stirng.encode_wtf16_array with start/end

[gkdn]

I noticed that string.encode_wtf16_array is not accepting the string range to encode (like sourceStartIndex and sourceEndIndex).
Is it efficient enough to do a slice first and then call encode_wtf16_array or would it be worthwhile to take additional parameters to encode_wtf16_array?

[gkdn]

@jakobkummerow

[jakobkummerow]

Creating a slice should be fairly cheap (allocation of a small object, could possibly be optimized out by engines in the long run). I'm also fine with adding parameters to encode_wtf16_array, or having a second version of that instruction that takes such parameters. Either way that instruction would probably belong in the world of stringview_wtf16 then.

I think it mostly boils down to: which scenario is more common in practice?

• wanting to encode an entire string, without having to spend binary size on providing (0, length) as range?
• wanting to encode part of a string, without having to explicitly create a slice first?

If it helps, we could also implement the second version as an experiment, then you could test with microbenchmarks to see if it makes a measurable difference.

[gkdn]

I don't think there will be any concerns around binary size here so it is ok to if there is only one instruction.

If the slice is effectively a view, I'm also not much concerned around the performance but if you would like to experiment in any case I can quickly adapt and provide you some numbers.