Comparison of ave, ddply and data.table

This is a copy of a post by me on the R-statistics blog.

Fortran and C programmers often say that interpreted languages like R are nice and all, but lack in terms of speed. How fast something works in R greatly depends on how it is implemented, i.e. which packages/functions does one use. A prime example, which shows up regularly on the R-help list, is letting a vector grow as you perform an analysis. In pseudo-code this might look like:

The problem here is that dum is continuously growing in size. This forces the operating system to allocate new memory space for the object, which is terribly slow. Preallocating dum to the length it is supposed to be greatly improves the performance. Alternatively, the use of apply type of functions, or functions from plyr package prevent these kinds of problems. But even between more advanced methods there are large differences between different implementations.

Take the next example. We create a dataset which has two columns, one column with values (e.g. amount of rainfall) and in the other a category (e.g. monitoring station id). We would like to know what the mean value is per category. One way is to use for loops, but I’ll skip that one for now. Three possibilities exist that I know of: ddply (plyr), ave (base R) and data.table. The piece of code at the end of this post compares these three methods. The outcome in terms of speed is:

It is quite obvious that ddply performs very bad when the number of unique categories is large. The ave function performs better. However, the data.table option is by far the best one, outperforming both other alternatives easily. In response to this, Hadley Wickham (author of plyr) responded:

This is a drawback of the way that ddply always works with data frames. It will be a bit faster if you use summarise instead of data.frame (because data.frame is very slow), but I’m still thinking about how to overcome this fundamental limitation of the ddply approach.

I hope this comparison is of use to readers. And remember, think before complaining that R is slow :).

ps This blogpost is based on discussions on the R-help and manipulatr mailing lists:
- http://www.mail-archive.com/r-help@r-project.org/msg142797.html
- http://groups.google.com/group/manipulatr/browse_thread/thread/5e8dfed85048df99

Tagged with: ,
Posted in R stuff

Leave a Reply

Your email address will not be published. Required fields are marked *

*

To create code blocks or other preformatted text, indent by four spaces:

    This will be displayed in a monospaced font. The first four 
    spaces will be stripped off, but all other whitespace
    will be preserved.
    
    Markdown is turned off in code blocks:
     [This is not a link](http://example.com)

To create not a block, but an inline code span, use backticks:

Here is some inline `code`.

For more help see http://daringfireball.net/projects/markdown/syntax

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code class="" title="" data-url=""> <del datetime=""> <em> <i> <q cite=""> <strike> <strong> <pre class="" title="" data-url=""> <span class="" title="" data-url="">