Why on earth are we hashing window names?

The "name" is only used as an id for the window. So we hash it on creation of the window. Then, nk_find_window() (which is called all the time), looks for a matching window hash (recreating the hash): and then checks the name (running strlen again!).

Would it not make a whole lot more sense to say, "use a unique integer identifier" for each window (or other item that needs an id), and just store that id ... thereby not needed to do a hash nor compare strings nor do strlen just to find a window?

You could even have a "nk_unique_id()" to return, per context, a new int that is unique (just increment a counter). No need to be fancy.

What am I missing?

Start by tracing window creation and nk_find_window(), focusing on where names are hashed, compared, and measured. Determine whether a context-scoped unique integer can preserve the existing lookup behavior; done requires a maintainer-approved design, since the issue currently poses a question rather than specifying an implementation.

Written by the indexing model from the issue text.