Recursive types error: Type instantiation is excessively deep

[vlazh]

Bug Report

I had got error Type instantiation is excessively deep and possibly infinite. in my JSON types - value to JSON transformation.
It's some new limitation introduced from v4.3.2?
Works fine with v4.2.4.

🕗 Version & Regression Information

• This changed between versions 4.2.4 and 4.3.2

⏯ Playground Link

I couldn't get to make a short and simple example because when I removing any recursive type reference then error will disappeared.

Playground link with relevant code

[andrewbranch]

Re-opened from #44281, possibly related: #45632, #45576

[andrewbranch]

@weswigham didn't we start doing more to relate conditional types around that time frame? I seem to recall some increase in these kinds of errors was expected. But this example takes a super long time to even finish checking, and it’s nearly instant in 4.2.

[weswigham]

We did definitely add rules for relating conditional type constraints around then, I think.

[andrewbranch]

#37208?

[andrewbranch]

Indeed, reverting that “fixes” this issue.

[jakearchibald]

I think this issue is impacting a couple of my libraries, as the IDBValidKey type references itself.

Here's the pattern that's causing a problem. It isn't a problem in 4.3.5.

[andrewbranch]

Spoke with @weswigham about this and he mentioned we might be able to apply a targeted fix by swapping the order of checks for conditional types.

[vlazh]

It looks like the issue was fixed by the version 4.6.2. Isn't it?

[jakearchibald]

This example still fails in 4.6.2. Not sure if that's intended.

[vlazh]

This example still fails in 4.6.2. Not sure if that's intended.

I found two deep recursions: IDBValidKey and IsExact -> DeepMakeRequiredForIsExact.
If splits IDBValidKey into two types the error goes away. But they not the same. Not allowed IDBValidKey[][][]... Example.

[jakearchibald]

But this seemed to work fine in 4.3.5

[hasgar-aot]

any fix? it's kind of annoying
v4.0.3
fails in 4.3.5 too.