[Rd] A couple of issues with colClasses/setAs

[Prof Brian Ripley]

On 8 Sep 2004, Peter Dalgaard wrote:

> Peter Dalgaard <p.dalgaard at biostat.ku.dk> writes:
> 
> > To make it actually work, we should probably fixup the "what" that is
> > being passed to scan a bit further upstreams.
> > 
> > 
> > > Might be a good idea to teach colClasses about "factor".
> > 
> > That's what I thought. Other ideas would be to predefine some standard
> > date classes (it's a bit annoying that there's no way to give
> > auxiliary information like formats), and maybe to allow a second
> > header line containing class names.
> 
> Done already, I see (except the class-header idea, which probably
> requires more thought). Thanks.

I've done some.  If you allow columns to be dropped it gets increasing 
hard to work out what as.is for example refers to, so I am looking at 
allowing named vectors.

> While we're in there should we perhaps consider adding pseudo-classes
> 
> "Ymd"
> "dmY"
> "mdY"
> "Y/m/d"
> "d/m-Y"
> "m/d/Y"
> 
> etc?  Or perhaps it is better with a catchall "Date-fmt", i.e., e.g.,
> "Date-%Y%m%d"? That'd be fairly easy to code:
> 
> else if (length(grep("^Date-",colClasses[i])))  
>         as.Date(data[[i]], sub("^Date-(.*)$","\\1", colClasses[i]))
> 
> although perhaps not so easy to document concisely.

Is it really so hard to convert such columns from character afterwards?


-- 
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