Ah... there really needs to be more literature written on switching to the
functional paradigm, IMHO.
That is really these guys haven't had an easy time with it.
On 4/11/07, Ryan Dickie
I thought I could resist this thread but I'll bite =:-()
The first language i learned was basic. No real functions, simple step by step instructions. I then learned hypercard, c, c++, python, assembly, vhdl, and too many others!
Now i've decided to learn haskell. I view it as a mathematicians language. I do research in the field of medical imaging.. particularly processing large cardiac data sets to figure out characteristics, diseases, etc. Why workflow generally starts out as 1) mathematical idea 2) turn that pure equation into a numerical recipe 3) implement, debug 5) analyze 6) goto step 1. I find doing my thinking in the continous domain makes things a lot easier.
But here's where i differ from everyone else. I already have the mathematical relationships all nice and tidy (hopefully!) in my head before i start. I literally just implement it. I don't want to care about threading, IO, message passing, or numerical stability. I have to care about performance but only so far as it hampers my productivity. Preferably the language will do it implicitly.
Your average programmer wants a language to do tasks. Having to think about the math and relationships behind it all is rather sickening to them (and me too!).
I am a new haskell programmer (basically a week into it!). It is by far the hardest language i've had to pick up. A lot of my code could be structured in a functional way.. but almost all reply on looping techniques (would take a lot of work to rethink my gradient descent method and make it fast!). Regardless of actually using haskell.. i like to transfer these techniques to c++. Before i even knew of haskell i knew some FP methods and i found that using these shrunk my code, shrunk the bugs, and did nice things for performance + concurrency. In fact, I just read a google paper on their batch system. They use to functions: map and reduce. They can easily split it up over their cluster etc... these are the ideas of FP that i like! That and the set-builder notation.
I also hate matlab to death. Is there any possibility of using haskell as a replacement using ghci? Mostly i care about linalg when it comes to using matlab.
ps: sorry if gmail butchered this reply. I had subscribed to the digest and turns out that was a mistake :D
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe