visible whitespace

[the-mikedavis]

Example with this rebased on top of #1796:

Adds some configuration options to config.toml:

# render whitespace characters (\t, \n, ' ')
[editor.whitespace]
render = "all"
# render = "none"    to disable visible whitespace

# or control whether each character is shown:
[editor.whitespace.render]
space = "all"
tab = "all"
newline = "none"

# control the grapheme used to represent the whitespace
[editor.whitespace.characters]
space = "·"
tab = "→"
newline = "⏎"

And a new scope for themes ui.virtual.whitespace, which should be set to something slightly off-background.

This makes way for a future "editor.whitespace.render" = "selection" as well, which could enable showing whitespace only in a selection rather than always.

TODOs:

• non breaking space character? (U+00A0F, default: ⍽)
• update existing themes to include ui.virtual
• add new config documentation to the book
• decide whether this should be off by default
• squash down commits 🔨

connects #1068
closes #1218
closes #688

[EpocSquadron]

I vote for on by default, this makes it fast easier to see when you are on the ending character, which will resolve a lot of confusion on how certain selection operations work. It suits the batteries included approach of helix better, IMO.

[Omnikar]

Thanks for continuing #1218. There's a bug I'd encountered when I first made that PR where when the cursor is at the very end of the file (after all characters in the file), the spot the cursor is on is rendered as a space; the bug seems to still be present here.

[the-mikedavis]

oh interesting, I see it too:

[the-mikedavis]

Looks like the space comes from here:

helix/helix-term/src/ui/editor.rs

Lines 343 to 346 in 05161aa

	// `unwrap_or_else` part is for off-the-end indices of
	// the rope, to allow cursor highlighting at the end
	// of the rope.
	let text = text.get_slice(start..end).unwrap_or_else(|| " ".into());

So I just pushed a commit that lets you configure which character that is and defaulted it to the braille blank (U+2800). It's a little brighter than a space, but I don't think it's very noticeable.

This approach seems kinda hacky though, what do you think?

[the-mikedavis]

Also eof seems like the wrong word for it, I suppose it's more like a trailing_cursor

[the-mikedavis]

actually it ended up being cleaner to detect when the highlight affected the trailing cursor and prevent substitution in that case: 45fd800

[archseer]

I wouldn't want this on by default but it should be easy to toggle at runtime. I had space+s set up to toggle character highlighting on/off in vim so I could check indentation if I needed to.

[archseer]

(So two config options: one for character map, one for on/off)

[sudormrfbin]

I was also thinking about a generic toggle mode, so for example <space>ts for toggling whitespace, <space>td for toggling diagnostics, etc.

[the-mikedavis]

I added this to the setting command so we can say :set whitespace.render all with the default being off. We could simplify the whitespace.render config struct to a boolean but OTOH I'd like to support the whitespace.render = "selection" and other config in #1068 eventually

[pickfire]

Should we just store this within editor?

[the-mikedavis]

Yeah I think there's a method or two in Editor (other than this one) that end up using the config in other ways as well. Seems wise to make that available so we don't need to pass it in all the time.

edit: actually I don't think we can have the editor config or whitespace config be a field on the EditorView struct, this call to render_text_highlights looks static

[pickfire]

I think we might need to check the tab substitution width, if it's substituted with two-width characters then it will look twice as large. But I doubt anyone would want to do that.

[the-mikedavis]

Is it possible to have a multi-column char? If so, we might be able to check for this when deserializing from config with helix_core::graphemes::grapheme_width

[pickfire]

What is this Specific for? Why do we need to differentiate with basic?

[the-mikedavis]

this comes from the original PR (#1218) - it's for this more advanced configuration: #1068 (comment)

I'm not too sure how valuable it is to have it be this configurable though. Maybe it'd be good enough to have a simple boolean that enables visible whitespace? You can also "turn off" visible whitespace for each kind of character by configuring substitutions to use spaces

[editor.whitespace.substitutions]
newline = " " # don't make newline visible

[pickfire]

I was also thinking about a generic toggle mode, so for example ts for toggling whitespace, td for toggling diagnostics, etc.

@sudormrfbin IIRC doom emacs have some triggering mechanism as well similar to what you said, maybe can steal from them.