Some small optimizations (helping type inference, avoid eager int to string conversion)

[KristofferC]

Benchmark used:

using Random
using StyledStrings

function generate_annot_string(n)
    colors = [:yellow, :green, :blue, :red]
    offset = 1
    annotations = Tuple{UnitRange{Int64}, Pair{Symbol, Any}}[]
    while true
        col = rand(1:length(colors))
        l = rand(1:10)
        offset + l > n && break
        push!(annotations, (offset:(offset+l), :face => colors[col]))
        offset += l
    end
    s = randstring(n)
    return Base.AnnotatedString(s, annotations)
end

str = generate_annot_string(100000)

using BenchmarkTools

@btime show(IOContext(devnull, :color => true), MIME("text/plain"), str)

Baseline:
27.456 ms (554582 allocations: 26.87 MiB)

First commit
13.629 ms (412776 allocations: 20.65 MiB)

Second commit:
12.911 ms (408186 allocations: 20.46 MiB)

Third commit:
9.508 ms (290571 allocations: 18.78 MiB)

[KristofferC]

This pattern avoids doing the dict lookup twice.

[KristofferC]

I don't actually know why this happens, it's just that doing this change fixes the issue reported by @code_warntype... I'm kind of scared of closures for reasons like this :(

[tecosaur]

This all looks good to go from my end. If there's nothing else that you're thinking of adding to this, I'm tempted to just rebase + tweak a few of the commit messages then merge 🙂

[KristofferC]

Sounds good to me!