[R] slower execution in R 1.9.0

[Prof Brian Ripley]

Rather a misleading subject, if it was slower under 1.8.1.

On Thu, 22 Apr 2004, Scott Bartell wrote:

> I have an R function (about 1000 lines long) that takes more than 20
> times as long to run under R Windows 1.9.0 and 1.8.1 than it does under
> 1.7.1.  Profile results indicate that the $<-.data.frame operation is
> the culprit, but I don't understand exactly what that is (assignment of
> data frame elements to another variable?), or why it's only a problem
> under 1.8.1 and 1.9.0.  Any advice?   

Don't use $<- on data frames, that is code like

foo$bar <- bah

for foo a data frame, if you don't want to pay the penalty of error
checking.  Versions prior to R 1.8.0 treated data frames as lists for that
construct, and often ended up with invalid data frames.

Sound like you could use a lists for your working data.

> 
> top 5 operations under 1.9.0
>                self.time self.pct total.time total.pct
> $<-.data.frame    110.64     93.3     110.68      93.3
> FMD                 5.38      4.5     118.60     100.0
> DIRinf              0.54      0.5      11.82      10.0
> INDinf              0.46      0.4       9.78       8.2
> pweibull            0.38      0.3       0.38       0.3
>   
> top 5 operations under 1.7.1
>                        self.time self.pct total.time total.pct
> pweibull                    2.62     55.7       2.62      55.7
> ^                           0.76     16.2       0.76      16.2
> CalcDistance                0.16      3.4       1.26      26.8
> FMD                         0.16      3.4       4.70     100.0
> matrix                      0.14      3.0       0.14       3.0

-- 
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