Pop quiz: what to do for Wednesday's lecture next week?

[Bryan O'Sullivan]

Hi, folks!

Unfortunately, Conal Elliott had to withdraw from giving Wednesday's lecture next week. That means we've a hole in the schedule to fill.

I think that rather than impose something on y'all by fiat, this might be a good opportunity to let people suggest topics they'd like to hear about.

• I already have one possible guest lecturer: Mark Lentczner. He's an accomplished language designer and implementor, currently employed by Google on its Caja language project. He would most likely give an "experience report" kind of talk, about his work developing a web-based teaching framework for Haskell named Barley: https://github.com/mtnviewmark/barley
• However, that's not the only possibility. We could instead talk about performance measurement, analysis and improvement, since that got bumped off the schedule in favour of yesterday's class on parsing.
• And finally, as I mentioned, I'm totally happy to take suggestions from you as to topics you'd like to hear about (or revisit).

So let's open the floor and see what gets people excited. Let us know!

Thanks,

Bryan.

[Drew Haven]

I'd love to know more about performance profiling and optimization if we aren't going to cover it in another lecture.  With Haskell it feels like it can be non-obvious what is contributing to long running times or memory.    I feel like I find myself making semi-random changes and testing them to figure out what performs better.  How do I know which fields to make strict and which ones won't make an impact?  How can I know that I am spending too much time allocating state data types when I should pass them as function arguments (like with Attoparsec)?  With the details of Haskell's run-time being so opaque, how can we peek in and get an idea of where to start our optimizations so we can avoid making needless ones?

I would also be interested in models of user input in Haskell.  How do you write a GUI, for example?  It's not something I've looked at yet, mostly because it seems like such a large topic.

Drew Haven
drew....@gmail.com

[Phaedon Sinis]

I agree, profiling & optimization would be awesome!