Support diff highlight with original file format syntax scheme

[asfaltboy]

This might be dupe, if so I appologize; I saw the discussion being had a number of times, but never as a proper feature request, so here we go.

In Github the diff with highlight is displayed very nicely, and I'd like to have it in sublime.

I know this can be done, but not sure how cleanly the code will remain after. The plugin sublime-compare-side-by-side does this by applying "highlight" to files that have been added to the view with the original/default syntax scheme. I understand that with side-by-side diff it is easier to keep syntax, and in "patch view" case we might need to apply the syntax line by line (by checking the original file's syntax for the row) + apply the diff highlight.

Related tickets, discussions:
#666
8c1313b#r121255619

[randy3k]

Actually, I tried the approach of applying the syntax line by line in the blame view, but it was incredibly slow and I had to abandon that.

[stoivo]

Have you tried to do it only for the content vissable? I think sublimes has an API for checking the viewport. Or is it slow on small files as well

[randy3k]

It is ok for small files, but it is an O(n) operation.

[stoivo]

And did you look at the option for only doing it for the current virwport?

[randy3k]

No. I am not sure if sublime has an event listener on change of viewport.

[asfaltboy]

We might be able to get the syntax for the whole file then, by using the "open, get syntax and close quickly" method. To avoid continuous flicker, we can also "cache" the syntax for each opened file, as described here: https://forum.sublimetext.com/t/get-syntax-associated-with-file-extension/26348/11

[asfaltboy]

There's also a discussion on a relevant feature request in sublimehq/sublime_text#2312, and a workaround provided by sublime_lib here.

[kaste]

@randy3k How did you apply the syntax line-by-line? Applying foreground regions token by token? Which tokenizer did you use?

[randy3k]

I used view.scope_name to extract the scopes from an ordinary view of the file and then use view.add_regions to apply the scopes to the blame view. It was quite slow for files > 1000 lines.

[asfaltboy]

Do you still have that branch? Maybe we can hack around it?

[randy3k]

I have discarded that unfortunately.