feat: translate on hover about linguist HOT 4 OPEN

Sounds good. I would implement it as another popup, not as a "selected text translation" feature.

How this feature can be implemented technically?

If you need to translate one word, we have to detect what word user hover on. We can track hovered element with pointeover event, but how to detect exact word? Have you some ideas?

from linguist.

How this feature can be implemented technically?

The extension that I mentioned appears to be doing some funky markup manipulation: https://github.com/artemave/translate_onhover/blob/12aab5217677b0ba3579106c1200443454c87a16/contentscript.js#L166, I'm not sure what's exactly going on there, maybe it transforms a text node into a bunch of text nodes each containing one word (which, I assume, may be the reason for some bugs like this one).

Another possible implementation would be to determine the word being hovered over by looking at the position of the cursor relative to the text node's rectangle. E.g. if a text node has two equally long words, we check whether the cursor is on the right or on the left side of the text element. But yeah, it's gonna be pretty inaccurate, at least for a very basic implementation. I'm sure this algorithm is implemented in some very low-level things like OS or some rendering engines (LibreOffice, game engines, text editors).

It makes sense to search for another solution.

from linguist.

@WofWca nothing to add, all your points are right. 1st solution unacceptable, because Linguist must never change original DOM structure, 2nd solution may be inaccurate by many reasons and we cannot to ensure nothing with this way.

But i like your idea, so i will keep it in mind, if you will find a possible solution, please return back here and add a new information

from linguist.

We could potentially clone hovered element to shadowRoot and split nodes there to calculate position, but this approach are not trivial, hard to maintain and may generate a lot of bugs like "texts styles are different in virtual element".

I imagine an ideal solution as some intersection API for chars - we detect hovered char and get chars around. Maybe we could hack text selection API.

But if such API does not exist, we could think how to simplify solution above

from linguist.