Dear all, I observed unexplainable things with rhdf5. I do not know if they could be avoided. I am trying to access datasets in a group using the function 'do.call' with "$" (since the name of the dataset will only be known at run-time)... and I have a hard time with that (example below). Does anybody have experience with that ? ## example from the help file > mad <- hdf5.open("microarray.h5") > # Create a new group inside the file > hdf5.group(mad,"chip1") > hdf5.dataset(mad$chip1,"raw", dim=c(534,534)) > > raw <- mad$chip1$raw > for(i in 1:534) + { raw[i,1:534] <- runif(534) } > > do.call("$", list("mad", "chip1")) NULL # uh, uh.... > mad$chip1 Name Type Date Comment [1] raw Dataset 21/07/2002 16:26 # let's try again... > do.call("$", list("mad", "chip1")) NULL > mad Name Type Date Comment [1] chip1 Group # show stopper > mad[[1]] Segmentation fault Thanks, Laurent -- -------------------------------------------------------------- Laurent Gautier CBS, Building 208, DTU PhD. Student DK-2800 Lyngby,Denmark tel: +45 45 25 24 85 http://www.cbs.dtu.dk/laurent

You should be able to use the standard "[[" semantics when accessing groups or datasets under rhdf5. this elimnates the need for the "do.call" Byron Ellis (bellis@hsph.harvard.edu) "Oook" - The Librarian Please finger bellis@hsph.harvard.edu for PGP keys On Sun, 21 Jul 2002, Laurent Gautier wrote: > Dear all, > > I observed unexplainable things with rhdf5. I do not > know if they could be avoided. > > I am trying to access datasets in a group using the > function 'do.call' with "$" (since the name of the > dataset will only be known at run-time)... and I have > a hard time with that (example below). Does anybody > have experience with that ? > > > ## example from the help file > > mad <- hdf5.open("microarray.h5") > > # Create a new group inside the file > > hdf5.group(mad,"chip1") > > hdf5.dataset(mad$chip1,"raw", dim=c(534,534)) > > > > raw <- mad$chip1$raw > > for(i in 1:534) > + { raw[i,1:534] <- runif(534) } > > > > > do.call("$", list("mad", "chip1")) > NULL > > # uh, uh.... > > > > mad$chip1 > Name Type Date Comment > [1] raw Dataset 21/07/2002 16:26 > > # let's try again... > > > do.call("$", list("mad", "chip1")) > NULL > > mad > Name Type Date Comment > [1] chip1 Group > > # show stopper > > > mad[[1]] > Segmentation fault > > > > > > Thanks, > > > > > Laurent > > > > > > > -- > -------------------------------------------------------------- > Laurent Gautier CBS, Building 208, DTU > PhD. Student DK-2800 Lyngby,Denmark > tel: +45 45 25 24 85 http://www.cbs.dtu.dk/laurent > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor >

Whoops, didn't scroll down that last bit...did you try mad[["chip1"]]? Byron Ellis (bellis@hsph.harvard.edu) "Oook" - The Librarian Please finger bellis@hsph.harvard.edu for PGP keys On Sun, 21 Jul 2002, Laurent Gautier wrote: > Dear all, > > I observed unexplainable things with rhdf5. I do not > know if they could be avoided. > > I am trying to access datasets in a group using the > function 'do.call' with "$" (since the name of the > dataset will only be known at run-time)... and I have > a hard time with that (example below). Does anybody > have experience with that ? > > > ## example from the help file > > mad <- hdf5.open("microarray.h5") > > # Create a new group inside the file > > hdf5.group(mad,"chip1") > > hdf5.dataset(mad$chip1,"raw", dim=c(534,534)) > > > > raw <- mad$chip1$raw > > for(i in 1:534) > + { raw[i,1:534] <- runif(534) } > > > > > do.call("$", list("mad", "chip1")) > NULL > > # uh, uh.... > > > > mad$chip1 > Name Type Date Comment > [1] raw Dataset 21/07/2002 16:26 > > # let's try again... > > > do.call("$", list("mad", "chip1")) > NULL > > mad > Name Type Date Comment > [1] chip1 Group > > # show stopper > > > mad[[1]] > Segmentation fault > > > > > > Thanks, > > > > > Laurent > > > > > > > -- > -------------------------------------------------------------- > Laurent Gautier CBS, Building 208, DTU > PhD. Student DK-2800 Lyngby,Denmark > tel: +45 45 25 24 85 http://www.cbs.dtu.dk/laurent > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor >

> > > > do.call("$", list("mad", "chip1")) > NULL here you want to refer to mad, not its name, do not use "mad" do.call("$", list(mad, "chip1")) should work also i have noted that there is a problem in some of the doc -- specifically the man page for hdf5.open -- that refers to hdf5.create.group. the latter has been supplanted, it seems, by hdf5.group. i did reproduce your segfault, that needs to be looked at. unfortunately the weather here is very nice....

On Sun, Jul 21, 2002 at 10:17:28AM -0400, Byron Ellis wrote: > The problem, I believe, is that HDF5 makes no guarrantee about the > ordering of the groups/datasets at any particular time (i.e. mad[[1]] > today is not neccessarily mad[[1]] tomorrow). At the very least, though, I > should go through and fix it so it fails gracefully :-) while you are into the segfault side of things, you may also want to try the following: > hdf5.open(file("whatever")) Segmentation fault (I know, the user is not suppposed to do that, unless he/she read the help file in diagonal ;) (hey, did I ?)... but doing that generates a " > library(rhdf5) Error in firstlib(which.lib.loc, package) : not an HDF5 file Error in library(rhdf5) : .First.lib failed " at the next try) L. > > Byron Ellis (bellis@hsph.harvard.edu) > "Oook" - The Librarian > > Please finger bellis@hsph.harvard.edu for PGP keys > > On Sun, 21 Jul 2002, Vincent Carey 525-2265 wrote: > > > > > > > > > do.call("$", list("mad", "chip1")) > > > NULL > > > > here you want to refer to mad, not its name, do not use "mad" > > > > do.call("$", list(mad, "chip1")) > > > > should work > > > > also i have noted that there is a problem in some of the > > doc -- specifically the man page for > > hdf5.open -- that refers to hdf5.create.group. > > the latter has been supplanted, it seems, by hdf5.group. > > > > i did reproduce your segfault, that needs to be looked at. > > unfortunately the weather here is very nice.... > > > > _______________________________________________ > > Bioconductor mailing list > > Bioconductor@stat.math.ethz.ch > > http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor > > > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor -- -------------------------------------------------------------- Laurent Gautier CBS, Building 208, DTU PhD. Student DK-2800 Lyngby,Denmark tel: +45 45 25 24 85 http://www.cbs.dtu.dk/laurent

On Sun, Jul 21, 2002 at 10:12:33AM -0400, Vincent Carey 525-2265 wrote: > > > > > > do.call("$", list("mad", "chip1")) > > NULL > > here you want to refer to mad, not its name, do not use "mad" > > do.call("$", list(mad, "chip1")) > > should work ... hohooo... it looks like it should be outside too (since I start to have dumb questions...) (unfortunately the weather is not particularly nice here) Thanks, L. > > also i have noted that there is a problem in some of the > doc -- specifically the man page for > hdf5.open -- that refers to hdf5.create.group. > the latter has been supplanted, it seems, by hdf5.group. > > i did reproduce your segfault, that needs to be looked at. > unfortunately the weather here is very nice.... > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor -- -------------------------------------------------------------- Laurent Gautier CBS, Building 208, DTU PhD. Student DK-2800 Lyngby,Denmark tel: +45 45 25 24 85 http://www.cbs.dtu.dk/laurent