[Rd] Making iconv portable?

[Milan Bouchet-Valat]

Le lundi 15 décembre 2014 à 13:49 -0500, Simon Urbanek a écrit :
> On Dec 15, 2014, at 1:37 PM, Spencer Graves <spencer.graves at prodsyse.com> wrote:
> > 
> > 
> >> On Dec 15, 2014, at 10:13 AM, Simon Urbanek <simon.urbanek at r-project.org> wrote:
> >> 
> >>> 
> >>> On Dec 15, 2014, at 12:21 PM, Kurt Hornik <Kurt.Hornik at wu.ac.at> wrote:
> >>> 
> >>>>>>>> Spencer Graves writes:
> >>> 
> >>>> Hello, All:  
> >>>> 	  What would it take to make “iconv” portable?  
> >>> 
> >>> 
> >>>> 	  I ask, because I want to convert accented characters to
> >>>> 	  vanilla ASCII, thereby converting, e.g., ‘Raúl’ to “Raul”, and
> >>>> 	  Milan Bouchet-Valet suggested on R-help that I use 'iconv(x,
> >>>> 	  “", "ASCII//TRANSLIT”)’.  This worked under Windows but failed
> >>>> 	  on Linux and Mac.  It’s part of the “subNonStandardCharacters”
> >>>> 	  function in the Ecfun package. The development version on
> >>>> 	  R-Forge uses this and returns “Raul” under Windows and NA
> >>>> 	  under Mac OS X (and presumably also Linux).
> >>> 
> >>> Hmm.
> >>> 
> >>> R> iconv("Raúl", "", "ASCII//TRANSLIT")
> >>> [1] "Raul"
> >>> 
> >>> seems to work for me on Linux ...
> >>> 
> >> 
> >> also on OS X:
> >> 
> >>> iconv("Raúl", "", "ASCII//TRANSLIT")
> >> [1] “Ra'ul"
> > 
> > 
> > 	  Thanks for the replies.  I should have checked my examples more carefully.  Consider the following example and a slight modification from help(“iconv”):  
> > 
> > 
> > > x <- c("Ekstr\xf8m", "J\xf6reskog", "bi\xdfchen Z\xfcrcher")
> > > Encoding(x) <- "latin1"
> > > x
> > [1] "Ekstrøm"               "Jöreskog"              "bißchen Zürcher"      
> > > iconv(x, "latin1", "ASCII//TRANSLIT")  # platform-dependent
> > [1] "Ekstrom"            "J\"oreskog"         "bisschen Z\"urcher"
> > > 
> > > x <- c("Ekstr\xf8m", "J\xf6reskog", "bi\xdfchen Z\xfcrcher")
> > > x
> > [1] "Ekstr\xf8m"            "J\xf6reskog"           "bi\xdfchen Z\xfcrcher"
> > > iconv(x, "", "ASCII//TRANSLIT")  # platform-dependent
> > [1] NA NA NA 
> > 
> > 
> > 	  This suggests a two-step fix to my problem:  (1) Check Encoding(x) and set to “latin1” if it’s “unknown”.
> 
> Well, that depends heavily on your source. In the above it is hand-crafted latin1 so if you don't declare it, the native encoding will be assumed - which can be anything and has nothing to do with your actual input in this particular case where you hand-constructed latin1.
> 
> 
> >  (2) Delete any new \” added by iconv.  
> > 
> 
> The whole point of translit is to create combinations of ASCII
> characters that represent the unicode characters, so " is just one
> many characters that can be used.
But it's quite unexpected that ö is transliterated to "o and ú to 'u.
Looks like iconv on OS X has a different idea of what ASCII
transliteration means than on Linux and Windows...

Anyway it's easy to remove " and ' if needed.


Regards

> Cheers,
> S
> 
> 
> > 
> > 	  Thanks again, 
> > 	  Spencer 
> > 
> >> 
> >> 
> >> 
> >>> -k
> >>> 
> >>> 
> >>>> 	 The “iconv” R code merely calls compiled code, which I’ve used very little in 30 years.   
> >>> 
> >>> 
> >>>> 	  Thanks, 
> >>>> 	  Spencer 
> >>> 
> >>> 
> >>> 
> >>>>> On Nov 30, 2014, at 2:32 AM, Spencer Graves <spencer.graves at structuremonitoring.com <mailto:spencer.graves at structuremonitoring.com>> wrote:
> >>>>> 
> >>>>> Wonderful.  Thanks very much.  Spencer
> >>>>> 
> >>>>> 
> >>>>> On 11/30/2014 2:25 AM, Milan Bouchet-Valat wrote:
> >>> 
> >>>> 	[[alternative HTML version deleted]]
> >>> 
> >>>> ______________________________________________
> >>>> R-devel at r-project.org mailing list
> >>>> https://stat.ethz.ch/mailman/listinfo/r-devel
> >>> 
> >>> ______________________________________________
> >>> R-devel at r-project.org mailing list
> >>> https://stat.ethz.ch/mailman/listinfo/r-devel
> > 
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel