Consider a list with two items, "table" and "data". "table" is a (n x m) data.frame, "data" is a (r x n) matrix. The matrix has as many columns as there are rows in data.frame. I want to select columns from "table" by a logical expression applied to "data" and get a list with the appropriately subsetted items.
More concrete example:
Let L be the list with item 1 having the columns "A", "B", ... being some factors.
To get a list that is subsetted on A=="some A level value", the "usual way" was:
indx = L$table$A == "some A level value"
L = list( table=L$table[indx,], data=L$data[,indx] )
I would like to get the subset simply by calling L[A=="some A level value"]
For this I will have to overwrite the susbetting method, so the list has to have a class attribute, let's call it "datalist". Then I thought I could define
setClass("datalist", representation="list")
"[.datalist" Error: object 'A' not found
This works:
L[,L$table$A=="a"]
Any ideas?
(btw: I notices that something quite similar is possible for data.tables (library "data.table"), but I could not figure out why it works there but not here...)
Thanks in advance!
Great, thanks!
It works.
Funnily, I also tried substitute, and -for testing purpose- did not use the argument in any way in the function and still got the same error message! It must have been due to some confusion of over-written (and not over-written) versions of the "["-function.
Gavin, I found the reason for my initioal problem... and the problem is in facte NOT solved :-(
I load the package "limma". After loading this, your code does NOT work anymore. I am completely puzzled. I have no idea what limma does to prevent this code from working.
> sessionInfo()
R version 2.15.1 (2012-06-22)
Platform: x86_64-pc-mingw32/x64 (64-bit)
locale:
[1] LC_COLLATE=German_Germany.1252 LC_CTYPE=German_Germany.1252
[3] LC_MONETARY=German_Germany.1252 LC_NUMERIC=C
[5] LC_TIME=German_Germany.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
Example code attached.
One more thing:
it is not only "limma" causing problems. These packages also have the same effect:
annotate, betareg, Biobase, colorspace, ggplot2, Matrix
Other packages do no harm. I tested:
beeswarm, boot, car, coin, data.table, lattice, MASS, modeltools, multcomp, mvtnorm, nlme, nnet, rgl, splines, statmod, stats4, survival
I tried to find some pattern, but I still have no idea :-(
Thanks for any help.
I carefully checked it in the source of "limma". There is no "[.datalist", but there are "["-methods defined for some other classes (specific to limma).
I checked other packages with no further dependencies (e.g. lattice). Same problem.
I used a supposedly unique name "jochensdatalist". Same problem.
Ok, still the error "object A not found" is dropped. Poorly I have no Idea how I can find out what is called instead of my method... (you can try with lattice)
Thank you for your help!
Here is what findMethods tries to tell me:
> findMethods("[")
An object of class “listOfMethods”
$ANY
.Primitive("[")
$nonStructure
function (x, i, j, ..., drop = TRUE)
{
value
I cant access the item "table" in the datalist instance with subset().
Things like
subset(L, A=="a")
subset(L, table$A=="a")
subset(L, L$table$A=="a")
(L is a datalist instance, see above) do not work.
If I define a function similar to "[.datalist" but name it differently, it works as expected. But I want to use the subsetting with "[".
I think declaring new classes, overriding methods etc. is overkill for this problem. Here is a one-liner that will do it:
with(L$table,list(table=L$table[A=='a',],data=L$data[,A=='a']))
A little clumsy to type each time, so here it is wrapped in a function:
dtsub
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package