take! on IOBuffer is not optimized for the common usecase where the buffer is not used again

[KristofferC]

See discussion in #27691 (comment).

[JeffBezanson]

Yes. take! currently tries to reset the IOBuffer to a state where it can still be used, but that requires allocating a new buffer, which sort of defeats the purpose. If you want the data without destroying the buffer, call read!. take! should just resize the vector, mark the iOBuffer as closed and unusable, and return the vector to you.

[StefanKarpinski]

That would make sense to me: I think as long as the IOBuffer object is not reusable that should work well. Reusing closed IOBuffers seems like a questionable and rare thing to do.

[stevengj]

Is there a non-breaking way to improve this in the meantime? There are dozens of usages of take! in Base and stdlib that could be improved if we provided a better function for this.