refactor: follow-up to leaf updater uses binary search

[gabriele-0201]

The following pr contains all the follow-ups to #529.

One specific thing I would like to ask is an opinion to the introduction of LeafNode::extract_key. It is a small abstraction that still requires some conceptual thoughts. The point is: the layout of the leaf node is mostly encapsulated within its module, and inside keep_up_to, I was extracting the key from the cell_pointers. However, I thought that a modification in the layout of the leaf node could silently break the leaf updater. That's why I added this helper function. In case a layout change happens, the entire leaf/node.rs code would be updated, including the way a key is extracted from the cell pointer.

[gabriele-0201]

• refactor: follow-up to leaf updater uses binary search #565 : 2 dependent PRs (#566 , #570 ) 👈 (View in Graphite)
• fix: update BaseLeaf when keep_up_to jumps to end #564
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.

[rphmeier]

One specific thing I would like to ask is an opinion to the introduction of LeafNode::extract_key. It is a small abstraction that still requires some conceptual thoughts.

This seems reasonable. We should avoid silent breakage (and ideally have a test which prevents those kinds of regressions). Helper functions are usually good, just not worth hiding the format of the leaf from the updater. as we will eventually have greater optimizations by making use of it.