On Fri, 9 Oct 2015, oleg_komashko(a)ukr.net wrote: Maybe not very likely, but for example ! rm -rf / If a gretl user has root privileges that would trash the hard drive. Or ! rm -rf ~/ even without root privileges would totally destroy the user's filespace. The shell is powerful! To date we have no evidence of malicious gretl scripts being distributed, but we'd rather not wait for a disaster to occur before imposing some default limitations. Allin

On Sat, 10 Oct 2015, oleg_komashko(a)ukr.net wrote: [...] Thanks for the positive comments on gretl that I have ellipsized. (But you're changing the thread topic once again!) I can agree with that. Anything gretl can do on Windows it should be able to do on Linux -- and also vice versa, since it seems that (like it or not) most gretl users are still on Windows. However, it seems to me that you just "got lucky" in being able to produce plots via R, running under gretl's "foreign" apparatus on Windows XP. I wouldn't expect this to work in general on any platform -- it's seriously stretching what "foreign" can do, and I would again suggest that you try a different approach, please! (I'd also be interested to know if after this plotting episode on XP you end up with zombie background processes running, as I did after trying your hansl script on Fedora Linux: it "worked" insofar as it put an R plot on screen, but it messed up the process table.) Allin Cottrell

It was designed for model-like classes of objects I didn't expected it would work with "forecast" class objects It was unintending pleasant side effect RforecastL("rwf(LRM,h=4)",X,1,1) # for objects of class "forecast" horizon is ignored # and should be given inside "mod" argument # gives 4 steps 10 жовтня 2015, 04:05:11, від "Allin Cottrell" <cottrell(a)wfu.edu&gt;:

Dear Riccardo, May be, I begin to understand your point I'd like to add several items I never intended to upload the package contained the shell command I even feared to upload it to e-disk I use for my course materials before Allin made the things clear for me: I will simply ban using shell command by my students I hope I was correct all the time during the discussion: I uploaded nothing containing even a hint to use third part soft to the staging area Also I hope I didn't waste anybody's time, since the discussion helped to clear Gretl policy And the last stroke. Below is a short code. How to tell, whether it uses third part soft, or it is a simple wrapper to Gretl foreign block? I fear it will ends in explicit statement in instruction for package writers not to use foreign command in packages. Of course I don't want to, but.... Oleh The code: function void Rmodel (string mod "quoted R function", list X "list of data series", string packages[null] "R packages to use, separated by commas", bool smr[0] "whether to output summary(model), default = no") if isnull(packages) packages = "" else packages = "library("~packages~")" packages = strsub(packages,",",");library(") endif sprintf h "%d", smr string path = $dotdir file1 = path~"/"~"coeff6789.mat" catch r1 = remove(file1) foreign language=R --send-data=X if (any(!class(gretldata)=="data.frame")) {mts2frame2<-function(x) {ddd<-data.frame(x); if (length(class(x))>1) {m<-ncol(x);for (i in 1:m) ddd[,i]<-ts(ddd[,i],start=start(x),frequency=frequency(x))} else ddd[,1]<-ts(ddd[,1],start=start(x),frequency=frequency(x)); ddd}; gretldata<-mts2frame2(gretldata)} attach(gretldata) @packages m1<-@mod nm1<-names(m1) ifcoef<-any(grepl("coeff",nm1,fixed=T)) if (ifcoef) { coeff6789<-as.matrix(as.numeric(m1$coeff)); co<-capture.output(gretl.export(coeff6789))}; h<-@h; { if (h==1) summary(m1) else m1 } end foreign catch coef = mread("coeff6789.mat",1) err = $error if err=0 print coef file1 = path~"/"~"routput6789.mat" r1 = remove(file1) endif end function Examples open beer.gdt ind = (i>30000) list X = q pb ind Rmodel("ets(q)",X,"forecast") Rmodel("lm(q~pb)",X,null,1) Rmodel("glm(q~pb)",X,null,1) Rmodel("auto.arima(q,xreg=pb)",X,"forecast",1) Rmodel("arfima(q)",X,"forecast") Rmodel("plm(q~pb,data=gretldata,model='within',index='ind')",X,"plm") 10 жовтня 2015, 10:53:35, від "Riccardo (Jack) Lucchetti" <r.lucchetti(a)univpm.it&gt;:

Dear Sven, A good idea, Notice, shell will not work without setting it on in via the GUI And not all remove() should be excluded: I think removing temporary files at $dotdir created while running a function should be admissible, e.g. temporary instructions files for gnuplot Oleh 10 жовтня 2015, 14:14:35, від "Sven Schreiber" <svetosch(a)gmx.net&gt;:

Dear Sven, May be, something like this: $dotdir will contain temp subdir after installing Gretl and writing and deleting files from packages would be only in this dir? 10 жовтня 2015, 15:26:26, від "Sven Schreiber" <svetosch(a)gmx.net&gt;: