I have set up a Node.js (Typescript) based Monorepo, and I am using Turborepo as the tool to manage the Monorepo. I am using pnpm as my package manager.
All my main projects are present in the apps folder, and reusable components that I want to share across projects are present in the package folder.
My project folder structure is as follows:
apps/
--- project 1/
---package.json
---app/
---node_modules/
...
--- project 2
---package.json
---app/
---node_modules/
...
packages/
---platform/
---package.json
---app/
---node_modules/
...
node_modules/
package.json
turbo.json
...
My code is written, and the build and lint are working as expected locally. I want to now proceed to deploy project 1 on my production system on AWS Lambda, and I tried with both Gitlab CI and AWS CodePipelines - My issues are similar.
My problem is that many of the dependencies are organised as symbolic/hard links. For example, project 1 depends on platform, but this dependency appears as a symbolic link. To publish project 1 to Lambda, I need it to be a standalone project with all dependencies present within the node_modules folder. Links are not useful when I zip project 1.
node_modules folder using cp -rL. However, that didn't seem to work with some errors:cp: cannot overwrite non-directory 'apps/omnibus-comments-handler/node_modules/././typescript' with directory 'node_modules/./typescript'
However, even then I am not sure if that's the right solution as it copies unnecessary dependencies.
npm instead as it installs dependencies in a flat style - However, even that doesn't seem to work due to similar challenges. For example, platform is a workspace dependency and is a symbolic link.My objective is to run a "build" command in such a way that all the dependencies are present in the node_modules folder for project 1 so that I can zip it and deploy it to Lambda. What is the best way to achieve this?
Please try:
turbo prune
I think it's made precisely for this purpose.