[R-SIG-Mac] A bug in the Mac GUI and a request

[Simon Urbanek]

Ivan,

On Sep 28, 2006, at 10:04 AM, Ivan Alves wrote:

> First the bug:
> R GUI version 1.17-pre (3821), Mac OS X system 10.4.7 on intel core  
> duo.  Whenever (or almost)  I save the content of the quartz device  
> (the graphics window) to a file, the window "Save Content of Quartz  
> Device to PDF File" remains (or rather comes back) after pressing  
> on the save button.  Furthermore this window cannot be closed,  
> except by quiting the R GUI.
>

Ok, we'll have a look. There are issues with Cocoa since 10.4 that  
are still plaguing us ...


> Now the request:
> UTF-8 encoding support.  Since I have started working with csv  
> files encoded in UTF-8 (which appears to be the native.enc in OS X)  
> I have had quite a headache subsetting dataframes by values of a  
> given factor (with text encoded in UTF-8 I pressume).  For instance  
> "Compañia" appears as "Compa\361ia".

The problem here is that your are *not* using UTF-8. UTF-8 encoding  
of "ñ" is (if printed as octal codes) "\303\261". So what you see is  
something that is in a different encoding - AFAIC it is latin1 as you  
can see easily:

In command-line R (without locale):
 > iconv("Compa\361ia","latin1","UTF-8")
[1] "Compa\303\261ia"

In the GUI:
 > iconv("Compa\361ia","latin1","UTF-8")
[1] "Compañia"

So when you are loading your table, you should be using something like
read.table(file("my.table",encoding="latin1"))
to get the correct result.

You may want to read Brian Ripley's article about encodings and  
localization in the R News Vol 5/1.

Cheers,
Simon