2 ms·
There is some utility in carving out a subset of a language to make the remaining part easy to comprehend and easy to contribute to. C++ is a large and complica
by kccqzy 7y ago
There is some utility in carving out a subset of a language to make the remaining part easy to comprehend and easy to contribute to. C++ is a large and complicated language but every team seems to be using a different subset of it. The same thing would happen with Haskell. Every one might have a different idea of where to draw the line for restricting advanced features. The author might decide that Haskell 98 plus OverloadedStrings is good enough. That's a valid stylistic choice. I would have drawn the line further; I personally think features to simulate dependent types aren't worth it (think singletons or DataKinds or TypeInType) but lens is absolutely necessary even though it is difficult for beginners to understand. But ask a colleague of mine, and you might receive a response that we should embrace complicated type-level programming like the kind seen in Servant, but instead restrict TemplateHaskell. All these are valid stylistic choices but they fragment the community. And eventually debugging other people's code bases (say dependencies) you would sooner or later have to face features you don't use.
Let us not forget the roots of Haskell as a research language where ideas in functional programming are to be tried out. When it emerged in the 1980s it was literally because a committee wanted a solid foundation to replace a disparate mélange of functional programming languages. It succeeded in that, and immediately introduced new features then considered highly novel (e.g. type classes). In this sense, Haskell will never be as practical and pragmatic as Go, where moderately modern language features aren't even in the language. Choosing a language like Go could be a valid choice, and so is choosing a language like Haskell.