Programming in R makes serious statistics serious fun

R is becoming huge in the bioinformatics world (I used it a lot for my own research), but I wouldn't go so far as to compare it to a general-purpose language like ECMAScript (and yes, ECMAScript is general-purpose when compared to R/SPlus). R is a good domain-specific language -- it has a great library of statistical methods for analysis and modeling -- but it's dog slow, a huge memory hog (i.e. forget large datasets), and has I/O facilities that are only a bit more advanced than what you get in FORTRAN 77.

My general research workflow was to do all of my custom logic in a faster, smaller language (i.e. Perl, C++), then to dump tab-delimited files that I could load into R for summary analysis and graphing. I would also sometimes prototype my code in R (if it involved tricky matrix math or statistical models), then re-write everything in a more capable language when it came time to work with real data.

That said, R has its strengths -- the graphics language is really powerful, once you take the time to learn it. Kind of like a functional interface to postscript...