Implement String prototype methods about boa

[neeldug]

@jasonwilliams I believe this list is missing replaceAll.

[calluw]

I have started to write some implementations for concat() and repeat() but have encountered an issue: currently none of the String prototype methods work, even existing charAt and charCodeAt:

Using:

let a = "Hello world".chatAt(2);
console.log(a);

Results in the buffer holding only: "Error: undefined"

Is this a known issue or needs to be resolved before prototype methods can be added?

[jasonwilliams]

Hey @CallumQuick hmm that’s interesting

could you try doing new String(“Hello world”) instead of “Hello world”.charAt() ? I think that may work.

When you make a string literal and immediately use a string prototype method on it, you’re using a string object prototype method on a literal string. Engines will box that value into a String object, Boa doesn’t do this yet, so that will most likely be the reason it doesn’t work.

There is still some work to do here around boxing and unboxing and the best approach.

https://github.com/getify/You-Dont-Know-JS/blob/master/types%20%26%20grammar/ch3.md#boxing-wrappers

[calluw]

Thanks @jasonwilliams, generating the string via new String(..) is working with charAt(), charCodeAt(). Will now test with my new implementations.

[calluw]

On the topic of the boxing of the primitive string, the only current issue with concat(), repeat(), slice() and the other methods that should return strings is that through to_value() they come out as "string literals" again. This means you can't do the following:

let a = new String("Hello world");
a.concat("! ").repeat(2);

Is there an easy way to cause concat, repeat etc to return not just string "values" but string objects? Would this be solved by auto-boxing string literals (#29)?

[jasonwilliams]

Now you're asking the interesting questions!
We can't do this because the return type of those methods are string types not objects.
Those strings would need to be re-boxed again (yep this is why boxing/un-boxing is a big performance hit and engines often try to find ways around it).

It would be against the spec to return a String instance instead of a string literal.
https://tc39.es/ecma262/#sec-string.prototype.concat

Would this be solved by auto-boxing string literals (#29)?

Basically yes. I think some research would need to be done in this area as i believe there are ways to optimise this operation rather than simply boxing and unboxing again.
In theory we can detect a method being called on a string, so at that point we could convert into a String instance call the method then throw away the instance (wasteful).

There is 1 place where i break this rule and just give back the value, that is when calling length on a string literal. https://github.com/jasonwilliams/boa/blob/master/src/lib/js/value.rs#L202-L208

[calluw]

@jasonwilliams Thanks for clarifying the spec's position (I had overlooked the subtlety of the concat/repeat/slice methods returning values only).

As a heads up, next I will probably take a look at the set of these methods which take a "searchString" argument (so not patterns yet): startsWith(), endsWith(), includes(), indexOf() and lastIndexOf().

[brian-gavin]

Hey, the links in this issue are out of date due to some code reorg!

Current location of String is here

[Razican]

Hey, the links in this issue are out of date due to some code reorg!

Current location of String is here

Good catch! I updated the links, but note that some of this is probably blocked until #419 lands :)

[HalidOdat]

@joshwd36 has implemented String.prototype[@@Symbol.iterator] in #704 , keeping track who has implemented what :)

[RageKnify]

Added split(), being worked on in #1026

[RageKnify]

@neeldug there's still a few methods in the check list, should this issue be closed?

[neeldug]

Shouldn’t be closed, only checked off.

[jedel1043]

Closing in favour of #1562.