issue with melt.data.table and patterns #1739
Description
Hi,
I stumbled upon this behavior of melt.data.table:
tmp=data.frame(val_a = sample(x=10,replace=F),val_b = sample(x=10,replace=F),val_c = sample(x=10,replace=F))
tmp$rank_b=rank(tmp$val_b)
tmp$rank_a=rank(tmp$val_a)
tmp$rank_c=rank(tmp$val_c)
str(tmp)
tmpT = melt(data.table(tmp),measure.vars=patterns("val_","rank_"),variable.factor=T,value.name=c("myValue","rank"))
str(tmpT)
ggplot(tmpT,aes(x=rank,y=myValue)) + geom_line(size=2) + facet_wrap(~variable) +theme_bw()
whereas I was expecting this plot:
tmp=data.frame(val_a = sample(x=10,replace=F),val_b = sample(x=10,replace=F),val_c = sample(x=10,replace=F))
tmp$rank_a=rank(tmp$val_a)
tmp$rank_b=rank(tmp$val_b)
tmp$rank_c=rank(tmp$val_c)
str(tmp)
tmpT = melt(data.table(tmp),measure.vars=patterns("val_","rank_"),variable.factor=T,value.name=c("myValue","rank"))
str(tmpT)
ggplot(tmpT,aes(x=rank,y=myValue)) + geom_line(size=2) + facet_wrap(~variable) +theme_bw()
the reason is the different order of the first two rank columns (rank_b/rank_a/rank_c vs. rank_a/rank_b/rank_c).
melt probably runs over the different patterns and puts the result together and hence the order of the rank column does not anymore match the order of the value column.
I was especially confused why this happens due to the new factor "variable" that seemed correct as it had 3 levels.
Maybe this simply need a simple mention in the documentation?
Would it be possible to get the factor levels correct and "joined" in the correct way?
Thanks for data.table!
steffen