Mixed types

[bengl]

From #1:

5. The API needs to appropriately consider the differences in interpretation of different TypedArray types. Specifically, it should be possible to concat only like-types (e.g. Uint8Array + Uint8Array, Float64Array + Float64Array, etc). Mixing types should likely be forbidden.

and from the slides:

Typical use case is concat'ing same type (u8 + u8, u32 + u32, etc) so handling mixed unnecessary

A common thing I do is encode data of multiple different numeric types into the same buffer. For example, an ArrayBuffer that's composed of data originating in a sequence looking like [Uint8Array[1], Float64Array[2], Int32Array[1], ...]. This also comes up when attempting to create or manipulate C structs within ArrayBuffers. Today, this usually involves something like a DataView or having multiple TypedArray views on the same ArrayBuffer and indexing appropriately. If I already have the sequence of TypedArrays of various types handy, which is often the case, it would be nice to let the language implementation take care of the concatenation bookkeeping, and even handle it lazily where possible.

In short, yes, mixing types for concatenation is useful. Perhaps, depending on the conclusion of #2, an ArrayBuffer.concat could be provided, taking in an iterator of arbitrarily-typed views.

[jasnell]

Hmm... yeah, I think the only path to being able to mix types would be having ArrayBuffer.concat as well. Otherwise we end up with too much risk of element length and interpretation issues, e.g. Uint32Array.concat([new Uint32Array(1), new Uint8Array(1)])