How can I get a stack trace back from a snow node after an error occurs? I'm getting errors when I use parSapply that do not occur when I use sapply. Snow is nice enough to give me the error message but it would be much more useful for me to have the kind of stack trace you can get from traceback(). So far I have tried:
options(showWarnCalls = T, showErrorCalls = T)
setDefaultClusterOptions(outfile = "/dev/tty")
and
options(error=traceback)
setDefaultClusterOptions(outfile = "/dev/tty")
without luck. I'm currently just testing with a local cluster ie:
makeSOCKcluster(c("localhost","localhost"))
but I will eventually be using an MPI cluster. Thanks.
Relevant versions:
snow package (version 0.3-3)
R 2.10.1
Mac OS X 10.5.8
Edit: moved version info to bottom of question and added OS X
For me, simple
options(error=traceback)
works, snow just captures the error message with traceback from slave and shows it on the master's output.
Related
I recently updated my MacBook air (2017) to the Mojave OS and proceeded to update my version of R and Rstudio. Unfortunately, ever since the 'plotTangentSpace' command no longer functions appropriately even when attempting to run the example 'plethodon' code from the Geomorph vingette, I always receive the following error in the output (bolded below):
data(plethodon)
Y.gpa<-gpagen(plethodon$land) #GPA-alignment
|====================================================================================| 100%
gp <- interaction(plethodon$species, plethodon$site)
plotTangentSpace(Y.gpa$coords, groups = gp)
Error in plotTangentSpace(Y.gpa$coords, groups = gp) :
unused arguments (Y.gpa$coords, groups = gp)
I have tried uninstalling and reinstalling R for a MacOS many times including versions 4.0.2, 3.6.3, 3.6.1, and 3.2.1. I have also uninstalled and reinstalled XQuartz several times.
If anyone has encountered a similar issue or has any other ideas I can try I would really appreciate the help. Thank you!
If you execute:
?plotTangentSpace
You should see that this is a deprecated function. That means you should not be using it. It says:
Notes for geomorph 3.3.0 and subsequent versions
I suspect you have not updated your copy of the vignette. There are instructions on the help page for an alternate approach. The example in the help page for the suggested function is:
data(plethspecies)
Y.gpa <- gpagen(plethspecies$land) #GPA-alignment
### Traditional PCA
PCA <- gm.prcomp(Y.gpa$coords)
summary(PCA)
plot(PCA, main = "PCA")
For the plethodon data they suggest:
PCA.w.phylo <- gm.prcomp(Y.gpa$coords, phy = plethspecies$phy)
summary(PCA.w.phylo)
plot(PCA.w.phylo, phylo = TRUE, main = "PCA.w.phylo")
So start a new session (to unload the currently loaded geomorph namespace, and execute this at the r session command line:
install.packages("geomorph")
You should be getting version 3.3.1 of the geomorph package.
I think I found the problem! In my case, the error was due to a old version of the package RRPP, which is required by Geomorph. After updating it, Geomorph is working perfectly! Hope this can be useful for you too.
I came across the same error but after updating the RRPP and rgl packages required for geomorph, the gm.prcomp() function worked for me. I hope this helps if you haven't figured it out already.
es function in smooth package has a weird behavior for some short time-series. Here is a case:
require(smooth) # version 2.4.7
#create two time series of length 10
vec1_ts<-ts(c(36,24,51,7,7,77,1,29,19,2),start=c(2016,1),frequency=12)
vec2_ts<-ts(1:10,start=c(2016,1),frequency=12)
#this works:
es(vec1_ts)
#but this doesn't:
es(vec2_ts)
# another non-working example with a shorter length
vec3_ts<-ts(c(23,5,4,12,2,5),start=c(2016,1),frequency = 12)
es(vec3_ts)
I got a "R encountered a fatal error" message, and then forced to terminate the session when running es for the second time-series.
This looks like a bug in the function of that package. Can you, please, submit the bug here: https://github.com/config-i1/smooth/issues
The StackOverflow is not for solving problems of specific packages, it is for the solution of more general problems.
I am working with the package random uniform Forest. I am trying to run the examples provided in the documentation:
data(iris)
XY = iris
p = ncol(XY)
X = XY[,-p]
Y = XY[,p]
iris.ruf = randomUniformForest(Species ~., XY, threads = 1)
But I get this error:
Error in checkForRemoteErrors(lapply(cl, recvResult)) :
7 nodes produced errors; first error: object '.doSnowGlobals' not found
I googled and found that this is because it is trying to use paralell computing and it does not find something it needs. I have never used paralell computing so I did not understand the explanations I have found and I do not know how to fix this problem. I also read "error: object '.doSnowGlobals' not found?".
According to the manual using "threads = 1" deactivated paralell computing, but I get the error anyways.
I have also checked and the packages paralell and doParallel are loaded.
I do not really need paralell computing and I do not know if I am "connected" to other computers, so I am not sure If that would even work. Would there be an easy way do deactivate paralell computing? Or another alternative for making this work?
The cause of the problem was that I was working on my university computer for which I do not have administrator rights. The randomUniformForest package makes use of parallel processing which uses the IP protocol (even when only one thread is used).
I tried the package on my private computer and it worked fine.
I am playing with cluster analysis in R and I would like to plot the cluster:
teste<-data.frame(v1=rnorm(100),v2=rexp(100),v3=runif(100) )
fit.c<- kmeans(teste,2)
aggregate(teste,by=list(fit.c$cluster),FUN=mean)
clusplot(teste, fit.c$cluster, color=TRUE, shade=TRUE,labels=2, lines=0)
But it gives me this error:
Error in clusplot.default(teste, fit.c$cluster, color = TRUE, shade = TRUE, :
4 arguments passed to .Internal(nchar) which requires 3
Does anyone know what is happening?
Thank you
I ran your code and got no problems. I am guessing that there is a version mismatch between your R and the packages you are using. Can you make sure that your cluster package is up to date (and any other packages you are using for that matter), and/or update to the latest version of R?
If that doesn't work, can you add the sessionInfo() information? That is often helpful for diagnosing these sorts of problems.
Currently I am applying one dataset to WGCNA codes for Network construction and Module detection. Here I have to use a function called "pickSoftThreshold" to detect the network topology. When I run that it shows me this error-
> sft = pickSoftThreshold(datExpr, powerVector = powers, verbose = 5)
pickSoftThreshold: will use block size 18641.
pickSoftThreshold: calculating connectivity for given powers...
..working on genes 1 through 18641 of 54675
Error in serialize(data, node$con) : error writing to connection
Any idea how to get rid of that?
Thanks in Advance!!
I myself just started using WGCNA a couple of days ago and am not really familiar with it yet. But the error looks like you are using too many genes (up to 55k): I think you should find a way to filter out some of them, if your computer isn't powerful enough.
(Ideas from http://labs.genetics.ucla.edu/horvath/CoexpressionNetwork/Rpackages/WGCNA/faq.html )