JuliaMono - a monospaced font for scientific and technical computing

JuliaMono is a monospaced typeface designed for programming in the Julia Programming Language and in other text editing environments that require a wide range of specialist and technical Unicode characters. It was intended as an experiment to be presented at the 2020 JuliaCon conference in Lisbon, Portugal (which of course didn’t happen).

JuliaMono is:

• free

• distributed with a liberal licence ^[1]

• suitable for scientific and technical programming as well as for general purpose hacking

• full of Unicode goodness

• easy to use, simple, unquirky, friendly, and approachable

• available for MacOS, Unix, and Windows ^[2]

This site uses JuliaMono for all text; if your browser can’t^[3] (or you didn’t allow it to) download and display web fonts, you’ll only see the font in action in the images. You’ll see a familiar formation of three dots here when/if the font has been downloaded:

To download and install JuliaMono, see the instructions here.

Type anything here. Nobody’s looking.

Screenshots

Editing code in Juno.

And in VS Code.

And in Vim:

And in Emacs:

Examples

The following examples will be rendered in JuliaMono by your browser (if it’s successfully downloaded the web font versions), so I hope what you see here is close to what I made.

The CSS markup applied to the following code uses two weights of the typeface, JuliaMono-Regular and JuliaMono-Medium, which is a smidgeon bolder:

using Zygote: @adjoint
function ignore(f)
  try return f()
        catch e; return 0; end
end
@adjoint function ignore(f)
  try Zygote._pullback(__context__, f)
  catch e
    0, ȳ -> nothing
  end
end

There are different weights of JuliaMono, so you can control the amount of contrast you have in your highlighted code: JuliaMono-Regular, JuliaMono-Medium, JuliaMono-Bold, JuliaMono-ExtraBold, and JuliaMono-Black. ^[4]

(There are also versions of two of the fonts with “Latin” in the name: these are stripped down versions supporting just the basic MacRoman/Windows1252 “Latin” character sets, intended for use as place-holders, which are of interest only if you want to have more control over font loading times in web browser-based applications.)

In the hands of a virtuoso (such as Dr Zygmunt Szpak, the author of the following Julia code fragment^[5]), the range of available Unicode characters can be quite expressive:

function T(𝛉::AbstractArray,
           𝒞::Tuple{AbstractArray,
           Vararg{AbstractArray}},
           𝒟::Tuple{AbstractArray, Vararg{AbstractArray}})
    ⊗ = kron
    l = length(𝛉)
    𝐈ₗ = SMatrix{l,l}(1.0I)
    𝐈ₘ = SMatrix{1,1}(1.0I)
    𝐓 = @SMatrix zeros(l,l)
    N = length(𝒟[1])
    ℳ, ℳʹ = 𝒟
    Λ₁, Λ₂ = 𝒞
    𝚲ₙ = @MMatrix zeros(4,4)
    𝐞₁ = @SMatrix [1.0; 0.0; 0.0]
    𝐞₂ = @SMatrix [0.0; 1.0; 0.0]
    for n = 1:N
        index = SVector(1,2)
        𝚲ₙ[1:2,1:2] .=  Λ₁[n][index,index]
        𝚲ₙ[3:4,3:4] .=  Λ₂[n][index,index]
        𝐦    = hom(ℳ[n])
        𝐦ʹ   = hom(ℳʹ[n])
        𝐔ₙ   = (𝐦 ⊗ 𝐦ʹ)
        ∂ₓ𝐮ₙ = [(𝐞₁ ⊗ 𝐦ʹ) (𝐞₂ ⊗ 𝐦ʹ) (𝐦 ⊗ 𝐞₁) (𝐦 ⊗ 𝐞₂)]
        𝐁ₙ   = ∂ₓ𝐮ₙ * 𝚲ₙ * ∂ₓ𝐮ₙ'
        𝚺ₙ   = 𝛉' * 𝐁ₙ * 𝛉
        𝚺ₙ⁻¹ = inv(𝚺ₙ)
        𝐓₁   = @SMatrix zeros(Float64,l,l)
        for k = 1:l
            𝐞ₖ = 𝐈ₗ[:,k]
            ∂𝐞ₖ𝚺ₙ = (𝐈ₘ ⊗ 𝐞ₖ') * 𝐁ₙ * (𝐈ₘ ⊗ 𝛉) + (𝐈ₘ ⊗ 𝛉') * 𝐁ₙ * (𝐈ₘ ⊗ 𝐞ₖ)
            # Accumulating the result in 𝐓₁ allocates memory,
            # even though the two terms in the
            # summation are both SArrays.
            𝐓₁ = 𝐓₁ + 𝐔ₙ * 𝚺ₙ⁻¹ * (∂𝐞ₖ𝚺ₙ) * 𝚺ₙ⁻¹ * 𝐔ₙ' * 𝛉 * 𝐞ₖ'
        end
        𝐓 = 𝐓 + 𝐓₁
    end
    𝐓
end

Languages

Here are some samples of various languages^[6] :

Unicode coverage

One of the goals of JuliaMono is to include most of the characters that a typical programmer or monospaced-font user would reasonably expect to find. (Except for all those emojis - they are best handled by the operating system.)

In JuliaMono, every character is the same width, because this is a monospaced typeface. Usually, typefaces with a lot of Unicode mathematical symbols are not monospaced, because they’re intended for use in prose and \( \LaTeX \) applications, rather than in programming code.

From a design perspective, forcing every character into the same size box is a problem. It’s like fitting every human being of whatever shape or size into identical airplane seats - some characters are bound to look uncomfortable. There’s never quite enough room for a nice-looking “m” or “w”.

UnicodePlots.jl uses various Unicode characters to plot figures directly in a terminal window. ^[7]

ImageInTerminal.jl is similarly awesome, conjuring images from Unicode characters:

JuliaMono is quite greedy^[8], and contains a lot of Unicode glyphs.

(Of course, size isn’t everything - quality can beat quantity, and other fonts will offer different experiences^[9]).

It’s also a good idea to support box-drawing characters and DataFrames.jl output (terminal permitting):

julia> df = DataFrame(A=samples, B=glyphs)
df = 10×2 DataFrame
│ Row │ A              │ B                   │
│     │ String         │ String              │
├─────┼────────────────┼─────────────────────┤
│ 1   │ sample 1       │ ▁▂▁▁▂▄▅▁▄▁▁▅▆▂▇▅▂▇  │
│ 2   │ sample 2       │ ▁▂▄▁▁▃▁▆▂▆▃▁▂▃▂▇▄   │
│ 3   │ sample 3       │ ▁▆▇▁▃▇▇▆▅▅▄▇▇▅▅▇▄▂  │
│ 4   │ sample 4       │ ▅▁▄▁▆▃▁▃▇▂▂▇▅▇▃▆▃▁  │
│ 5   │ sample 5       │ ▆▂▁▂▇▆▃▅▅▄▆▇▄▇▆▁▇   │
│ 6   │ sample 6       │ ▁▁▇▂▂▇▃▅▂▂▆▂▄▄▁▄▂▇▆ │
│ 7   │ sample 7       │ ▂▃▂▁▁▇▁▂▆▂▁▇▁▄▃▂▁▄  │
│ 8   │ sample 8       │ ▄▄▁▂▄▁▅▁▅▁▂▂▇▂▁▃▄▄  │
│ 9   │ sample 9       │ ▁▁▁▂▁▆▃▄▄▁▂▂▃▂▁▅▁▆▃ │
│ 10  │ sample 10      │ ▁▇▄▂▅▃▇▁▇▇▆▄▇▅▄▂▄▅▄ │

(Can you spot the little used and sadly mathematically-unsupported “times” (×) character?)

For a comparison of JuliaMono with other math-capable monospaced fonts, visit mono-math.netlify.app, which shows how Unicode math symbols look.

JuliaMono is quite greedy^[8], and contains quite a few Unicode glyphs.

(Of course, size isn’t everything - quality can beat quantity, and other fonts will offer different experiences^[9]).

If you want to know whether you can use a Unicode character as an identifier in your Julia code, use the undocumented function Base.isidentifier(). So, for example, if you have the urge to use a dingbat (one of the classic Herman Zapf dingbat designs) as a variable name, you could look for something suitable in the output of this:

julia> for n in 0x2700:0x27bf
			Base.isidentifier(string(Char(n))) && print(Char(n))
	   end
✀✁✂✃✄✅✆✇✈✉✊✋✌✍✎✏✐✑✒✓✔✕✖✗✘✙✚✛✜✝✞✟✠✡✢✣✤✥✦✧✨✩✪✫✬✭✮✯✰✱✲✳✴✵✶✷✸✹✺
✻✼✽✾✿❀❁❂❃❄❅❆❇❈❉❊❋❌❍❎❏❐❑❒❓❔❕❖❗❘❙❚❛❜❝❞❟❠❡❢❣❤❥❦❧➔➕➖➗➘➙➚➛➜➝➞➟➠➡
➢➣➤➥➦➧➨➩➪➫➬➭➮➯➰➱➲➳➴➵➶➷➸➹➺➻➼➽➾➿

julia> ❤(s) = println("I ❤ $(s)")
❤ (generic function with 1 method)

julia> ❤("Julia")
I ❤ Julia

Contextual and stylistic alternates

JuliaMono is an OpenType typeface. OpenType technology provides powerful text positioning, pattern matching, and glyph substitution features, which are essential for languages such as Arabic and Urdu. In English, OpenType features are often seen when letter pairs such as fi in certain fonts are replaced by a single glyph such as fi. These ligatures have been used ever since printing with moveable type was invented, replacing the occasional awkward character combination with a better-looking alternative.

To be honest, I’m not a big fan of their use in coding fonts (and I’m not the only one^[10]). I like to see exactly what I’ve typed, rather than what the font has decided to replace it with. But, there are a few places in Julia where suitable Unicode alternatives are not accepted by the language, and where I feel that the ASCII-art confections currently used can be gently enhanced by the judicious use of alternate glyphs. There are also a few places where some subtle tweaks can enhance the readability of the language without introducing ambiguity.

In JuliaMono, the following substitutions are applied when the contextual alternates feature is active:

You can see these in action in the following code fragment:^[11]

julialang = true # (!= 0)
(x, y) -> (x + y)
f(p::Int) = p * p
@inbounds if f in (Base.:+, Base.:-)
    if any(x -> x <: AbstractArray{<:Number})
         nouns = Dict(
            Base.:+ => "addition",
            Base.:- => "subtraction",
        )
    end
end
df2 = df |>
    @groupby(_.a) |>
    @map({a = key(_), b = mean(_.b)}) |>
    DataFrame # <|

OpenType fonts also offer you the ability to choose different designs for certain characters. These are stored as a ‘stylistic set’.

All the options are stored in the font, and are often referred to by their internal four letter code (not the best user-oriented design, really). For example, the contextual alternates listed above are collectively stored in the calt feature.

Sometimes, an application will show the options more visually in a Typography panel^[12], usually tucked away somewhere on a Font chooser dialog.

Here’s a list of the stylistic sets currently available in JuliaMono.

All this fancy technology is under the control of the application and the operating system you’re using. Ideally, they will provide an easy way for you to switch the various OpenType features on and off.

Browser-based editors such as Juno and VS Code support many OpenType features in their editor windows, but not in the terminal/console windows. They provide a settings area where you can type CSS or JSON selectors to control the appearance of features, and you’ll have to know the feature codes. Some features are opt in, others are opt out; this too can vary from application to application.

Terminal/console applications also vary a lot; on MacOS the Terminal and iTerm applications try to offer controls for OpenType features, with varying degrees of success. On Linux, some terminal applications such as Konsole and Kitty offer quite good support, but others such as Alacritty offer little or none, as yet. ^[13]

If the application allows, you should be able to switch the calt contextual ligatures off, particularly since quite a few people won’t like any of them in their code. For the following listing, I switch the calt set off using CSS (see here), and then enable some of the alternative stylistic sets: compare characters such as the 0, g, a, j, and @ with the previous listing:

julialang = true # (!= 0)
(x, y) -> (x + y)
f(p::Int) = p * p
@inbounds if f in (Base.:+, Base.:-)
    if any(x -> x <: AbstractArray{<:Number})
         nouns = Dict(
            Base.:+ => "addition",
            Base.:- => "subtraction",
        )
    end
end
df2 = df |>
    @groupby(_.a) |>
    @map({a = key(_), b = mean(_.b)}) |>
    DataFrame # <|

(I originally liked the idea of a more circular @ sign, but in practice it doesn’t work at small point sizes, as the details disappear. But I’ve kept it anyway.)

If you really don't like any ligatures

It's possible that you don't like even the few ligatures and alternate characters provided in JuliaMono. If switching them off isn't possible, or not drastic enough, you can strip them out of the font with the following terminal magic (courtesy of @fredrikekre):

$ cat Dockerfile
FROM python
RUN pip install fonttools
COPY run.sh /scripts/
ENTRYPOINT ["/scripts/run.sh"]

$ cat run.sh
#!/bin/bash
set -euo pipefail

VERSION="${1-}"

mkdir -p /juliamono-source
mkdir -p /juliamono-output
mkdir -p /juliamono

# Download release version
curl -L "https://github.com/cormullion/juliamono/releases/download/v${VERSION}/JuliaMono.tar.gz" | tar -xzvC /juliamono-source

# Strip glyphs
for f in /juliamono-source/*.ttf
do
    pyftsubset "$f" '*' --output-file=/juliamono-output/$(basename "$f")  --layout-features-=calt,liga
done

# Pack it up
tar -C /juliamono-output -czvf /juliamono/JuliaMono-${VERSION}.tar.gz $(ls /juliamono-output)

$ cat Makefile
.PHONY: image strip
VERSION := 0.021
image:
        docker build -t juliamono-strip:latest .

strip:
        docker run --rm -v ${PWD}:/juliamono juliamono-strip:latest "${VERSION}"

There are a few areas of the Unicode system that have been officially kept empty and are thus available to store characters that are not part of the standard. These are called the Private Use Areas, and there are three: \ue000 to \uf8ff, \UF0000 to \UFFFFD, and U100000 to U+10FFFD.

Each typeface can do its own thing in these areas. In JuliaMono, for example, if you look around \ue800 you’ll find a few familiar shapes:

julia> foreach(println, '\ue800':'\ue802')




The obvious drawback to using characters in a Private Use Area is that you have to have installed the font wherever you want to see them rendered correctly, unless they’ve been converted to outlines or bitmaps. If the font isn’t installed (eg on github), you have no idea what glyph - if any - will be displayed.

You can define these to be available at the Julia REPL. For example, say you want the Julia circles to be available in the terminal when you type \julialogo in a Julia session with the JuliaMono font active. Run this:

using REPL
REPL.REPLCompletions.latex_symbols["\\julialogo"] = "\ue800"

Now you can insert the logo in strings by typing \julialogo:

julia> println("Welcome to ")
Welcome to 

It’s usually possible to type Unicode values directly into text. This is a useful skill to have when you’re not using the Julia REPL... On MacOS you hold the Option (⌥) key down while typing the four hex digits (make sure you’re using the Unicode Hex Input keyboard). On Windows I think you type the four hex digits followed by ALT then X. On Linux it might be ctrl-shift-u followed by the hex digits.

Thanks!

Thanks to: Thibaut Lienart for his Franklin.jl web site builder; to Jérémie Knüsel who provided invaluable suggestions and advice; to Dr Zygmunt Szpak for his cool maths code; to Simeon Schaub for the issues and PRs, and to others in the Julia community for help and suggestions.

Footnotes