feat: add number styling

[ttytm]

Thanks for the maintenance. It's nice to see the repositories active development.

This PR extends the options for buffer indices and numbers with the style options 'superscript' and 'subscript'. Current boolean values will continue to work as they are.
I did my best to follow the conventions I've seen in the repo.

[romgrk]

You know what, pass those digits as an argument to style_number and we can keep the string constants. E.g. style_number(num, style, digits). The digits would be either a table or nil. The string comparison happens once per render, I'm cool with that.

[Iron-E]

Good call, that can be resolved outside the loop.

[ttytm]

Wouldn't this give us two of those?

buffer_index.text = style_number(
  i,
  icons_option.buffer_index,
  icons_option.buffer_index == 'superscript' and SUPERSCRIPT_DIGITS
    or icons_option.buffer_index == 'subscript' and SUBSCRIPT_DIGITS
    or nil) .. ' '

Partial reverting the full encapsulation I did in the second commit might be more elegant. Only calling style_number() when superscript or subscript is specified but not passing digits as arg.

buffer_number.text = ((icons_option.buffer_number == "superscript" or icons_option.buffer_number == "subscript")
        and style_number(bufnr, icons_option.buffer_number)
        or tostring(bufnr)) .. ' '

What you think?

edit: Ah I see, it would still take two comparisons then!
If you like a helper function for the one comparison let me know.

[romgrk]

You'd assign digits outside of the loop.

local digits = nil
if opt == 'super' then digits = SUPER end
if opt == 'sub' then digits = SUB end

for ...
  text = style_number(..., digits)
end

[ttytm]

I see, didn't realized the outer loop. Thought the loop that needed to be solved was solved with iron-e suggestions. Thanks!

[Iron-E]

No matter what, certain options have to be resolved in the loop. That commit just increased the perf of doing that.

If @romgrk is okay with it, you can implement the with style_numbers resolving digits internally and do a benchmark to see what the impact is.

You can use this code to do that

If the impact is significant, we can just have a number_style option that can be resolved externally. Less configurability, but makes things faster.

[ttytm]

Thanks for the snip! On my box it just had little impact during normal usage while running style_number in a loop of 1000 where we currently set buffer_index.text. Highest 0.0036389999999997s. Lowest 0.00029000000000146s.

Metrics comparing function calls:
10.000 loops to just get a number from index.
0.0017560000000003s
10.000 loops of style_number, digits passed as arg
0.0039189999999996s
10.000 loops of style_number, digits resolved in a condition inside the function (like the current code does)
0.0038720000000017
10.000 loops of style_number(internal condition), returning a "non-styled" number.
0.0017619999999953s

So there doesn't seem to be any significance resolving digits in style_number or passing them as arg.

[Iron-E]

Thanks for doing that. We can let romgrk decide from here 🙂

[romgrk]

Let's pick whatever option is more simple, doesn't seem to have a significant performance impact.

[ttytm]

I think the state now that includes the suggestion after the initial PR is fairly concise and readable 🙂:+1:

[Iron-E]

I can solve the merge conflicts, don't worry