Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

I posted this as a response to a question on R-help. I think the idea of a “collect” function could be useful both in the context of unreliable functions that sometimes error out and also in filtering contexts where currently one creates a list containing good elements and some sort of sentinel, usually NULL, which has itself to be filtered out in a separate subsetting operation after the main filtering loop.

Here’s an example:

    d <- runif(20, min=-2, max=8) # test data

aFunc <- function(x) {  # gives error occasionally
if (x > 0)
x
else
}

collect <- function(x, FUN, skip_error=TRUE, args_list=NULL)
{
if (!is.vector(x))
stop("arg x must be a vector")
fname <- deparse(substitute(FUN))
xvar <- deparse(substitute(x))
i <- 1
j <- 1
result <- vector(mode=mode(x), length=length(x))
while (i <= length(x)) {
tryCatch({
args <- list(x[i])
if (length(args_list))
args <- c(args, args_list)
ans <- do.call(FUN, args)
result[j] <- ans
j <- j + 1
}, error=function(e) {
if (!skip_error) {
msg <- paste("collect\n",
"call to", fname, "failed at",
paste(xvar, "[",  i, "]\n", sep=""),
"Message:\n", conditionMessage(e))
stop(msg, call.=FALSE)
}
NULL
},
finally={i <- i + 1})
}
if (j > 1)
result[1:(j-1)]
else
vector(mode=mode(x), length=0)
}

## Example

collect(d, aFunc, skip_error=FALSE)
Error: collect
call to aFunc failed at d[2]
Message:

collect(d, aFunc, skip_error=TRUE)
[1] 7.7380303 0.7554328 1.8352623 0.5136118 4.4231091 2.5368103 1.8656615
[8] 2.9244200 2.1364120 7.6711189 0.2141325 7.8216620 5.8347576 5.3939892