microsoft / microsoft/TypeScript
Import type over generic argument
Nobody has claimed this yet.
- Dominant language
- Go
- Stars
- 111k
- Forks
- 14.3k
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 132
Description
Search Terms
import type, generic, module string type
Suggestion
To allow for giving a generic argument as an import() type's argument, something like
function <Mod extends string>(obj: unknown, modPath: Mod): obj is import(Mod);
Use Cases
Would allow generic functions that type inference from any module path string literal to the type exported by that module. This is helpful for encapsulating good module loading practices. For instance, in a large AMD application I work on, we need to manage the application footprint by not loading modules that aren't needed for the current execution path. There are a number of utility functions we could write around handling modules to support good behavior on this front.
Examples
One of the signatures of AMD's require() could be typed as:
function require<Mod extends string>(path: Mod): import(Mod);
This actually does work today if that declaration is put in a .d.ts file and skipLibCheck is set to true. If a string literal is passed, the return type is inferred from the path; if non-literal, the return type is any. In fact, if this worked but disallowed anything other than a module path, it would be super handy for typo avoidance. I'm pretty sure what's happening now is accidental, though, since any more complex use yields a "string literal expected" error on the import type.
Likewise, I'm pretty sure a really useful signature for define() could be done with a mapped type and tuple typed-rest params, something like:
type ImportsOf<Paths extends string[]> = { P in keyof Paths: import(Paths[P]) };
function define<Paths extends string[]>(paths: Paths, callback: (...args: ImportsOf<Paths>)=>void);
Specifically, the function I wanted to write with this was a type-narrowing instanceof check over a module path. Given a module that exports a constructor:
function isModuleInstance<Mod extends string>(obj: unknown, path: Mod): obj is InstanceType<import(Mod)> {
const modValue = require(path);
return modValue && obj instanceof modValue
}
If the module for a constructor hasn't been loaded, we can assume no object has been constructed by it, so this technique is very handy for keeping down the module load footprint in applications with lots of modules.
A question for discussion would be the value of an import-type on a non-literal argument, or a non-module argument for that matter. My initial impression is if the import() type can't be resolved, it should evaluate to never and produce an error just like import('some_garbage_string_i_made_up')". That might require the type checker to be given a stricter bound than "extends string", though. If necessary, a user-visible built-in type that means "a string representing a known module" seems like it might be useful in lots of contexts.
Checklist
My suggestion meets these guidelines:
- [X ] This wouldn't be a breaking change in existing TypeScript/JavaScript code
- [X ] This wouldn't change the runtime behavior of existing JavaScript code
- [X ] This could be implemented without emitting different JS based on the types of the expressions
- [X ] This isn't a runtime feature (e.g. library functionality, non-ECMAScript syntax with JavaScript output, etc.)
- [Seems like it to me] This feature would agree with the rest of TypeScript's Design Goals.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reproducing the generic import() examples in a .d.ts file with skipLibCheck enabled, including literal and non-literal module paths. Compare the current behavior with the requested signatures and define what should happen for unresolved or non-module arguments; done means the intended cases are accepted with the expected inferred types and invalid cases remain appropriately diagnosed.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- typescript
- Domain
- compilers
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100