Make editor remember the latest search register

[xJonathanLEI]

Resolves #5112.

This PR makes changes to the behavior of search commands:

• The editor now remembers the search register used to kick start a search/rsearch/global_search command;
• search_next, search_prev, extend_search_next, and extend_search_prev now uses the search register to continue the search;
• search_selection now uses the selected register. It also updates the search register;
• make_search_word_bounded defaults to using the search register to make the search_selection-make_search_word_bounded flow more ergonomic, while also accepting register selection.

Overall, this PR should make the whole family of search commands play more nicely with register selection.

[xJonathanLEI]

Update:

Recent changes in the editor fixed the part where commands do not respect immediate register selection (i.e. "bn now uses register b as expected). However, the UX is still bad as this requires the user to enter the "b prefix every time to navigate through the search results.

That's why this PR is still relevant IMO, as the 2nd part of the PR is to make the editor remember the latest register used to kick off a search, and uses that by default unless a prefix is explicitly specified. In this case users can drop the prefix and simply press n to navigate.

Also update the PR title to better reflect the only part left for this PR.

[pascalkuthe]

yeah I think this is a nice QOL improvement and makes sense. Implementation is pretty small/unintrusive too 👍

[the-mikedavis]

This seems potentially too sticky to me: / is meant to be the default when you don't provide a register but this change makes any register that you select the default until you explicitly select another one. I would prefer to see this as last_search_register: Option<char> on the editor. The functions that start searches like searcher and search_selection would set/clear the field if a register is/isn't selected while the functions that continue a search like search_next_or_prev_impl and make_search_word_bounded would only read from the field.

[xJonathanLEI]

this change makes any register that you select the default until you explicitly select another one

Not really though? If you start a new search with just / (no explicit prefix), it changes back to the / register. See this line, which is what starts a search. It doesn't even read from search_register.

I too agree that if it were that sticky it's a problem, but that's not the case here. I also just tested my branch and it's indeed working as expected (i.e. reverts back to / when performing a new search without prefix).

[the-mikedavis]

Ah yep I see now. Let's rename it to last_search_register though, the current name is confusing to me.

[xJonathanLEI]

Renamed the field the last_search_register as suggested.

(Oops. Maybe I shouldn't have rebased... Now the compare button is pretty much useless)

[pascalkuthe]

Diff is small enough to review quickly so no worries
:) Maybe one more thing: we could move last_search_register to editir.registers. This avoids adding more field to Editor

[xJonathanLEI]

Sure thing! But do you mean making this a pub field of Registers? Looks like that struct does not have any public field yet. Should I make it the first? Or should I keep it private but expose a method with the same name?

Edit: I pushed the pub version optimistically.

[pascalkuthe]

A public field is probably fine in this case 👍 I don't think it makes sense to encapsulate here. For me it's just about bundling related fields together instead of making Editor the kitchen sink for everything