Remove custom __getClass method

[saghul]

No description provided.

[bnoordhuis]

👍

[chqrlie]

Why not use a.constructor === RegExp or possibly a?.constructor === RegExp ?

[bnoordhuis]

That won't work if they're from different realms/contexts, right?

[chqrlie]

Why not use a.constructor === RegExp or possibly a?.constructor === RegExp

I guess we could use a.constructor.name === 'RegExp' or a?.constructor?.name === 'RegExp' then.

A simpler way to test this is a instanceof RegExp. RegExp should be the original value of globalThis.RegExp, like the other builtin object constructors. If these constructors are different in a different realm, using the constructor name solves this for actual instances. Too bad if the user refines a RegExp constructor that does something incompatible.

Same problem with a instanceof Date.

[saghul]

I followed what seems to be the most used convention to test for Regex objects.

[chqrlie]

I followed what seems to be the most used convention to test for Regex objects.

OK, classical idiom I suppose. I tend to prefer a simpler idiom that does not involve a function call nor allocate memory.

[bnoordhuis]

Right, but this is in the repl where performance isn't much of a concern.

[chqrlie]

Right, but this is in the repl where performance isn't much of a concern.

I still suggest a?.constructor?.name === 'RegExp' and a?.constructor?.name === 'Date'`

[saghul]

Changed it to instanceof like we do for Date. It's a repl, so we don't need to go the extra mile IMHO. If the user wants to monkey around with objects, more power to them.

[chqrlie]

Changed it to instanceof like we do for Date. It's a repl, so we don't need to go the extra mile IMHO. If the user wants to monkey around with objects, more power to them.

Ben's objection was that objects created in a different realm would not be instances of the same constructors. This is correct, but we do not have support in repl for different realms at the moment.

V8 has a global object Realm with functions to create and handle different realms. This seems to be an extension that is not proposed for tc39. Should we implement it?

[saghul]

Any chance ShadowRealm is close to that? https://github.com/tc39/proposal-shadowrealm

[chqrlie]

Any chance ShadowRealm is close to that? https://github.com/tc39/proposal-shadowrealm

It goes in the same direction, but with a different approach: different object/function names and semantics

[saghul]

Since the old one never got anywhere I'd say we shouldn't support it.