percent() with zero value yields NaN

[vsalmendra]

> library(scales)
> percent(0.1)
[1] "10%"
> percent(0)
[1] "NaN%"

> sessionInfo()
R version 3.1.0 (2014-04-10)
Platform: x86_64-pc-linux-gnu (64-bit)

locale:
[1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
[3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
[5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
[7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
[9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] scales_0.2.4

loaded via a namespace (and not attached):
[1] colorspace_1.2-4 munsell_0.4.2    plyr_1.8.1       Rcpp_0.11.2    

[BrianDiggs]

The problem ultimately arises because plyr:::round_any(0, 0) return NaN. (See https://github.com/hadley/plyr/blob/master/R/round-any.r#L28). I'm not sure what round_any should return when accuracy == 0; I'm thinking that it should just return x unaltered. Or a vector of 0 the same length as x.

# not real code because we are talking about a function in another package.
plyr:::round_any.numeric
function (x, accuracy, f = round) 
{
    if(accuracy == 0) {
        x
    } else {
        f(x/accuracy) * accuracy
    }
}

[dougmitarotonda]

I think that is a good solution. percent similarly does not work for negative numbers and I submitted a pull request for that here: #40. It would be great to see both these changes made so that percent works on all real numbers.

[hadley]

@BrianDiggs I suspect there's a better rounding algorithm that will also work with small x. Hopefully twitter can help: http://twitter.com/hadleywickham/status/608679186597695488

[mvpp]

Any update on this?
The weird part is that it returns "0%" properly when passed in a vector with at least one no-zero value:

percent(c(0,2))
[1] "0%" "200%"
percent(0)
[1] "NaN%"
percent(c(0,0))
[1] "NaN%" "NaN%"

[Stefan2015-5]

I must say it is embarrassing that this bug happened in the first place but it is more embarrassing that it is not fixed. I nearly had a report going out with "NaN%" in there. Sorry, if this comes across angry, but it makes me sad when you can't rely on these easy things.

my.percent<- function(x) {
if(length(x)==1) if(x==0) return(paste0(0,"%") )
return(percent(x) )
}

[krlmlr]

IMO, the precision() function (which is used only by percent()) should be improved:

> precision(c(0.1, 100))
[1] 10
> precision(c(-0.1, 100))
[1] 100
> precision(0)
[1] 0

precision(0) should return 1 (this would fix the issue here), but for the other two I'm not sure.

[Braonan]

Could use mvpp's observation for a fix by creating a vector with dummy value and using only first numeric e.g.,
percent(c(0,1))[1]
bit of a dirty fix; but simple and works