Do I need lerna to use auto for a monorepo?

[pomber]

After reading the docs it isn't clear what happens when I run auto shipit on the root of a monorepo.

It does say it works with lerna:

The npm plugin works out of the box with lerna in both independent and fixed mode. auto works on a repo basis and should be run from the root of the repo, not on each sub-package. No additional setup is required.

But is lerna required?

(I'm currently using yarn as the 'installer' and wsrun as the "task-runner")

[hipstersmoothie]

At the moment yes it's required. In my opinion lerna is a pretty bad installer, an okay task runner, and a really good publisher. Auto uses lerna to do most of the tricky parts of publishing releases.

@zephraph had an idea to make a tmp lerna.json to get around the user actually having to use lerna

[pomber]

Thanks, it makes sense. Maybe I can run a different workflow on [pr] to add the canary comment to the pr body.

[hipstersmoothie]

If you find a nice way to do that we can add it to the GitHub actions docs

[hydrosquall]

One compelling reason to explore ways to use auto in a monorepo without needing lerna is that lerna does not support yarn 2 workspace:* links.

lerna/lerna#2564

This makes it difficult to use auto in a yarn 2 monorepo where some packages depend on others in the same monorepo, given that the workspace:* format is the recommended way to specify a local dependency. (Otherwise, you run the risk of importing a remote rather than local version), and will need to manually update all dependents any time you update the version of a core package.

https://yarnpkg.com/features/workspaces#workspace-ranges-workspace

[hydrosquall]

Drilling into this a bit more, currently the main blocking command to trying to use auto in a yarn 2 monorepo with workspace links is here:

auto/packages/core/src/utils/get-lerna-packages.ts

Line 15 in 909ae8b

const response = await execPromise("npx", ["lerna", "ls", "-pla"]);

If this line were replaced with the option to use yarn workspaces list (or some yarn 2 equivalent with flags to get version numbers / private status / etc), it might be possible to use auto without lerna (although it's possible that some other commands would fail afterwards)

[hydrosquall]

Update: I was able to use lerna and auto with Yarn 2 by replacing all my workspace:* links with explicit version ranges (like ^0.2.0)

[bguillen-tcs]

Hello @hipstersmoothie and @hydrosquall ,

I am new to auto and publishing packages, so forgive me if my question(s) are already answered above. Things are just going over my head right now. 😂 Anyway, I am still a little confused about this.

Essentially, I've managed to setup a nice proof of concept utilizing auto to publish independent packages within a monorepo. It truly works like a charm. Feel free to check it out here -- the workflow and repo setup.

Now, after taking a step back with the team, I wanted to see if we could use [email protected], as opposed to lerna for managing our monorepo packages -- essentially, publishing them independently, like how I am doing with lerna. Is this possible? To see what I am currently trying to do, checkout the repo here. The main thing there is the release workflow; I am basically running yarn install from the root, compiling TS modules, and then running auto shipit. However, it does not seem to publish properly -- the root package.json is getting the version bump, rather than actual package package.json.

I've taken a look through doc website, and have only come across monorepo documentation with lerna, here.

Anyway, is publishing independent packages within a monorepo using yarn possible? Is this something that is more of a limitation of yarn and not the auto software, or both?

Let me know if you need me to clarify anything.

All the best and thanks for this awesome tech,

Bryan