compilerOptions.useUnknownInCatchVariables introduces by #967

[stefanpenner]

re: compilerOptions.useUnknownInCatchVariables introduces by #967

Do we:

a) prefer to keep this disabled
b) update our code to correctly handle catch (e: unknown) {

---

I like (b) ... but it is more work, for example:

TS docs suggest e instanceof Error unfortunately, especially with more then one vmcontext there can be more then one error instance, and more technically today we duct-type these caught variables, changing them to a subclass model may not always be possible / work / etc.

One approach (type guard):

isWhatWeWant(e: unknown) : e is { code: string } {
  //... implementation
}

catch (e) {
  if (isWhatWeWant(e)) {
    if (e.code === 'ENOENT') {
       // do stuff
    } else {
       // do other stuff
    }
  } else {
    throw e; // we have no clue what is up
  }
}

Another approach: catch(e: any) { } loose and goosy

There are other approaches, WDYT?