Paul Chiusano

Functional programming, UX, tech


About my book

My book, Functional Programming in Scala, uses Scala as a vehicle for teaching FP. Read what people are saying about it.

Popular links

Unison: a friendly programming language from the future the worldwide elastic computer (coming soon)
Type systems and UX: an example
CSS is unnecessary

Unison update 4: more editor interactions

[   fp   unison   ]            

Here’s a video of the latest progress. Watch me write the expression 1 + 1, then evaluate it!! Further explanation below.

There are a number of additional editor interactions shown here:

Not shown:

I think there’s a lot more work that could be done to improve fluidity, but already, it feels pretty nice!


When the explorer shows terms which are ill-typed but which match the query, this is the closest we get to a ‘compile error’ when writing code in Unison. When your compiler is run in batch mode, the question of what information to display in the compile errors is very important. What if we miss giving some information that’s relevant to fixing the error? Thus compilers like GHC will often dump out a screenful of text for even very simple errors. There’s also entire lines of research devoted to the question of how to minimize error messages, make them maximally informative (see type error slicing), and so on.

In Unison, we don’t need to decide what information is relevant in an “error message” because the user can explore their program interactively. You are shown the type of your search result, the admissible type, the current type, and the types of all local variables, which is often enough to make it clear how to modify your code. But if you desire more information, you can navigate around to get the types of any related expressions you care about, and this can be done very quickly. The question of ‘what information to display in the compile error’ becomes irrelevant! There are no compile errors, and the user can extract the information they deem relevant interactively.

Here’s a question: should the name of the identity function be id or identity? Maybe you have some opinions about this. In Unison, we can actually store both names in the metadata associated with the term, and the search will return a term if any of the names match the search query. Regardless of our choice, it’s not an additional typing burden to have a longer name—the programmer can type id followed by two spaces to accept identity (or id, <enter>, etc). Name resolution is always type-directed, and the user specifies the minimum information necessary to disambiguate.

In working on Unison, I’ve found it’s very interesting to think about the distinction between a language feature vs an editor feature. We often conflate the two. The syntax for “imports”, name resolution, etc, are all considered part of the language. In a semantic editor, many of these concerns are better dealt with by the editor, which can incidentally be much more flexible (allowing multiple names for a given term is just one instance of this). Raw text program editing has been holding us back!

That’s all for now, more updates soon!

comments powered by Disqus