lerna VS changesets

On Netlify's enterprise tier, approximately 46% of builds are monorepos, with the majority leveraging Nx and Lerna. Recognizing this trend, Netlify has focused on enhancing the setup and deployment experiences for monorepo projects. In particular they worked on an "automatic monorepo detection" feature. When you connect your project to GitHub, Netlify automatically detects if it's part of a monorepo, reads the relevant settings, and pre-configures your project. This eliminates the need for manual setup. This feature also extends to local development via the Netlify CLI.

This is the exact problem monorepos were born to solve. Not only will a monorepo let you share UI components, you'll be able to gradually add shared application logic as well (for instance, do all of your apps have their own logic for connecting to a database? you could roll that into a shared library with a monorepo). There are a lot of tools for accomplishing this in JS, but probably the most popular is lerna, which is built on top of NX (though lots of teams roll their own monorepo in nx without lerna, which IMO is a totally valid option).

To begin, you need to prepare your environment. A few ways to build a React package include tools like Bit, Storybook, Lerna, and TSDX. However, for this tutorial, you will use a zero-configuration bundler for tiny modules called Microbundle.

If so, are you using a monorepo tool like Nx or Lerna? If not, start there and see if it solves your problem.

I mean I don't know what your monorepo looks like, but for example infernojs (actually written with typescript) uses lerna, and lerna seems simpler than typescript references

Tools for publishing, such as Lerna (when using the --conventional-commit flag), follow this convention when incrementing package versions and generating changelog files.

We could improve part of this by using something like Lerna. With the right configuration, Lerna can be really helpful.

It feels dirty to be adding express as a dependency to a react project, but I'm pretty sure TRPC requires all of client and server code to be in the same node.js project, since types are shared. I've read that you can use a tool like Lerna to share types between node projects, but it requires a build step, which would diminish the benefits of TRPC.

For more information, check out the PR

For versioning we use changeset, but with fixed-packages. We have found that the user thinks updating and managing our dependencies much easier if they can just bump each package to the same version. (storybook, sanity follows this appoach). While it feels unintuitive that each package gets a major if just the CSS has breaking changes, the user don't really care. They just want to finish the chore of updating their dependencies fast, without having to understand the system that deeply. This also makes documenting the changes much easier!

Thanks for sharing this. Why should I choose that instead of Changesets that work like a charm?

https://github.com/changesets/changesets works well for generating changelogs.

I found a tool named Changesets for versioning and publishing monorepo in PNPM’s docs. After some study on its philosophy, I believe it’ll be a good fit for monorepos:

The CI/CD pipeline is based on GitHub Actions and Changesets

This can be a lot of work — particularly maintaining a useful changelog. Which is why I use Changesets, which — together with its great GitHub Action — does pretty much all of this work for me.

Even better, with the help of GitHub Actions and Changesets, you can automatically manage versioning and publish the UI library to NPM.

Was the CI to slow before? Have you ever considered using something like changesets?

Use a library to generate automatic changelogs every time your code is updated. Chakra UI used Changesets.