Partitioning a dataset with a for loop - r

Question

I am trying to make a function in R which returns a list of dataframes which are subsetted by each factor level.
An example to help explain what I am trying to do;
#Creating a dataset for my example
f1<-c("a","a","b","b","c","c")
f2<-c("x","y","x","y","x","y")
v1<-c(1:6)
v2<-c(7:12)
factors<-as.data.frame(cbind(f1,f2))
integers<-as.data.frame(cbind(v1,v2))
df<-cbind(factors,integers)
#The function
partition<-function(data){
factors<-Filter(is.factor,data) #Splitting data into factors
subsets<-list(NULL) #Creating an empty list where I
will put the subsets nm=0
for( i in 1:ncol(factors)){
nm=nm+nlevels(factors[,i])
}
nm
for( i in 1:ncol(factors)){
for(j in 1:nlevels(factors[,i])){
for(k in 1:nm){
subsets[[k]]<-df[which(factors[,i]==levels(factors[,i])[j]), ]
}
}
}
return(subsets)
}
partition(df)
This yields:
[[1]]
f1 f2 v1 v2
2 a y 2 8
4 b y 4 10
6 c y 6 12
[[2]]
f1 f2 v1 v2
2 a y 2 8
4 b y 4 10
6 c y 6 12
[[3]]
f1 f2 v1 v2
2 a y 2 8
4 b y 4 10
6 c y 6 12
[[4]]
f1 f2 v1 v2
2 a y 2 8
4 b y 4 10
6 c y 6 12
[[5]]
f1 f2 v1 v2
2 a y 2 8
4 b y 4 10
6 c y 6 12
As you can see, these are all the same datasets. By removing the loop over k, all the datasets are different and subsetting properly however it only gives me three datasets (as there are two levels in the last factor variable we keep the subset where f1 == "c").
Removing the for loop over k we get;
[[1]]
f1 f2 v1 v2
1 a x 1 7
3 b x 3 9
5 c x 5 11
[[2]]
f1 f2 v1 v2
2 a y 2 8
4 b y 4 10
6 c y 6 12
[[3]]
f1 f2 v1 v2
5 c x 5 11
6 c y 6 12
Where we are missing the subsets where f1 == "a" and f1 == "b"
Note I should get 5 dataframes as we have 2 + 3 factor levels (This is calculated as nm in the first for loop before the subsetting.
So my question is, how do I get the above to run without overiding what had already be subsetted?
For some background, this is working towards making a classification model where It would yield nfactor(df) predictions, then I will run a GLM to weight each prediction.
Thank you for any insight into my problem.
Update
The first answer from Glen simplifies my code which may make the problem I am having more apparent. Here is the updated code (note it runs much more effectively on large datasets with the split() function so thank you Glen.
for(k in 1:nm){
for( i in 1:ncol(factors)){
for( j in 1:nlevels(factors[,i])){
subsets[[k]]<-split(df,factors[,i])[j]
}
}
}
Returns the same as my original question . The problem is that when I run a loop over k through nm, the loop over writes what was already generated. How do I stop this from happening?

Answer 1

If I understand your question correctly. You can do this real easily with the split function.
f1<-c("a","a","b","b","c","c")
f2<-c("x","y","x","y","x","y")
v1<-c(1:6)
v2<-c(7:12)
factors<-as.data.frame(cbind(f1,f2))
integers<-as.data.frame(cbind(v1,v2))
df<-cbind(factors,integers)
tmp1=split(df,f1)
tmp2=split(df,f2)
c(tmp1,tmp2)

Answer 2

library(plyr)
library(foreach)
x<-foreach(i= colnames(Filter(is.factor,df)), .combine='c') %do%
plyr::dlply(df, i)
Returns a list of 5 dataframes. c is used to combine each result of the foreach loop (which is a list in itself). Without this we get a list of lists. With c, it combinds all the lists into 1 lists.