Revolutions

In case you missed it: August roundup

In case you missed them, here are some articles from last month of particular interest to R users.

This post listed seven reasons of an anthropology professor for using R.

This post linked to an intro of the ply package for performing SQL-like "group by" operations on data frames.

This post linked to two recently released books: Hadley Wickham's ggplot2 book and the O'Reilly book Beautiful Data.

This post reviewed Friedrich Leisch's 2008 tutorial on creating packages for R.

This post linked to an R-based example on Wikipedia demonstrating pie chart failure.

This post announced that the foreach package now supports snow as a parallel backend with the release of the doSNOW package.

This post announced the "applications" category on Revolutions, showcasing applied uses of R.

This post reviewed the e-book Portfolio Analysis with R/Rmetrics.

This post linked to a "mathesaurus" of idioms translated between R, Matlab, Python, and numpy.

This post linked to an introduction to R for actuarial use.

This post linked to Drew Conway's video presentation on social network analysis in R.

This post introduced Google's coding standards for R.

This post linked to an R analysis of fish imports/exports.

This post announced the formation of a DC-area user group, and called for speakers for the inaugural meeting in October.

This post gave an example of parallel-processing a data frame in blocks with "isplit" from the iterators package.

This post linked to two upcoming R courses on graphics and data mining.

This post noted some highlights of the "fortunes" package.

This post revealed that the tech industry analysts TechCrunch now use R to produce their reports.

This post linked to some recent news articles mentioning R and REvolution R.

This post was my belated review of DSC 2009 in Copenhagen.

This post linked to Barry Rowlingson's map of CRAN mirrors.

Other non-R-specific stories in the last month covered Bonnie Tyler, errors in published journal articles, XML and big data , Chernoff midriffs, deep space, massive trees, the Neflix prizewinners, the translation equilibrium game, career prospects for Stats grads, and crows.

The R Community Calendar has also been updated.

It's been a busy month for the blog: the pie charts article was linked by the popular lifestyle blog Lifehacker, and we've surged past 1000 RSS subscribers according to Google. (You can follow the blog with any RSS feed reader like Google Reader.) As always, thanks for the comments and please send any tips to me at david@revolution-computing.com.  You can also follow me on Twitter as @revodavid.

New R User Group in Madison, WI

A new local R user group has just been formed in Madison, Wisconsin by Yury Bukhman. The first meeting will be on September 29, when Brian Yandell will give a presentation on creating R packages. Sign up to the Google Group at the link before for more info.

Google Groups: Madison Area R Users Group

Grand Central Dispatch comes to Snow Leopard

With gmail down, I've been catching up on some reading, namely Ars Technica's unfathomably in-depth review of the new MacOS update, Snow Leopard. It installed on my two Macs like a dream (only problem: had to recreate a printer that disappeared during the update, fixed in 30 seconds), but once done there's not much new to look at. Most of the changes are under the hood.

One such new core technology is Grand Central Dispatch. As Ars explains, it's Apple's new core technology to promote concurrent computing, and to get rid of the dreaded "beach ball" of an unresponsive Mac once and for all. The basic problem it solves is this: threads are a great way of making use of the processing power available in multi-core machines, but they are fiddly to work with and it's hard to know how many to use to optimize performance. GCD solves both these problems elegantly: there's a nice compiler-level syntax for parallel programming, and an eponymous central overseer to make sure that the optimal number of threads is being used at any one time.

It's purely coincidental, but I was struck by some of the design philosophies common to both GCD and our foreach package for parallel programming in R. Both are designed to optimize use of available computing resources: GCD knows how many cores are available and doles out threads accordingly; with foreach you tell it the maximum number of R processes to spawn and it doles them out in much the same way. (On the other hand, foreach's R processes still contend with other processes on the system; GCD has the advantage of being able to manage resources on a global basis. On the other other hand, foreach can span multiple workstations in a cluster, while GCD's domain is one Mac.) Both are designed to reduce the complexity of parallel programming; with GDC you can change one line of code to make a "for" loop's iterations run in parallel, and likewise with foreach. And this summation from Ars on GCD echoes my thoughts on foreach, too:

Those with some multithreaded programming experience may be unimpressed with the GCD. So Apple made a thread pool. Big deal. They've been around forever. But the angels are in the details. Yes, the implementation of queues and threads has an elegant simplicity, and baking it into the lowest levels of the OS really helps to lower the perceived barrier to entry, but it's the API built around blocks that makes Grand Central Dispatch so attractive to developers. Just as Time Machine was "the first backup system people will actually use," Grand Central Dispatch is poised to finally spread the heretofore dark art of asynchronous application design to all Mac OS X developers.

Also new to Snow Leopard is OpenCL, a new programming API opening up the computational muscle of the GPU to developers. It would be nice to exploit this for REvolution R somehow; and perhaps, if OpenCL becomes a cross-platform standard (as it seems it might), it will be possible sometime soon.

Ars Technica: Mac OS X 10.6 Snow Leopard: the Ars Technica review

Package Update Roundup: Aug 2009

This is a list of new or updated packages that were released for R in August, as announced on the r-packages mailing list. To include other updates on this list, please email David Smith. For a complete list of all updates on CRAN, see the CRANberries archive for August 2009. Follow package name links for ratings and other information on crantastic.org.

dcemri, a new package for medical image analysis, was released.

doSNOW, a new backend to foreach allowing parallel computing with snow, was released. 

Deducer, a cross-platform data-analysis GUI, has been released.

interval, a new package for weighted logrank tests on interval censored data, was released.

KFAS, a new package for multivariate Kalman filtering, smoothing, simulation and forecasting, was released.

parser, an alternate implementation of the parse function, was released.

perm, a new package for exact linear permutation tests including two-sample, k-sample and trend tests, was released. 

RSQLite, a package interfacing R with SQLite, has been updated. 

TraMineR, a package for mining, describing and visualizing sequences of states or events, has been updated.