[R] Subset() within function: logical error

[David Winsemius]

On Jun 29, 2015, at 5:03 PM, Rich Shepard wrote:

>  Moving from interactive use of R to scripts and functions and have bumped
> into what I believe is a problem with variable names. Did not see a solution
> in the two R programming books I have or from my Web searches. Inexperience
> with ess-tracebug keeps me from refining my bug tracking.
> 
>  Here's a test data set (cleverly called 'testset.dput'):
> 
> structure(list(stream = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L), .Label = c("B", "J", "S"), class = "factor"),
>    sampdate = structure(c(8121, 8121, 8121, 8155, 8155, 8155,
>    8185, 8185, 8185, 8205, 8205, 8205, 8236, 8236, 8236, 8257,
>    8257, 8257, 8308, 8785, 8785, 8785, 8785, 8785, 8785, 8785,
>    8847, 8847, 8847, 8847, 8847, 8847, 8847, 8875, 8875, 8875,
>    8875, 8875, 8875, 8875, 8121, 8121, 8121, 8155, 8155, 8155,
>    8185, 8185, 8185, 8205, 8205, 8205, 8236, 8236, 8236, 8257,
>    8257, 8257, 8301, 8301, 8301), class = "Date"), param = structure(c(2L,
>    6L, 7L, 2L, 6L, 7L, 2L, 6L, 7L, 2L, 6L, 7L, 2L, 6L, 7L, 2L,
>    6L, 7L, 2L, 1L, 2L, 3L, 4L, 5L, 6L, 7L, 1L, 2L, 3L, 4L, 5L,
>    6L, 7L, 1L, 2L, 3L, 4L, 5L, 6L, 7L, 2L, 6L, 7L, 2L, 6L, 7L,
>    2L, 6L, 7L, 2L, 6L, 7L, 2L, 6L, 7L, 2L, 6L, 7L, 2L, 6L, 7L
>    ), .Label = c("Ca", "Cl", "K", "Mg", "Na", "SO4", "pH"), class = "factor"),
>    quant = c(4, 33, 8.43, 4, 32, 8.46, 4, 31, 8.43, 6, 33, 8.32,
>    5, 33, 8.5, 5, 32, 8.5, 5, 59.9, 3.46, 1.48, 29, 7.54, 64.6,
>    7.36, 46, 2.95, 1.34, 21.8, 5.76, 48.8, 7.72, 74.2, 5.36,
>    2.33, 38.4, 8.27, 141, 7.8, 3, 76, 6.64, 4, 74, 7.46, 2,
>    82, 7.58, 5, 106, 7.91, 3, 56, 7.83, 3, 51, 7.6, 6, 149,
>    7.73)), .Names = c("stream", "sampdate", "param", "quant"
> ), row.names = c(NA, -61L), class = "data.frame")
> 
>  I want to subset that data.frame on each of the stream names: B, J, and S.
> This is the function that has the naming error (eda.R):
> 
> extstream = function(alldf) {
>    sname = alldf$stream
>    sdate = alldf$sampdate
>    comp = alldf$param
>    value = alldf$quant
>    for (i in sname) {
>        sname <- subset(alldf, alldf$stream, select = c(sdate, comp, value))


Never use the form dfrm$colname as the argument to the subset argument of subset. You can see that 'stream' is a factor, right? Perhaps 

Furthermore, by inspection you can see that there is no colname =='sdate', so I would guess that would be your next error. Or 'comp' or 'value' for that matter. Oh now I see, you made them outside of `alldf`. Then how is that supposed to work. The subset function is supposed to be looking inside `alldf` to find those column names.


Perhaps:

    subset(alldf, stream %in% c('B', 'J', 'S'),  .... 

....   but have not figured out why you used 'subset' if you wanted: select = c(sdate, comp, value))


Furthermore, it is generally error prone to use `subset` inside functions. The help page warns against the practice. Better to use "[".

>        return(sname)
>    }
> }
> 
>  This is the result of running source('eda.R') followed by
> 
>> extstream(testset)
> Error in subset.data.frame(alldf, alldf$stream, select = c(sdate, comp,  :
>  'subset' must be logical
> 
>  I've tried using sname for the rows to select, but that produces a
> different error of trying to select undefined columns.

Right. Those are not column names in any dataframe.
> 
>  A pointer to the correct syntax for subset() is needed.

No. A pointer to the correct use of "[" is needed.

-- 

David Winsemius
Alameda, CA, USA