I'm sure at least some of these will turn out to be horribly mis-informed, but here's my take on R having messed with it for about two hours and not written anything that would qualify as an assignment even for an undergrad.
Like most interpreted languages, it supports stream of consciousness exploration quite well. Also, like most interpreted languages, it's very easy to get 20 lines in and realize that you should have done things in a completely different order.
I have no idea how you apply Test Driven Design to this language. Maybe you're not supposed to.
That statistical library appears to be very deep. And, if you can't find it in the standard libs, there's a huge 3rd-party collection to draw from.
Making matrices column major by default seems weird. I guess that's because extracting a column from a data table is a much more common operation than extracting a row (rows are observations, columns are attributes) and they wanted to keep matrices and data tables similar.
The default vector multiplication operation seems odd, to say the least. I struggle to think of a time when you would want to create a vector whose elements were the products of the corresponding elements of the arguments. And, extending it to a mixed-length operator by wrapping the shorter one boggles my mind. Why not define it as inner or outer product?
I think my next step is to use it to crank through some practical problems from a stats text. My Design of Experiments book from undergrad has problems in a lot of different forms. I may start to see some patterns to using the language.
No comments:
Post a Comment