type assertions affect the type narrowing of the subsequent code

[a145789]

🔎 Search Terms

Type Assertion type Narrowing

🕗 Version & Regression Information

• This changed between versions 5.3.3 and 5.4.3

⏯ Playground Link

https://www.typescriptlang.org/play?ts=5.4.3#code/JYWwDg9gTgLgBAbzjAnmApnAwhcEB26+MANMmpgAoA2ArgObD5wC+cAZlLnAOQButdDwBQwgPRjyGbHAC82XJELEA2j3wBDEEIC6w9AA9IsKZhx5lMAOrAYACwCS+AM4wN1agB4AKgD45cN5wAGRwNAxM+kbQ8KjSNvZOru7U4YzM8kgA+mBQ6OzABgD8AFxwrlBM9Kyihsbw7LT4AMYwwARwAO62ji5uHp5YcIYwRAAmzgoWRKSBwwaj+BNwCb3JHmlMvgAUwnBwzYoEM2VDAD7YKjoke8gaUPToMKWBwgCUZRVViLeHfQdHSyrJL9agBbZuB5POBFIoA6bEN5wDSTcxKGbAvopHy+X4EVzw9HESbyACCUCgGhQADpgM5yZSUNtDgiYEi4SyifAyipOcdiHo8f8kJptKwAnzLM4VAAGHRwW7Adhwbai9BIhC3fZ5GC0KDMNW3FiibVPPXMABEFuExuEQA

💻 Code

import { type Component, type Plugin } from 'vue'

// type C = Component['name']
export type ComponentWithInstall<T> = T & Plugin
export type WithInstallPlugin = { _prefix?: string }

export function withInstall<C extends Component, T extends WithInstallPlugin>(
  component: C | C[],
  target?: T
): string {
  const componentWithInstall = (target ?? component) as ComponentWithInstall<T>
  const components = Array.isArray(component) ? component : [component]

  const { name } = components[0] 
  if (name) {
    return name
  }

  return ""
}

🙁 Actual behavior

the name property is underlined in red, indicating an error.

Property 'name' does not exist on type 'Component'.(2339)

The type C = Component['name'] is valid, as the name property exists within the Component type. When I remove const componentWithInstall = (target ?? component) as ComponentWithInstall<T>, everything returns to normal. Alternatively, if I uncomment type C = Component['name'], even though it's not used anywhere, the error disappears. Type assertions should not affect the type narrowing of subsequent code, but an error occurs. This code worked fine in version 5.3.3, but started to error in version 5.4.3.

🙂 Expected behavior

Ts don't report the error.

Additional information about the issue

No response

[MartinJohns]

This is a crash

This is no crash. Crash means the compiler crashes, not you receiving an error.

This changed between versions 5.3.3 and 5.3.4

You probably mean 5.4.3? There is no version 5.3.4.

[a145789]

This is a crash

This is no crash. Crash means the compiler crashes, not you receiving an error.

This changed between versions 5.3.3 and 5.3.4

You probably mean 5.4.3? There is no version 5.3.4.

Excuse my carelessness; I have made revisions.

[ahejlsberg]

That's very strange. I'll take a look.

[jcalz]

crossref Will type assertions affect the type narrowing of the subsequent code?

[ahejlsberg]

The issue here appears to be inconsistent caching of union and intersection property lookups, specifically the caching logic in getUnionOrIntersectionProperty and its handling of the skipObjectFunctionPropertyAugment parameter. It doesn't have anything to do with narrowing.

[ahejlsberg]

I see what's happening. For various reasons we have two caches for resolved property lookups on unions and intersections, one that includes property augmentations from Object and Function, and one that doesn't. Logic was added in #54753 to propagate entries from the non-augmented cache to the augmented cache (see here), but that's incorrect when properties are marked with CheckFlags.Partial. I'll put up a PR with a fix.