Programmers don't understand that it doesn't matter what the default target is for a type like Either as long as there is one and it never changes. I go into some detail and justification of a programming language design that makes decisions like this inherent to the structure of the type rather than author's (arbitrary and harmful) preference. The Left and Right of Either do not mean anything in and of themselves.
Haskell programmers that often forget how hard it was to learn and use Haskell. They also forget that without the benefit of a code example that does precisely what one wants, it can be nearly impossible for someone to make forward progress unless they have a solid foundation in the language itself. We justify how this reality influences the way we write the Haskell Programming from first principles book.
Sometimes building a package with dependencies on OpenGL can be a little confusing for people new to Haskell. I will use a simple game on Hackage as an exercise for demonstrating how you might do this on Ubuntu. Note: this was written before Stack was a thing.
I often get asked why I do or do not recommend a particular resource for learning Haskell. To save some time and to record more detail than I'll remember off the top of my head, I've written a round-up of the issues with various resources and books people use to learn Haskell.
This is a simple follow-along with a Python tutorial, with my version in Haskell and a running commentary. I wrote this partly to hint at why one might want a statically typed language regardless of how one thinks prototyping is best done. It was also partly cathartic.
This is a quasi-literate version of the very simple URL shortener I wrote in Haskell with Scotty.
I answered a question on Stack Overflow that I thought might be worth sharing here so that others might get an idea of how to refactor code into point-free style.
I respin an article originally written in Scala into Haskell, then veer off into the stratosphere.
I rewrote somewhat difficult to understand templating code that was originally in Clojure into much simpler Haskell and yielded a large performance benefit for my trouble.
Written in Scotty. The code is not great.
The previous edition of this post was cleaned up by a suggestion from another Haskeller. I share it with you here.
Aeson is a real joy to use once you get into the swing of things, but there are some patterns out there that end-users are left to discover for themselves. One of those is how to deal with data that has a lot of instances of the Maybe type.
This is an extraction of a conversation between multiple Haskellers in an IRC channel. I'm not identifying who said what and the ordering will be more topical than temporal. I'll add annotations for context as appropriate. I edited liberally.
Need to parse data with varying structure? Once again we resort to our old friend for handling exclusive possibilities, the sum type.
Sometimes one knows only part of the structure to be parsed out of JSON ahead of time, with some of that structure being defined by a user or consumer of the API. The solution to this in general and when using Aeson to make the wrapper type parametric.
A simple demonstration of extracting boilerplate from Aeson code handling a Just and a Nothing case of a Maybe value.
Recently I had an experience where in the course of helping somebody with a problem, I developed an example that I thought would help people understand sum types, value constructors, and type constructors better.
In my last post, I described IO (IO ()) as being a sign you might've made a mistake unless you knew it was what you wanted. There are patterns which involve using a closed-over mutable reference for things like counters. This naturally leads to nested IO actions.
While playing around with querying ElasticSearch I bumped into something that I hadn't really understood explicitly before about monads, nesting, and IO. Rather than blather on, I'm going to share a "literate" ghci session that demonstrates the point. Main editing change I made was to remove duplication in the output from querying the type :t in ghci.