Chapter 7 comment about gzip could do with clarification about efficient-programming-with-components HOT 1 CLOSED

It wasn't totally clear to me what this was trying to convey.

Agreed, I will rewrite.

At that point the InputIterator analogy gets a bit strained!

It is very common for InputIterator algorithms to store the previous k elements (for a small fixed k)
to do some kind of analysis. See for example: https://en.cppreference.com/w/cpp/algorithm/adjacent_difference
So yes, gzip does build up a window, but it's still a constant size memory usage.

A compression algorithm which wouldn't work this way is using the globally optimal huffman tree.
You need to observe all the data first to construct the optimal tree, then go back through the data and compress with the constructed tree. You can guess,and adjust on the way, but it fundamentally is not an InputIterator algorithm.

I suppose you could then argue that doing that for small chunks (like gzip) is just rearranging that ...

from efficient-programming-with-components.