[Rcpp-devel] more on stats functions

[Romain Francois]

Le 09/12/10 04:54, Christian Gunning a écrit :
> On Wed, Dec 8, 2010 at 1:18 AM, Romain Francois
> <romain at r-enthusiasts.com>  wrote:
>>
>> It might make sense to get rid of the log = argument and maybe add a ldnorm,
>> or log_dnorm, or dnorm_log instead. so that there is no ambiguity.
>
> One nice part about the current setup is that R-exts.pdf chapter 6
> serves as a (mostly) definitive guide for arguments and behavior,
> since the Rf_* functions are at the back-end.

Makes sense. It is easy to just require that people use double and not 
int to avoid the confusion.

> The only d* example I
> found that didn't work (given that all non-missing scalar arguments
> were doubles) was:
>
> wrong = cxxfunction(signature(x='numeric'),
>    'NumericVector xx(x) ;
>    return( xx = dunif( xx, true ));',
>    plugin='Rcpp'
> )
>
> vv<- seq(0, 1, by = 0.1)
> identical(wrong(vv), dunif(vv, log=TRUE))

Thanks. rev 2749 fixes it.

>> It would be great to recycle this effort of yours into testing. Can you have
>> a look at the runit.stats.R file and add some test cases.
>
> See attached for patch with some added d* cases.  Patch is relative to
> rcpp/pkg/Rcpp/inst/unitTests.

Applied in rev 2750. Thanks.

> Is there a convenient way to run only and all of the tests in
> runit.stats.R?  Other than running *all* the Rcpp unit tests, the best
> I found was doing individual tests with:
>
> rm(.rcpp.stats)
> rm(.setUp)
> source('~/src/rcpp/pkg/Rcpp/inst/unitTests/runit.stats.R')
> .setUp()
> test.stats.dnorm()
>
> best,
> xian


-- 
Romain Francois
Professional R Enthusiast
+33(0) 6 28 91 30 30
http://romainfrancois.blog.free.fr
|- http://bit.ly/fT2rZM : highlight 0.2-5
|- http://bit.ly/gpCSpH : Evolution of Rcpp code size
`- http://bit.ly/hovakS : RcppGSL initial release