Add new region flag for dimmed foreground

[jwortmann]

Problem description

This is a feature request that was already briefly discussed in #817 (comment), but I think it deserves its own issue so that it doesn't get buried in that thread.

It would be great to add a new region flag for the View.add_regions API to enable syntax highlighting with a dimmed foreground color.

The use case I have in mind for this are certain diagnostics in LSP with a special DiagnosticTag.Unnecessary property, which is utilized for unused parameters in function definitions and unused variables. Many language servers support this kind of diagnostics. These diagnostics might also be used for unreachable code blocks spanning multiple lines, but I haven't yet seen that in practice.

Preferred solution

class RegionFlags(enum.IntFlag):
    # [...]

    DRAW_STIPPLED_UNDERLINE = 1024
    """ Draw a stippled underline below the regions. """
    DRAW_SQUIGGLY_UNDERLINE = 2048
    """ Draw a squiggly underline below the regions. """
+   DRAW_DIMMED = 4096
+   """ Draw the regions with a dimmed foreground color. """

This should look exactly like the way control characters are already drawn, for example the <NBSP> in the following screenshot:

(I used a custom syntax definition for this example, which applies the listed scopes to the entire line.)

Notably, if a given region spans multiple scopes with different colors, the text in the region should keep its respective colors, just with a dimmed opacity. Text which is bold or italic should keep its font style.

Alternatives

none

Additional Information

One aspect to think about is what should happen if the given region starts or ends in the middle of a token used for syntax highlighting. This particular case should never happen for the LSP diagnostics mentioned above, so I imagine that it wouldn't be any problem to extend the dimmed highlighting to the start and end of syntax tokens, if that makes the implementation easier or more efficient.

[rchl]

These diagnostics might also be used for unreachable code blocks spanning multiple lines, but I haven't yet seen that in practice.

pyright does it (it's basedpyright on the screenshot but I think both work the same):

[rwols]

Another datapoint for unreachable code blocks: LSP-clangd uses it for code wrapped in preprocessor macro blocks that aren’t active.