The purpose of these functions is to allow a byte coded function to be converted back into a fully interpreted function as a temporary work around for issues in byte-code interpretation.
unByteCode(fun)assignEdgewise(name, env, value)
unByteCodeAssign(fun)
All three functions return a copy of the modified function or assigned value.
function to be modified
object name
namespace
new function body
Gregory R. Warnes greg@warnes.net
unByteCode
returns a copy of the function that is directly
interpreted from text rather than from byte-code.
assignEdgewise
makes an assignment into a locked environment.
unByteCodeAssign
changes the specified function in its source
environment to be directly interpreted from text rather than from
byte-code.
The latter two functions no longer work out of the box because assignEdgewise
(which unByteCodeAssign
uses) makes use of an unsafe unlockBinding
call, but running assignEdgewise()
will
These functions were inspired as a work-around to R bug https://bugs.r-project.org/show_bug.cgi?id=15215.
data(badDend)
dist2 <- function(x) as.dist(1 - cor(t(x), method = "pearson"))
hclust1 <- function(x) hclust(x, method = "single")
distance <- dist2(badDend)
cluster <- hclust1(distance)
dend <- as.dendrogram(cluster)
if (FALSE) {
## In R 2.3.0 and earlier crashes with a node stack overflow error
plot(dend)
## Error in xy.coords(x, y, recycle = TRUE) : node stack overflow
}
## convert stats:::plotNode from byte-code to interpreted-code
## (no longer available unless assignEdgewise is defined by the user)
## unByteCodeAssign(stats:::plotNode)
## illustrated in https://stackoverflow.com/questions/16559250/error-in-heatmap-2-gplots
# increase recursion limit
options("expressions" = 5e4)
# now the function does not crash
plot(dend)
Run the code above in your browser using DataLab