Ignore accents

[dbmrq]

I don't know how hard it would be to do this, but it would be really really great if Sneak could consider characters with or without accents as the same.
Just now I wanted to go to the word "ética", for instance, and I instinctively tried Sneak with et, but it doesn't work. This happens to me all the time.

PS: I love Sneak, thanks for making it!

[justinmk]

I think this is what :lnoremap is for. Let's merge this with #138

If you really like this kind of thing, easymotion probably has a feature for it already, I think. The goal of sneak is less "do what I mean" (DWIM) and more narrow.

Thanks too :)

[dbmrq]

I've tried easymotion, but I don't like it as much. :/

lmap is just like a regular mapping but for more than one mode, isn't it? Maybe I'm missing something, but I don't see how that's relevant.
What I want in my example is that Sneak et would match not only "ét", but both "et" and "ét".

Here I mapped f to use one-character Sneak and did fo:

Here I'd hope "ó", "õ" and "ô" to be highlighted too.

I think what we would need is something like this: http://vi.stackexchange.com/a/7394

Searching for [[=e=]] matches "e", "é", "ê" and so on. So we'd need an option that would cause Sneak to search for [[=e=]] when you type e, [[=a=]] when you type a etc. I don't see any drawbacks in doing this (except perhaps for the performance of the plugin), but it could be changed with an option anyway.

I don't use lmap or keymap, I use combining characters for accents, so to type "ética" I actually type 'etica. Sneak doesn't count 'e as "é", though, so I can't search for that character at all, and the same goes for all accented characters. And even if that did work it would require an extra key stroke. Using collection classes would fix that and many other possible problems IMHO, making Sneak a lot more flexible.

This would even be useful for english speakers I think… there are many loan words from french, for instance, and what would you do if you wanted to go to the word résumé? It would be a lot easier to just do Sneak re.

[justinmk]

Hmm, equivalence class looks useful. It shouldn't be much code to support this so I'll keep it open.

[dbmrq]

Great, thanks! 🎉

[dbmrq]

I can try to implement this if you want, but I don't know much about Vimscript, so I'd need some help.

Now I just looked into your search.vim and tried replacing the line that says

    let self.search = escape(a:input, '"\')

with

    let self.search = substitute(a:input, '\a', '[[=\0=]]', 'g')

But it doesn't work, I don't know why. Doing that never get any matches anymore.

[justinmk]

Sorry, it might be awhile before I can help with this.

[dbmrq]

No problem, I'll keep trying and I'll post here if I get anywhere, and you can look into it whenever you have the time. Thanks again. :)

[dbmrq]

Aha! I think I got it.

The problem was the escaping, and then I tried to escape just the characters in the original code, "\, but it wasn't working.

Besides escaping those characters, now I tried adding a backslash before the whole equivalence class too:

    let escapedstring = escape(a:input, '"\')
    let self.search = substitute(escapedstring, '\a', '\\[[=\0=]]', 'g')

And it works! Here's sae:

I also tried some other random characters, slashes, quotes, braces, it seems like it's all working as it should.

I didn't make a PR already because I don't know if this has any unintended consequences or if there's something else we'd have to change too, I got to this via wild guesses and I'm not familiar with the rest of the code at all. But if you think it's alright I can make a PR too.

For now I'll keep using it like this and if anything weird shows up I'll let you know.

[justinmk]

Nice! Try it for a few days and let me know. This is simple enough that it's worth adding.

[dbmrq]

Ok, I used this a lot over the last month and it seems to be working great. :)
I still don't know why it needed that extra back slash, but it works perfectly and it made sneak a hundred times better for me.
Should I make a PR or something?

[justinmk]

@danielbmarques PR would be welcome.

[ggandor]

In case someone is reading this now and interested: the extra backslash at the beginning of '\\[[=\0=]]' is necessary because the final search pattern is very nomagic (self.prefix sets \V), but we need a collection-delimiting pair of brackets (\[]) around the equivalence class, not literal ones.