Comments on: A Fast Intro to PLYR for R

By: JD Long

JD Long — Thu, 07 Apr 2011 18:49:12 +0000

Good point Wabe. I didn’t use summarize in the example because I wanted to illustrate a design pattern applicable to functions which are not part of summarize. But I agree that summarize is pretty dang handy!

By: Wabe

Wabe — Thu, 07 Apr 2011 18:22:46 +0000

A neater way to achieve the same result and at the same time set appropriate column names is to write:
ddply(dd,c(“dim1″,”dim2″),summarize,mean1=mean(v1),mean2=mean(v2),mean3=mean(v3),sd1=sd(v1))

By: Ethan Brown

Ethan Brown — Mon, 24 Jan 2011 05:50:20 +0000

Dearest Sir,

Once again you were the right search result at the right time. What a handy tool and a handy post! My night is saved.

BTW, it was good to meet you many moons ago at useR! 2010 at the RUG session–we now have a lively and excellent local group here in Denver thanks in part to your inspiration.

By: Lavinia

Lavinia — Mon, 18 Oct 2010 05:42:49 +0000

Thanks for the worked example, very useful, I’ve looked at plyr but never really appreciated how it worked.

By: JD Long

JD Long — Wed, 18 Aug 2010 14:06:15 +0000

plyr is certainly not as fast as sqldf() which uses the sqlite engine. However, this is only a relevant comparison in situations where an operation can be done in both plyr and sqldf(). These situations are actually quite few. plyr works not only on data frames, but also on lists, matrices, arrays, etc. In addition, sqlite (and SQL in general) only supports a very limited set of aggregation functions. plyr supports any aggregation function you can write plus all the functions included in base R and the thousands of CRAN packages. So it’s great to compare speed, but be forewarned that this comparison is only relevant in a very small subset of cases!

Thanks for reading and posting your thoughts!

By: dontpanic

dontpanic — Wed, 18 Aug 2010 08:54:54 +0000

Thanks for the article, very interesting…

i’ve made a litte performance test with a quiet large dataframe (~500000 rows, 5 cols). It looks like sqldf() is much faster (9 secs) than ddply() (15 secs) grouping by two columns and then sum() of a third column.