Support for font ligatures

[jasonlong]

This might very well be more trouble than it's worth to implement, but I was wondering if it would be possible to (re)implement font ligatures so that fonts like Fira Code would work.

According to the Hasklig project that inspired Fira Code, MacVim can't render ligatures because of this old commit.

It sounds like there are side effects to simply reverting this commit, but if they aren't too nasty, it would be awesome to have ligature support.

[HealsCodes]

@jasonlong - the commit you're referencing disables ligatures in the old ATSU renderer.
The current CoreText renderer never hat ligatures in the first place..

But since I'm in the same boat and I'd love to use Fira Code I wrote a patch to make ligatures available
in the CoreText renderer and also provide an option to turn them of and on:

I've opened PR #56 witch includes my changes.

[jpetrie]

@douglasdrumond This can probably get closed since #56 was merged.

[HealsCodes]

👍 however I still have a little improvement (rounding of character positions) that didn't make it into the PR. The only difference being that the current version can have a slight "shimmer" when moving between lines which results from minor differences in the character positions if not rounded.

[jpetrie]

Closed in #56.

[jordwalke]

Sometimes when I move my cursor over the rendering, it reverts back to the original characters? Should this issue be reopened until the issues are resolved?

(Great work implementing it though!)

[HealsCodes]

That's a known limitation of the current implementation (also visible in the demo above).
If you like you can take a look at PR #56 for more details.

[jordwalke]

Is there another open github issue tracking it?
I experimented with adding text editor ligatures, and I found that drawing all content in the editor one line at a time (and only ever one line at a time) was the easiest way to get everything working consistently - apologies if that suggestion is incredibly naive.

[HealsCodes]

As it would imply some deep changes to the current renderer there is not other issue tracking it.
It was agreed so on so far that the splitting on moving the cursor over a ligature is an acceptable behavior for now.

MacVims CoreText rendering is using a lot of partial redraws to keep an acceptable speed, which in turn can result in liguatures splitting back up if the cursor moves over them (the ligature detection only receives the text to be drawn not the whole line). I'm afraid there is no simple way to fix that.
But feel free to give it a try 👍

[jordwalke]

I think your gif demo shows something different than what I'm describing. In your example, I think it doesn't seem as bad because the moment you leave the line, it renders correctly. What I'm seeing is that even if you leave the line - it still is not repaired. I believe the reason is that you have line highlighting on (highlighting the current line) - whereas I do not. The behavior I'm seeing would warrant leaving this issue open - whereas if I had the experience you demonstrated, I would say that it is somewhat usable. Currently, with what I'm seeing it's not. Would you try disabling your current line highlighting and seeing if you agree?

[jordwalke]

Also, as you can see here, there seems to be some other serious issue. That line only has one semicolon at the end, yet two are rendered because - for some reason - when the == glyph is broken back into the two separate characters, it makes the line shorter and draws the new semicolon, while leaving the old one.

[HealsCodes]

Would you be willing to compile MacVim from source?
The issue you describe looks like part of what was discussed in the original PR after it was merged :(
I have a gist of MMCoreTextRenderer.mm that should at least get rid of the spacing issues (I've seen them before).

[jordwalke]

Yes I would - do you have a branch that I should try?

[HealsCodes]

You can use the current MacVim master or my branch for the original PR. But you need to replace MMCoreTextRenderer.mm with the file from my gist.