"Special" intersections (such as `string & {}`) are no longer order-independent since 4.8

[Andarist]

Bug Report

🔎 Search Terms

string autocomplete catch all intersection special

🕗 Version & Regression Information

• This changed between versions 4.7 and 4.8

⏯ Playground Link

Playground link with relevant code

💻 Code

declare function a(arg: 'test' | (string & {})): void
a('') // try string completions here
declare function b(arg: 'test' | ({} & string)): void
b('') // try string completions here

🙁 Actual behavior

Since 4.8 the second function doesn't provide completions despite using an intersection with the same members as the first one.

🙂 Expected behavior

I would expect both to behave the same, regardless of the order of the intersection members.

[fatcerberus]

Was looking for the dupe for like 10 minutes because I knew this had come up very recently. Felt stupid when I finally realized it was your own PR.

[andrewbranch]

I’m pretty sure I recall Anders saying this was intentional, though #49119 doesn’t say outright.

[Andarist]

Ye, at the end of the day - I don't think this is a big deal. However, I actually lost like 15 minutes of my life, scratching my head why this thing doesn't work - only to realize later that I've written those two in reverse order. A general rule of thumb is that the order of the intersection shouldn't matter, right?

[jakebailey]

A general rule of thumb is that the order of the intersection shouldn't matter, right?

This is what I thought, and I spent a bunch of time trying to make them sorted, but it does matter for overload ordering and that can affect basically anything. I can't think of any reason for it to matter for explicitly just string & {} though.

That being said, it seems cheap to just compare them both ways.

[RyanCavanaugh]

@ahejlsberg any specific thoughts?

[RyanCavanaugh]

Very close to making a specific "Not sure if bug, but Andarist can certainly send a PR if interested" label...

[jakebailey]

Too late: #52782

[ahejlsberg]

I don't feel strongly about fixing this, but the fix certainly looks fine to me.

[ahejlsberg]

That said, we should of course run the full battery of tests on the fix to make sure it doesn't break anything.