Hacker News Reader

Comment by geebee | Hacker News Reader

Julia does seem promising. Interesting that you see it as more along the lines of python than R.

I just finished the data science coursera class, and while we used Python, we didn't get into R. I've played around a little R on my own, and while I certainly don't think it's too difficult to learn for a programmer with a math background, as a programmer I feel much more at home with Python. Given the choice, I'd rather use Python than R just because it feels more natural to me.

If Julia is a truly a programming language, I'd agree that it would be more of a competitor with python than R (in the sense that its audience would be people like me who would lean toward python)... but I think it could be very successful by making it very easy for a programmer to stay within a programming language. In other words, it wouldn't compete with R, it would compete with python by bringing what you get from R to a programming language.

View on HN

(Hi there, I took the coursera course too!) I disagree with your compare and contrast of R as not-a-programming-language. R is absolutely a real language in every sense of the term. R, and its precursor SPlus, both have a very lispy feel. If you should ever try scheme after learning R, the biggest superficial difference you'll find is that the parentheses move:

function(x1, x2) becomes (function x1 x2)

R has a number of warts as a language (e.g. how it deals with copying objects in memory, amongst others), however it's just as "real" as Python, which also has its warts.

I do like the direction Julia's heading, even if I'd prefer to see something like Clojure/Incanter, or better still, something like Racket be the next baton holder.

Oh, I'm not too surprised to hear that R isn't a full fledged programming language. I found python very natural to use, since I currently program mainly in Ruby (and, Java/C++ a while ago). Python felt easier for me than R, but that's probably related to my background.

Some of this may also simply be a result of doing some exercises in Python (in the coursera class), and not doing any R. I was pretty bummed that we didn't't get at all into R in the coursera class, because sometimes just getting set up a little bit (like having some sample skeleton files for homework) and a few exercises can make a big difference. Not meant as a knock on the course, which was free and (I thought) very good. But a bit of exposure to R would be helpful (while I appreciate the importance of visualization, I do think that given the choice between tableau and an intro to R, I'd definitely have preferred R).

I was pleasantly surprised by Tableau. While I feel ggplot2 in R is the platinum standard for sane statistical visualizations, it was clear that the Tableau people have put some thought into some kind of underlying grammar. I'd love to see a ggplot style interface to Tableau.

R was really where I started to see the world through functional eyes. I originally came to R with a background involving a little C, a little Perl, a little ObjC, and a whole lot of Python.

Eventually, I hit a point with R where I was trivially doing pretty complicated things that would have been... ugly in Python. Even my Python now looks more like R. This was partly due to finding Peter Norvig's Udacity course, but also partly due to the fact that his somewhat un-"pythonic" python style really, really resonated with my R-tainted brain.

>Eventually, I hit a point with R where I was trivially doing pretty complicated things that would have been... ugly in Python.

I'm curious; can you give any examples?

There have been others, but this was a pretty good example of 1.) munging data easily 2.) having libraries available that did what I was looking for

https://gist.github.com/pschmied/5903410

I make no guarantees about the quality of my hackish code :-)

I agree that Tableau was more impressive than I expected, however I also think mere pivot tables in Excel are under appreciated for perhaps their biggest strength—many mathematically-minded and analytical in starting positions who do not work in the IT department supporting desktop workstations. For people in these positions even Python may be considered too low level, yet, they already have Excel, and if they have the budge for Tableau they can get that too without threatening any sys admin types.

Anyway, on the topic of R, I highly suggest the Johns Hopkins data related courses on Coursera. Many of them use R as the central tool, and the three I've taken really stood out for how much the instructors reminded me of getting a private tutorial from a bright colleague on an area of their own expertise.

Based on your description, you'd probably already know most of what would be covered, but the Roger Peng class on "Computing for Data Analysis"[1] is starting again soon, and it includes an overview of R that might have some gems for you. I liked that the lectures were relatively succinct, and the assignments put it in practice.

[1] https://www.coursera.org/course/compdata

oh man. Too late to edit, but I meant to say R is a full fledged language, not isn't.

be the next baton holder

For X to be the next baton holder, X would need to be quite widely adopted, and the superficially matlab-like look and feel of Julia should really help in this.

The (self-described) bland, Python-esque syntax is absolutely a boon for mass adoption amongst those who feel that Python-eqsue syntax is "natural." That, however, doesn't change my preferences for parentheses, or my wish for people to stop fearing the S-Exprs!

Reply on news.ycombinator.com