Design Questions

[tzachar]

Hi there.

Im trying to understand some of the design choices, as Im having trouble with some of the current behaviour. To be more precise, why are you calling MagmaBuffer.clear_interface on every CursorMoved? I would expect this to only be called when the cursor exited a cell / entered a different cell. This causes an excess of ui redraws, and also make it difficult to properly place images using Kitty.

[dccsillag]

That's because on CursorMoved there is a chance that the output window should change its position (e.g., the cell has a new line). It seemed quite tricky to do some sort of output display update without clearing everything and starting over from scratch.

Doing this for the floating window seemed fairly cheap; as for images, not so much; so what I did (in the Ueberzug canvas) was to have add_image and clear only queue image add/remove operations, which are only done on present, where we do not do anything when the image is to be both hidden and shown.

[tzachar]

Ok, but I still have an issue. The kitty graphics protocol only support image placement where the cursor is. So, I have to move the cursor b4 positioning the image. But this results in a CursoreMoved event, which gets me into an infinite loop or jumped out of the cell and the image is removed. I need some way to tell magma to disable catching the CursorMoved events so I can update the display properly.

Any idea?

[dccsillag]

Well, Vim has :noautocmd. Try using that? That would ensure that no autocmds are triggered at all, not only Magma ones.

[dccsillag]

Actually, it looks like all you'd need to do is change the eventignore option, which is much easier to do from a remote plugin (via self.nvim.options["eventignore"]).

eventignore_save = self.nvim.options["eventignore"]
self.nvim.options["eventignore"] = "all"
# ... do stuff
self.nvim.options["eventignore"] = eventignore_save

[tzachar]

10x, it seems to work. Updated #16

Would appreciate you looking at it.

[dccsillag]

Great! Will review in a bit.