Comments for Nor Talk Too Wise http://nortalktoowise.com Tue, 01 May 2012 15:09:25 +0000 hourly 1 http://wordpress.org/?v=3.3.2 Comment on The Best Statistical Programming Language is …Javascript? by Tony http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-382 Tony Tue, 01 May 2012 15:09:25 +0000 http://nortalktoowise.com/?p=1632#comment-382 Which is precisely why I would much prefer generating plots with Highcharts than ggPlot2 (for example). However, I'd hate to try to run regressions using Javascript. Which is precisely why I would much prefer generating plots with Highcharts than ggPlot2 (for example). However, I’d hate to try to run regressions using Javascript.

]]> Comment on The Best Statistical Programming Language is …Javascript? by Tony http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-381 Tony Tue, 01 May 2012 15:06:55 +0000 http://nortalktoowise.com/?p=1632#comment-381 I couldn't agree with you more. I'm not proposing actually using Javascript for analysis (although there's a strong case to be made that it is the language for visualization). That's kind of the point of this post. There's no logical reason to do analysis in Javascript--I can read in data, run a variety of regressions, and generate breathtaking plots of the data in ~5 lines of R (2 if I'm feeling ambitious). It would take me hours of reading documentation, trying out new libraries, and running likely more than 100 lines of code to accomplish the same in Javascript. Which brings me to the ultimate point of the post, which has nothing to do with Javascript: I suspect I'll prefer R to Julia for a long time yet for precisely these reasons, despite Julia's prodigious benchmark results. I couldn’t agree with you more. I’m not proposing actually using Javascript for analysis (although there’s a strong case to be made that it is the language for visualization). That’s kind of the point of this post. There’s no logical reason to do analysis in Javascript–I can read in data, run a variety of regressions, and generate breathtaking plots of the data in ~5 lines of R (2 if I’m feeling ambitious). It would take me hours of reading documentation, trying out new libraries, and running likely more than 100 lines of code to accomplish the same in Javascript. Which brings me to the ultimate point of the post, which has nothing to do with Javascript: I suspect I’ll prefer R to Julia for a long time yet for precisely these reasons, despite Julia’s prodigious benchmark results.

]]> Comment on The Best Statistical Programming Language is …Javascript? by Alex http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-380 Alex Tue, 01 May 2012 00:56:34 +0000 http://nortalktoowise.com/?p=1632#comment-380 How fast programs run is only one of the many factors that determine the experiences with a programming language. It is not even just the programming experience, user experience should be included as well. Borrowing a popular word, I will say that it is the entire ecosystem that is the most important. There needs to be a good committee deciding where the language goes. There needs to be a creative army of developers writing libs and extensions. There needs to be a large number of people who are rather familiar with the language and are eager to spend time helping others out. So on and so forth. Even when we talk about being fast. How quick it runs 1 million loops is still only part of the story. I work in an environment where programs need to be created and modified very frequently. These programs interact with a lot of stuff but are not computationally intensive in a hardcore sense. So, to me, nice structure is important and flexibility is crucial. Saving 20 seconds while running loops is nice but one could easily waste days if using a slightly more awkward tools. How fast programs run is only one of the many factors that determine the experiences with a programming language. It is not even just the programming experience, user experience should be included as well.

Borrowing a popular word, I will say that it is the entire ecosystem that is the most important. There needs to be a good committee deciding where the language goes. There needs to be a creative army of developers writing libs and extensions. There needs to be a large number of people who are rather familiar with the language and are eager to spend time helping others out. So on and so forth.

Even when we talk about being fast. How quick it runs 1 million loops is still only part of the story. I work in an environment where programs need to be created and modified very frequently. These programs interact with a lot of stuff but are not computationally intensive in a hardcore sense. So, to me, nice structure is important and flexibility is crucial. Saving 20 seconds while running loops is nice but one could easily waste days if using a slightly more awkward tools.

]]> Comment on The Best Statistical Programming Language is …Javascript? by Teaching code, production code, benchmarks and new languages | Quantum Forest http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-378 Teaching code, production code, benchmarks and new languages | Quantum Forest Mon, 30 Apr 2012 09:37:05 +0000 http://nortalktoowise.com/?p=1632#comment-378 [...] Tony Boyles asked, tongue in cheek, ‘The Best Statistical Programming Language is …Javascript?’ Well, if you define best as fastest language running six benchmarks that may bear some resemblance to what you’d like to do (despite not having any statistical libraries) maybe the answer is yes. [...] [...] Tony Boyles asked, tongue in cheek, ‘The Best Statistical Programming Language is …Javascript?’ Well, if you define best as fastest language running six benchmarks that may bear some resemblance to what you’d like to do (despite not having any statistical libraries) maybe the answer is yes. [...]

]]> Comment on The Best Statistical Programming Language is …Javascript? by Reygun http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-377 Reygun Mon, 30 Apr 2012 01:00:32 +0000 http://nortalktoowise.com/?p=1632#comment-377 JavaScript may not be the environment with more functions but it should be noted, that to be able to use online gives us the possibility of doing that ideas are not left in the role, but become useful tools for those who want to. JavaScript in web design is a revolution and it can also be for analysis statistical if only given the chance is only question for adherents to improve their current conditions. Ideas on the internet are more useful. JavaScript may not be the environment with more functions but it should be noted, that to be able to use online gives us the possibility of doing that ideas are not left in the role, but become useful tools for those who want to. JavaScript in web design is a revolution and it can also be for analysis statistical if only given the chance is only question for adherents to improve their current conditions.

Ideas on the internet are more useful.

]]> Comment on The Best Statistical Programming Language is …Javascript? by Tony http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-374 Tony Sun, 29 Apr 2012 16:49:27 +0000 http://nortalktoowise.com/?p=1632#comment-374 I didn't run the experiment myself, just reported the results. If you're curious, the code is all on <a href="https://github.com/JuliaLang/julia/tree/master/test/perf" rel="nofollow">github</a>. I don't know Javascript Very well, but It looks like they're all untyped arrays. I didn’t run the experiment myself, just reported the results. If you’re curious, the code is all on github. I don’t know Javascript Very well, but It looks like they’re all untyped arrays.

]]> Comment on The Best Statistical Programming Language is …Javascript? by Tony http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-373 Tony Sun, 29 Apr 2012 16:43:36 +0000 http://nortalktoowise.com/?p=1632#comment-373 1) The table comes from the Julia Homepage. 2) I've come to the conclusion that nothing is typical in any sort of programming. So maybe they are a fair representation, and maybe they aren't. I won't lose sleep over it. 3) I didn't run the experiment myself, just reported the results. I have no idea. 1) The table comes from the Julia Homepage.
2) I’ve come to the conclusion that nothing is typical in any sort of programming. So maybe they are a fair representation, and maybe they aren’t. I won’t lose sleep over it.
3) I didn’t run the experiment myself, just reported the results. I have no idea.

]]> Comment on The Best Statistical Programming Language is …Javascript? by Tony http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-372 Tony Sun, 29 Apr 2012 16:34:39 +0000 http://nortalktoowise.com/?p=1632#comment-372 My Mistake! I was swayed to think of it as a fork because it stands alone so easily (which is the only way I've used it). My Mistake! I was swayed to think of it as a fork because it stands alone so easily (which is the only way I’ve used it).

]]> Comment on The Best Statistical Programming Language is …Javascript? by Bruce Durling http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-368 Bruce Durling Sat, 28 Apr 2012 23:37:04 +0000 http://nortalktoowise.com/?p=1632#comment-368 Incanter isn't a fork of clojure. It is a clojure library. Incanter isn’t a fork of clojure. It is a clojure library.

]]> Comment on The Best Statistical Programming Language is …Javascript? by Robert Young http://nortalktoowise.com/2012/04/the-best-statistical-programming-language-is-javascript/#comment-365 Robert Young Sat, 28 Apr 2012 14:24:19 +0000 http://nortalktoowise.com/?p=1632#comment-365 R is unusual, if not unique, in its schizophrenic nature: stat pack command language and "general" programming language, all with, more or less, the same syntax. Reviews of R's codebase, older and recent, usually are surprised at 1) how much of R is written in R or 2) how much of R is written in C/Fortran or linked to routines written in C/Fortran. Serving two masters seldom works out well, and R demonstrates that. With Rcpp and similar, I expect we'll see more of R packages being written "real" languages, while R syntax will be used more as a command language, a al SPSS/SAS/etc. R is unusual, if not unique, in its schizophrenic nature: stat pack command language and “general” programming language, all with, more or less, the same syntax. Reviews of R’s codebase, older and recent, usually are surprised at 1) how much of R is written in R or 2) how much of R is written in C/Fortran or linked to routines written in C/Fortran. Serving two masters seldom works out well, and R demonstrates that.

With Rcpp and similar, I expect we’ll see more of R packages being written “real” languages, while R syntax will be used more as a command language, a al SPSS/SAS/etc.

]]>