ignore concealed characters

[justinmk]

A search may fail even though the visible text appears to be a match, because the actual text contains a concealed character that prevents the match. Maybe Sneak can provide an option to match only visible/non-concealed text.

[justinmk]

after some thought, this doesn't have many use cases and will be more confusing that helpful

[chiphogg]

I think there's actually a major use case: vim-dirvish.

The directory buffer in dirvish uses concealed text: each line contains the full path, but only the filename is visible. This can cause sneak to fail hard. In many cases, every line becomes a match. In fact, depending on the way folders are laid out, each line can effectively be multiple matches!

I think the confusion consideration cuts both ways on this one. When I use sneak, I'm usually jumping to something I can see. It was confusing at first to be jumping to lots of intermediate results, none of which were even visually highlighted. Now, of course, I understand how it works, and have for many years, so I'm used to it. But at some point, I realized that the thing I was initially trying to do (i.e., jump among visible results) might actually be the more desirable behaviour intrinsically. 🙂

@justinmk: WDYT about reopening and reconsidering this?

[ggandor]

I think it would be best to set conceallevel to 0 on SneakEnter, and restore it on SneakLeave. It's not ideal, the sudden shift of characters is annoying, and might be disorienting, but still better than the confusing behavior of getting unexpected matches or worse, not jumping to an expected match.

Edit: nevermind, I forgot that Sneak actually uses conceal to display the labels (on Vim)

[justinmk]

I forgot that Sneak actually uses conceal to display the labels (on Vim)

Yeah, though not in Nvim 0.5+: 3ff1033

It probably makes sense to special-case dirvish buffers.

In any case, first need to figure out how to only search "visible" text.

[chiphogg]

I had a thought this morning: can we simply discard any matches after the fact, if it turns out that their text is completely concealed? (I have no idea how these things work, but if it's possible to tell whether text is or is not concealed, then we should be able to use this as a post-processing filter without needing to mess with the "core" search algorithm.)

[ggandor]

It probably makes sense to special-case dirvish buffers.

It would be great for help files too. In fact, cannot think of a single situation where it would make sense to match hidden characters with such a jumping plugin (well, for label mode at least).