Introduction to Filament, a Humanist Language

Filament is a humanist programming environment. Think of it as a mixture of Mathematica, Scratch, and open datasets, swirled together into modern UX .

Filament is meant to be accessible to all ages and levels of skill, from children and novice programmers to artists, scientists, and even professionals who are doing exploratory coding. The syntax is designed to be simple to understand and forgiving of errors.

Though Filament uses some advanced computational concepts from existing array , functional , and object oriented programming languages, these concepts are presented in a form that is simple as possible while still being useful.

Filament can be used by itself, but is generally meant to be used within a notebook interface , typically on the web.

Notable features

  • All math operations are really functions and can be applied to scalars, lists, and larger objects, as in array programmming langauges like APL ). ex: 4 * [1,2,3] = mulitply(4,List(1,2,3)) = [4,8,12]
  • Identifiers and numbers are case-insensitive and can have _'s anywhere in them without changing the meaning. ex: is_prime = IsPrime and 4_000_000 = 4000000 = 4*1000*1000 .
  • All numbers have optional units and can perform unit arithmetic and conversions. ex: How fast can Superman fly around the world? earth.radius * 2 * π / 4000ft/s as hours = 9.11 hours
  • Reactive like a spreadsheet. When a function or variable is re-evaluated to return a new result, any dependent expressions are automatically re-evaluated as well.
  • Variable assignment and function application are flexible so that the code can be shaped depending on needs of the problem. Ex: draw 10 circles with random radiuses lined up horizontally .

    • Rather than nesting a series of functions

    draw(row(map(random(10),(r)=>circle(radius:r))))

    they can be pipelined with the >> operator

    range(10) >> map(r=>circle(radius:r)) >> row() >> draw()

    making them easier to understand.

    To set a variable either << or >> can be used. This emphasizes that data is moving into the variables and then used later. Same example as above, with intermediate variables.

    nums << random(10) nums >> map(radius => circle(radius)) >> circles circles >> row() >> draw()
  • Multiple syntaxes . The underlying AST can be represented in one of three ways: a pure ASCII version (as in examples above), a print ready version that uses unicode glyphs where appropriate but would be harder to type without editor support (pi = π , theta = ø , << and >> = arrow symbols, etc.), and finally a visual block syntax similar to Scratch ) (appropriate for younger programmers who are just getting started). All three syntaxes are fully interchangeable.
  • Keywords and operators are very opinionated, and chosen to be understandable by non-programmers. The boolean operators are and , or , and not rather than && , || , and ! , though these may optionally be available for advanced users if they want. The same with << >> used for assignment and pipelines rather than bit shifting. Many of the more confusing C derived standard operators have been removed, like ++ and += .
  • Humanist APIs. The standard library APIs are designed to be consistent and easy to combine, working together in a holistic way that feels natural. For example, the graph label function can take text, numbers, colors, images, or urls and will just do the right thing. Functions generally do the best they can with whatever you give them, and complain visually when they have can't. Check out some of the APIs
  • Built in Datasets. The environment and language are designed to easily access local and remote datasets to do interesting computations. Furthermore it will come with built in datasets of known facts, such as countries of the world, planets in the solar system, scrabble word scores, and more.

    • Hypothesis

    With the right abstractions and syntax is it possible to make a programming environment that is simple enough for kids but powerful enough for domain experts. It must have a gentle learning curve and provide value at every step.

    Risks

  • Problems that fall outside the happy path become very complex to solve. For example, currently the language has no control flow or custom functions. You can get pretty far without these, but not everwhere. It could end up not being very general purpose.

    • Filament takes inspiration from

  • Frink
  • APL, J, Kerf
  • Jupyter & Python notebook programming
  • Mathematica
  • TallyCat , an earlier unit based calculator I wrote.

    • More info:

  • Tutorial
  • API
  • prototype implementing just the unit arithmetic.