[R] regexpr and portability issue

[Prof Brian Ripley]

On Tue, 2 Aug 2005, Marco Blanchette wrote:

> I am still forging my first arms with R and I am fighting with regexpr() as
> well as portability between unix and windoz. I need to extract barcodes from
> filenames (which are located between a double and single underscore) as well
> as the directory where the filename is residing. Here is the solution I came
> to:
>
> aFileName <-
> "/Users/marco/Desktop/diagnosticAnalysis/test/MA__251329410021_S01_A01.txt"
> t <- regexpr("__\\d*_",aFileName, perl=T)
> t.dir <- regexpr("^.*/", aFileName, perl=T)
> base.name <- substr(aFileName, t+2, t-2 + attr(t,"match.length"))
> base.dir <- substr(aFileName, t.dir, attr(t.dir,"match.length"))
>
> My questions are:
> 1) Is there a more elegant way to deal with regular expressions (read here:
> more easier, more like perl style).

Yes, use sub and backreferences.  An example from the R sources doing 
something similar:

             wfile <- sub("/chm/([^/]*)$", "", file)
             thispkg <- sub(".*/([^/]*)/chm/([^/]*)$", "\\1", file)

However, R does have functions basename() and dirname() to do this!

> 2) I have a portability problem when I extract the base.dir Windoz is using
> '\' instead of '/' to separate directories.

That is misinformation: Windows (sic) accepts either / or \ (see the 
rw-FAQ and the R FAQ).  Use chartr("\\", "/", path) to map \ to /.

The `portability problem' appears to be of your own making -- take heart 
that R itself manages to manipulate filepaths portably.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595