Integrate bencher for benchmarks

[tusharmath]

The idea here is to integrate Bencher by @epompeii and keep track of performance.

It would be great if we could push data of older commits for the last 2-3 months and visualize it on something like — https://bencher.dev/perf/advent-of-code

[tusharmath]

/bounty 50$

[algora-pbc]

## 💎 $50 bounty • Tailcall Inc.

### Steps to solve:
1. Start working: Comment /attempt #1300 with your implementation plan
2. Submit work: Create a pull request including /claim #1300 in the PR body to claim the bounty
3. Receive payment: 100% of the bounty is received 2-5 days post-reward. Make sure you are eligible for payouts

🙏 Thank you for contributing to tailcallhq/tailcall!
🧐 Checkout our guidelines before you get started.
💵 More about our bounty program.

Attempt	Started (GMT+0)	Solution
🔴 @neo773	Mar 8, 2024, 6:33:24 AM	WIP
🟢 @murtaza030	Mar 10, 2024, 6:52:02 AM	WIP
🟢 @alankritdabral	Mar 10, 2024, 1:19:24 PM	WIP

[neo773]

/attempt #1300

Algora profile	Completed bounties	Tech	Active attempts	Options
@neo773	31 tailcallhq bounties + 61 bounties from 18 projects	TypeScript, Rust, JavaScript & more		Cancel attempt

[algora-pbc]

@neo773: The Tailcall Inc. team prefers to assign a single contributor to the issue rather than let anyone attempt it right away. We recommend waiting for a confirmation from a member before getting started.

[epompeii]

Thank you for following up @tusharmath!
This is a follow on issue from a discussion on #436

I will create a PR that runs on pushes to the main branch, following the first example shown here: https://bencher.dev/docs/how-to/github-actions/
A working example of this approach is also available here: https://github.com/bencherdev/example/blob/main/.github/workflows/benchmarks.yml
The perf page for that example is here: https://bencher.dev/perf/example

In order for this to work, you all need to have BENCHER_API_TOKEN set as a repository secret.
It will be a bit tedious to add accurate historical data. Though the Bencher API supports backdating reports, getting that historical data would require replaying each past push to main in GitHub Actions in chronological order.
With that said, the PR I'll be posting will be forward looking only.

Currently you all have implemented relative benchmarking, comparing a single run on main to a single run on a PR here: #762
This can also be accomplished with Bencher as well, and this first PR will lay the ground work for you to use more advanced Threshold for detecting performance regressions.

[murtaza030]

/attempt #1300

Options

• Cancel my attempt

[algora-pbc]

@murtaza030: The Tailcall Inc. team prefers to assign a single contributor to the issue rather than let anyone attempt it right away. We recommend waiting for a confirmation from a member before getting started.

[alankritdabral]

/attempt

Algora profile	Completed bounties	Tech	Active attempts	Options
@alankritdabral	1 tailcallhq bounty + 1 bounty from 1 project			Cancel attempt

[algora-pbc]

@alankritdabral: The Tailcall Inc. team prefers to assign a single contributor to the issue rather than let anyone attempt it right away. We recommend waiting for a confirmation from a member before getting started.

[alankritdabral]

Thank you for following up @tusharmath! This is a follow on issue from a discussion on #436

I will create a PR that runs on pushes to the main branch, following the first example shown here: https://bencher.dev/docs/how-to/github-actions/ A working example of this approach is also available here: https://github.com/bencherdev/example/blob/main/.github/workflows/benchmarks.yml The perf page for that example is here: https://bencher.dev/perf/example

In order for this to work, you all need to have BENCHER_API_TOKEN set as a repository secret. It will be a bit tedious to add accurate historical data. Though the Bencher API supports backdating reports, getting that historical data would require replaying each past push to main in GitHub Actions in chronological order. With that said, the PR I'll be posting will be forward looking only.

Currently you all have implemented relative benchmarking, comparing a single run on main to a single run on a PR here: #762 This can also be accomplished with Bencher as well, and this first PR will lay the ground work for you to use more advanced Threshold for detecting performance regressions.

Hey @epompeii i saw your comment and created a 1367, i took a little bit inspiration from 3849. I would be grateful if you could have a look to the pr.

[epompeii]

@tusharmath I have create a pull request to add Bencher to track the performance of the main branch in the tailcall CI: #1441

I very much appreciate the generous offer of a bounty. My goal in doing these sorts of integrations is to spread the word about Bencher. All that I ask in return is that you help spread the word in your own way, however large or small 😃

[epompeii]

I'm aware of the failure occurring here: https://github.com/tailcallhq/tailcall/actions/runs/8332008049/job/22800159465#step:5:1
And I'm investigating.

Reverted in the mean time by: #1505

[epompeii]

@tusharmath I'm sorry for the delay getting back to this and my failure to fully test #1441, that is totally on me.

The reason for the delay is that I was heads down working on Bencher v0.4.5. This new version completely refactors the way that branch selection is handled and it deprecates some of the options that I would have used here. I wanted to finish that work before proceeding so I didn't add options that were going to be deprecated a few days later.
Plus, the new way of doing things is much more elegant!

With that said, it looks like @alankritdabral has done a pretty nice job adding Bencher in #1679. Though he is using the now deprecated options/way of doing things.
I have left a review on the PR for the things that need to be updated: #1679 (comment)
If he is willing to update things, then using his PR is alright with me. Otherwise, I will do my best to get this knocked out 😃

[tusharmath]

Thanks! Would love to see this integrated into Tailcall soon.

[alankritdabral]

@tusharmath I'm sorry for the delay getting back to this and my failure to fully test #1441, that is totally on me.

The reason for the delay is that I was heads down working on Bencher v0.4.5. This new version completely refactors the way that branch selection is handled and it deprecates some of the options that I would have used here. I wanted to finish that work before proceeding so I didn't add options that were going to be deprecated a few days later. Plus, the new way of doing things is much more elegant!

With that said, it looks like @alankritdabral has done a pretty nice job adding Bencher in #1679. Though he is using the now deprecated options/way of doing things. I have left a review on the PR for the things that need to be updated: #1679 (comment) If he is willing to update things, then using his PR is alright with me. Otherwise, I will do my best to get this knocked out 😃

Thankyou for your review means a lot ❤️ . Almost done with the changes!

[alankritdabral]

@tusharmath @epompeii done with the changes. #1679 is good to go.

[epompeii]

Okay, lets finally get this knocked out!
@tusharmath I have created a new pull request: #1725
Again, my apologies for the hubris to not fully test things out the first time around. 🤦🏽‍♂️

Unfortunately, after taking a closer look at @alankritdabral PR #1679, I noticed some fundamental security vulnerabilities in his approach for handling PRs from forks. Bencher has built in defenses that probably would have prevented you all from getting pwn requested, but it would have broken your build (again). So I just went ahead and got things implemented as recommended in the docs: https://bencher.dev/docs/how-to/github-actions/#pull-requests-from-forks

[github-actions]

Action required: Issue inactive for 30 days.
Status update or closure in 7 days.

[epompeii]

@tusharmath is there anything else needed to close this out, now that #1725 has merged?

[github-actions]

Action required: Issue inactive for 30 days.
Status update or closure in 7 days.

[github-actions]

Issue closed after 7 days of inactivity.