This week I’ve been attending SPLASH 2018. It’s already been quite an interesting week, and it’s only just begun. I thought I’d write up a quick report on some of the things that have been particularly interesting to me, and some of the ideas that they’ve sparked off.
Teaching programming (and Rust!)
I really enjoyed this talk by Felienne Hermans entitled “Explicit Direct Instruction in Programming Education”. The basic gist of the talk was that, when we teach programming, we often phrase it in terms of “exploration” and “self-expression”, but that this winds up leaving a lot of folks in the cold and may be at least partly responsible for the lack of diversity in computer science today. She argued that this is like telling kids that they should just be able to play a guitar and create awesome songs without first practicing their chords1 – it kind of sets them up to fail.
The thing that really got me excited about this was that it seemed very connected to mentoring and open source. If you watched the Rust Conf keynote this year, you’ll remember Aaron talking about “OSS by Serendipity” – this idea that we should just expect people to come and produce PRs. This is in contrast to the “OSS by Design” that we’ve been trying to practice and preach, where there are explicit in-roads for people to get involved in the project through mentoring, as well as explicit priorities and goals (created, of course, through open processes like the roadmap and so forth). It seems to me that the things like working groups, intro bugs, quest issues, etc, are all ways for people to “practice the basics” of a project before they dive into creating major new features.
One other thing that Felienne talked about which I found exciting was the idea that – in fields like reading – there are taxonomies of common errors as well as diagnostic tools that one can use to figure out where your student falls in this taxonomy. The idea is that you can give a relatively simple quiz that will help you identify what sorts of mental errors they are making, which you can then directly target. (Later, I talked to someone – whose name unfortunately I do not remember – doing similar research around how to categorize mathematical errors which sounded quite cool.)
I feel like both the idea of “practice” but also of “taxonomy of errors” applies to Rust quite well. Learning to use Rust definitely involves a certain amount of “drill”, where one works with the mechanics of the ownership/borrowing system until they start to feel more natural. Moreover, I suspect that there are common “stages of understanding” that we could try to quantify, and then directly target with instructional material. To some extent we’ve been doing this all along, but it seems like something we could do more formally.
Borrow checker field guide
Yesterday I had a very nice conversation with Will Creigthon and Anna Zeng. They were presenting the results of work they have been doing to identify barriers to adoption for Rust (they have a paper you can download here to learn more). Specifically, they’ve been surveying comments, blog posts, and other things and looking for patterns. I’m pretty excited to dig deeper into their findings, and I think that we should think about forming a working group or something else to continue this line of work to help inform future directions.
Talking with them also helped to crystallize some of the thoughts I’ve been having with respect to this “After NLL” blog post series. What I’ve realized is that it is a bit tricky to figure out how to organize the “taxonomy of tricky situations” that commonly result with ownership as well as their solutions. For example, in reading the responses to my previous post about interprocedural conflicts, I realized that this one fundamental conflict can manifest in a number of ways – and also that there are a number of possible solutions, depending on the specifics of your scenario.
I’ve decided for the time being to just press on, writing out various blog posts that highlight – in a somewhat sloppy way – the kinds of errors I see cropping up, some of the solutions I see available for them, and also the possible language extensions we might pursue in the future.
However, I think that once this series is done, it would be nice to pull this material together (along with other things) into a kind of Borrow Checker Field Guide. The idea would be to distinguish:
- Root causes – there are relatively few of these, but these are the root aspects of the borrow checker that give rise to errors.
- Troublesome patterns – these are the designs that people are often shooting for from another language which can cause trouble in Rust. Examples might be “tree with parent pointer”, “graph”, etc.
- Solutions – these would be solutions and design patterns that work today to resolve problems.
- Proposals – in some cases, there might be links to proposed designs.
The idea is that for the root causes and troublesome patterns, there would be links over to the solutions and proposals that can help resolve them. I don’t intend to include a lot of details about proposals in particular in this document, but I’d like to see a way to drive people towards the “work in progress” as well, so they can give their feedback or maybe even get involved.
An Eve retrospective
Chris Granger gave an amazing and heartfelt talk about the Eve effort to construct a more accessible model of programming. I had not realized what a monumental effort it was. I had two main takeaways: first, that the crux of programming is modeling and feedback. Excel works for so many people because it gives a simple model you can fit your data into and immediate feedback – but it’s inflexible and limited. If we can change that up or scale it, it could be a massive enabler. Second, that the VC approach of trying to change the world over the course of a few ridiculously high-pressure years sounds very punishing and doesn’t feel just or right. It’d be wonderful if Chris and co. could continue their efforts without such personal sacrifice. (As an aside, Chris had a few nice things to say about Rust, which were much appreciated.)
Incremental datalog and Rust type checking
Finally, I spent some time talking to Sebastian Erdweg and Tamás Szabó about their work on incremental datalog. They had a very cool demo of their system at work where they implemented various analyses – unreachable code and out-of-bounds index detection – and showed how quickly they could update as the input source changed. Sebastian also has a master’s student (whose name I don’t know – yet! but I will find out) that implemented a prototype Rust type checker in their system; I look forward to reading more about it.
Their system is much finer grained than anything we’ve attempted to do in rustc. It seems like we could easily port Polonius to that system and see how well it works, though it seems like it would also make sense to compare against Frank McSherry’s amazing differential-datalog system.
I’ve been thinking for some time that the next frontier in terms of improving rustc is to start formalizing and simplifying name resolution and type-checking. Talking to them did get me more inspired to see that work proceed, since it could well be the foundation for super snappy IDE integration. (But first: got to see Polonius and Chalk over the finish line!)
Finally, I had a very long conversation with Will Byrd and Michael Ballantyne about how Chalk works. We discussed some details of how MiniKanren’s search algorithm works and also some details how Chalk lowering works. I won’t try to summarize here, except to say that Will gave me an exciting pointer to something called “default logic”, which I had never heard of, but which seems like a very good match for specialization. I look forward to reading more about it.
Me talking about Rust
I am posting this in the morning on Thursday. Today I am going to give a talk about Rust – I plan to focus on both some technical aspects but also how Rust governance works and some of the “ins and outs” of running an open source project. I’m excited but a bit nervous, since this is material that I’ve never tried to present to a general audience before. Let’s see how it goes![^video] (Tonight there is also a joint Rust-Scala meetup, so that should be fun.)
My daughter insists she can do this… let’s just say she’s lucky she’s so cute. =) ↩