Join elements into a single string without `.toArray()`

[tats-u]

Some other languages provide a way to join elements into a single string without having to convert the iterator to an array.
It is disappointing that this proposal does not provide a similar API.

• Java Stream API: iterator.collect(Collectors.joining(delimiter))
• Kotlin: iterator.joinToString(delimiter)
• C#/VB LINQ: String.Join("delimiter", iterator)

With the current specification, we have to call .toArray().join(). If JS engines do not estimate the length of the result string in advance, this array allocation is extra and unnecessary. They should.

---

P.S.

• Users may expect that the iterator has .join method like arrays or the above languages
• .toArray() in .toArray().join() is a little bit verbose and looks like a little boilerplate

[Josh-Cena]

You can use iterator.reduce((acc, cur) => `${acc}${cur}`, "") FWIW

[tats-u]

I wonder if It allocates a new intermediate string every call of the lambda, or JS engine optimize it without allocation of them.

[ljharb]

Do you have concrete benchmarks that this approach actually has memory problems? because you can’t assume that an implementation necessarily does or does not do that allocation.

[tats-u]

No, I am just concerned. I wonder if we have no choice other than trusting optimization of .toArray().join(delim) (or .reduce((acc, cur) => `${acc}${cur}`)) by JS engines (in the future).
I forgot to add "users may expect that the iterator has .join method like arrays or Java/C#."
This suggestion (supplementation) can be another proposal like https://github.com/tc39/proposal-async-iterator-helpers.

[ljharb]

I expect the optimization of the former, at least, to be identical to that of a .join method on iterators.

[tats-u]

I wish someone would perform a bench with an iterator with so many elements using the latest Node 21 or Chrome, where this API is available.
"It should be done by you!"? It makes sense...

[Josh-Cena]

How can a native join be any more efficient than a series of string concatenation when the total length is unknown AOT?

[tats-u]

Right, it does not need reallocation.
If the result string size is strictly calculated, the algorithm equivalent to .toArray().join() is required.

[michaelficarra]

This proposal is stage 3 and will not be taking on any new additions. If you'd like to discuss this idea further, I'd suggest using the Discourse. Closing.