-
Notifications
You must be signed in to change notification settings - Fork 12.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Expose configuration hook for custom module resolution #18896
Comments
Note that this is already an option for non- I'm not sure there's anyone who would want this ability who isn't already loading TS through some more advanced build pipeline. |
@RyanCavanaugh I want to continue using VS Code as a regular editor, but also be able to use JSPM/ bower/custom webpack without doing a lot of fiddling with It would also be nice if I could more generally hook into the module resolution pipeline without having to implement a full language server and wrapping extension. |
I think this could become quite flexible and workable without having too many options. Currently typescript uses a fallback strategy when resolving modules and that forgiving behavior doesn't need to change as a result of this feature. For example, let's say I'm using moment and_bootstrap_ in my application. I can't provide a path mapping for bootstrap because of how its declaration file references jQuery's. so maybe I have installed that declaration using npm but I have moment installed with JSPM and moment comes with its own type declaration and I that is the one I want to use and I can today just manually. so when TypeScript resolves moment, I could via this feature suggest |
I've also been trying to cobble together a working vscode / typescript / jspm project and eventually came to a similar conclusion as this issue describes. My thoughts were a new compilerOption which would identify a custom resolver module that would return an array of objects, mapping containingFiles / moduleNames keys to module locations.
When should such a function be called? It could be called early on in the process and cached but if design time support is required then the map would need updated dynamically. I suspect Typescript already has a mechanism to do this. |
I believe this could also solve some issues with remote/network file systems at larger organizations (e.g. #16426). |
I'd be very interested in a hook here for jspm resolver support. |
@masaeedu I played around a bit with modifying ts.sys to use"virtual modules"/type providers. I'd be very interested if this API was flexible enough to support virtual modules in a more general way. |
Just saw this issue - I'd be quite interested as well to have such a hook. Would allow me to start making experiments to integrate PnP with Typescript 🙂 |
Would a new "moduleResolution" option make sense for this? Maybe something like |
Hi. I propose some solution in #28624 that allow to provide in compilerOptions custom file resolution function (from node module or file). It works fine with tsc and language server. |
The current situations makes this a no-go for us to use typescript. |
Any progress on this? Just had to convert a project in a monorepo to yarn-pnp to resolve a dependency problem, and because I now have no node_modules in that project, tsserver does not find any of my dependencies :( |
I was thinking about it a little bit, and given sufficient metadata about the types in the codebase, you could also use this as a sort of primitive typeclass system. As an example: type eq<a> = { equals: (x: a, y: a) => boolean }
type arrEq = <a>(a: eq<a>) => eq<a[]>
const arrEq: arrEq = a => {
const equals = (xs, ys) => all(zipWith(a.equals, xs, ys))
return { equals }
}
const strEq: eq<string> = { equals: (x, y) => x === y }
type t = eq<string[]>
const arrStr: t = require("<resolve>")
// Here we introspect over the types of things in the codebase, and
// produce a module that exports `arrEq(strEq)`
arrStr.equals(["foo", "bar"], ["foo", "bar"]) // => true Of course, this is a rather nutty idea, and maybe not the best use case for this sort of feature, but as Jurassic Park teaches us, if you can, you should. What could go wrong? |
Any move on this? |
@RyanCavanaugh This could actually solve my feature request #31703 |
@masaeedu may be you can attempt library of typescript-module-alias |
Any? 🔢 |
It would be really great to have this on TS. Webpack and Babel both support writing custom resolvers. ESLint also has, through eslint-plugin-import. There's a similar issue with some good insights on its usefulness: typescript-eslint/typescript-eslint#2771 (I actually think it should've been here as it's not typescript-eslint job to solve this but tsc). |
Year 2021, still looking forward on this. This document https://www.typescriptlang.org/docs/handbook/module-resolution.html#module-resolution-strategies now is some complex, why not just give a hook function for user to do the resolve. |
Would it be the ideal solution for nodejs to stabilize the loaders api, given that typescript's module resolution mimics node's module resolution algorithm? |
This would allow to type check one project by types from others, for example i have a I would like to import it in |
Following on from #6012 and #5039, it'd be nice to have the ability to add TypeScript to projects with heterogeneous module resolution/packaging systems and have typechecking work.
The
paths
andbaseUrl
configuration options get us a long way, but they don't deal well with module loaders that are less widely used, frequently changing, or that allow for complex module resolution rules. Letting the user entirely outsource all module resolution concerns using the configuration would allow type checking to work regardless of the user's preference of npm, bower, requirejs, browserify, jspm, webpack, etc.A simple approach might be to have a
tsconfig
option for a parametrized command line invocation (similar to npm scripts) that is able to resolve modules. Each package manager can distribute a tool that implements module resolution according to its own needs, which the users can install and point to in theirtsconfig.json
.Exactly how extensive the API needs to be is up for debate. It could be something as simple as a function that maps requested module names to file system paths, or a more abstract implementation of something like
System
from here, which could open up more interesting use cases (e.g. "virtual" modules built using code generation, or F# style type providers).The text was updated successfully, but these errors were encountered: