3 ms·
I’m not totally convinced that being ‘secretly a lisp’ is what was good about R. I think the easy vectorisation is good, and the consequences of the bizarre fun
by dan-robertson 3y ago
I’m not totally convinced that being ‘secretly a lisp’ is what was good about R. I think the easy vectorisation is good, and the consequences of the bizarre function argument evaluation are good. I don’t know of lisps that do the vectorisation stuff so naturally, and while I guess fexprs are a thing, I think they are possibly too general in the syntax they can accept – basically the simplicity of lisp syntax allows macros to have more tree-structured input in a way you wouldn’t want for a language with non-lisp syntax (where the head lives outside the list), and I think the flexibility makes the syntax more confusingly non-uniform.
- levocardia 3y agoA lot of R's popularity and usefulness has to do with the libraries that are available in it. I would put up with almost any amount of BS from base R to use ggplot and the tidyverse, and ditto for a number of modern stats algorithms. In many cases, Python implementations of the same techniques are either woefully outdated or completely nonexistent.
- dan-robertson 3y agoWhen I’ve seen attempts at ggplot2 or dplyr in other languages, one issue is bad performance or bugs, but it’s also been a problem that the language features seem to allow those libraries to be much more ergonomic. eg I found Julia much less nice to use for those sorts of things despite it seeming light it ought to be well suited (making a reasonably good claim to a lot of CL heritage for example)
- disgruntledphd2 3y agoPlotnine is a pretty good implementation of ggplot2, apart from the necessity to quote variable names it seemed to work almost perfectly.