I have data similar to these:
data <- data.frame(x1=rexp(10,2),y1=rnorm(10,1),x2=rexp(10,2),y2=rnorm(10,1),x3=rexp(10,2),y3=rnorm(10,1),x4=rexp(10,2),y4=rnorm(10,1)) I would like to order all y1,y2... variables in increasing order.
I have tried this among others:
data[with(data,order(y1,y2,y3,y4,decreasing=FALSE)), ] There is number of SO contribution for example here about order, yet I cannot make it work.
You could try
indx <- do.call(order, c(data[grep('^y', names(data))], decreasing=TRUE)) data[indx,] By default, the option is decreasing=FALSE. If that is what you wanted, just do
indx <- do.call(order, data[grep('^y', names(data))]) With data.table package you can either sort by reference and pass several variable names at once to the setorder function (this function isn't mentioned in the above link)
library(data.table) setorderv(setDT(data), grep("^y\\d+", names(data), value = TRUE)) If you want decreasind order you can specify the order, e.g.,
setorderv(setDT(data), grep("^y\\d+", names(data), value = TRUE), order = -1L) The nicest thing about it, is that you can pass a whole vector into order that will sort each column differently, for example
indx <- c(1L, -1L, -1L, 1L) setorderv(setDT(data), grep("^y\\d+", names(data), value = TRUE), order = indx) While everything is done by reference (without creating copies using <-)
decreasing=TRUE here?
data.table answer in the link mentioned by the OP
I think the cited solution does exactly what it is supposed to do - sort first on y1, then (for same y1-values) on y2, and so on.
If you want to have each x-y pairs sorted independently, you have to break up the data.frame.
data.frame(data[with(data,order(y1,decreasing=FALSE)),c('x1', 'y1') ], data[with(data,order(y2,decreasing=FALSE)),c('x2', 'y2') ], data[with(data,order(y3,decreasing=FALSE)),c('x3', 'y3') ], data[with(data,order(y4,decreasing=FALSE)),c('x4', 'y4') ]) 
y1is order not the resty2...xcolumns change too?do.call(order, c(data[grep('^y', names(data))], decreasing=TRUE)