison to its performance issues relative to Hardhat’s node.

Running the same tests with Hardhat’s node takes only 5 seconds on my machine.

Running tests with zksync node takes 11m 48s.

Same test suite.

I can reproduce this behaviour consistently across macOS, Linux, and WSL, so I don't believe machine specs are the issue, but let me know if they would be helpful.

Repo Link (Optional)

https://github.com/safe-global/safe-smart-account

Additional Details

to reproduce, run:

1. npm run test for hardhat tests
2. npm run test:zk for zksync tests

compare run times

[dutterbutter]

Thanks for posting, and apologies for the frustration endured using hardhat-zksync-node. I will investigate the above issue and see if I can provide a solution or fix.

If you were not aware, we delivered EVM equivalence in the v27 upgrade so you can make use of L1 tooling (e.g. hardhat) without needing any of the custom zksync hardhat plugins. The EVM experience will also be further improved when ZKsyncOS lands.

[mmv08]

Thanks, we had two production contract versions released pre-EVM compatibility, and we would like to keep the tests for them to test flows such as upgrading the contract from zksync one to EVM one, etc

[dutterbutter]

Okay so I did some fixes that results in a ~1.8x speed up of the hardhat-zksync-node hardhat test runner and properly sets --quiet param so we no longer spill tx summary output etc to stdout.

Once these 2 PRs (matter-labs/anvil-zksync#729, matter-labs/hardhat-zksync#1790) land, and released you will see an improvement.

That being said, there is still one area we can get an even more speed up (batch closing) but will require additional refactoring that has no immediate time table but is on the radar.

I'm going to mark this as answered for now, feel free to re-open if there is more you'd like to resolve / discuss or open a new discussion that will be triaged accordingly. Really appreciate you posting! This is how we improve, albeit incrementally 🙏