warnErrList {utils}

R Documentation

Collect and Summarize Errors From List

Description

Collect errors (class "error", typically from tryCatch) from a list x into a “summary warning”, by default produce a warning and keep that message as "warningMsg" attribute.

Usage

warnErrList(x, warn = TRUE, errValue = NULL)

Arguments

x	a list, typically from applying models to a list of data (sub)sets, e.g., using tryCatch(*, error = identity).
warn	logical indicating if warning() should be called.
errValue	the value with which errors should be replaced.

Value

a list of the same length and names as the x argument, with the error components replaced by errValue, NULL by default, and summarized in the "warningMsg" attribute.

See Also

The warnErrList() utility has been used in lmList() and nlsList() in recommended package nlme forever.

Examples

## Regression for each Chick:
ChWtgrps <- split(ChickWeight, ChickWeight[,"Chick"])
sapply(ChWtgrps, nrow)# typically 12 obs.
nlis1 <- lapply(ChWtgrps, function(DAT) tryCatch(error = identity,
                      lm(weight ~ (Time + I(Time^2)) * Diet, data = DAT)))
nl1 <- warnErrList(nlis1) #-> warning :
## 50 times the same error (as Diet has only one level in each group)
stopifnot(sapply(nl1, is.null)) ## all errors --> all replaced by NULL
nlis2 <- lapply(ChWtgrps, function(DAT) tryCatch(error = identity,
                      lm(weight ~ Time + I(Time^2), data = DAT)))
nl2 <- warnErrList(nlis2)
stopifnot(identical(nl2, nlis2)) # because there was *no* error at all
nlis3 <- lapply(ChWtgrps, function(DAT) tryCatch(error = identity,
                      lm(weight ~ poly(Time, 3), data = DAT)))
nl3 <- warnErrList(nlis3) # 1 error caught:
stopifnot(inherits(nlis3[[1]], "error")
        , identical(nl3[-1], nlis3[-1])
        , is.null(nl3[[1]])
)

## With different error messages
if(requireNamespace("nlme")) { # almost always, as it is recommended
 data(Soybean, package="nlme")
 attr(Soybean, "formula") #-> weight ~ Time | Plot  => split by "Plot":
 L <- lapply(split(Soybean, Soybean[,"Plot"]),
             function(DD) tryCatch(error = identity,
                 nls(weight ~ SSlogis(Time, Asym, xmid, scal), data = DD)))
 Lw <- warnErrList(L)
} # if <nlme>

---