Ordering bars in a stacked bar plot using ggplot - r

Question

The following is a simplified version of my dataframe (without too much loss in generality)
sales<-data.frame(ItemID=c(1,3,7,9,10,12),
Salesman=c("Bob","Sue","Jane","Bob","Sue","Jane"),
ProfitLoss=c(10.00,9.00,9.50,-7.50,-11.00,-1.00))
which produces
ItemID Salesman ProfitLoss
1 1 Bob 10.0
2 3 Sue 9.0
3 7 Jane 9.5
4 9 Bob -7.5
5 10 Sue -11.0
6 12 Jane -1.0
The following produces a stacked bar plot of each salesman's sales, ordered by the overall profit for each salesman.
sales$Salesman<-reorder(sales$Salesman,-sales$ProfitLoss,FUN="sum") #to order the bars
profits<-sales[which(sales$ProfitLoss>0),]
losses<-sales[which(sales$ProfitLoss<0),]
ggplot()+
geom_bar(data=losses,aes(x=Salesman, y=ProfitLoss),stat="identity", color="white")+
geom_bar(data=profits,aes(x=Salesman, y=ProfitLoss),stat="identity", color="white")
This works exactly as I desire. My issue arises when one of the salesmen has a profit but no loss, or a loss but no profit. For instance, changing sales to
sales<-data.frame(ItemID=c(1,3,7,9,10),
Salesman=c("Bob","Sue","Jane","Bob","Sue"),
ProfitLoss=c(10.00,9.00,9.50,-7.50,-11.00))
and reapplying the previous steps produces
So, the salesman are clearly out of order. For this example I can cheat and plot my profits before losses like
ggplot()+
geom_bar(data=profits,aes(x=Salesman, y=ProfitLoss),stat="identity", color="white")+
geom_bar(data=losses,aes(x=Salesman, y=ProfitLoss),stat="identity", color="white")
but that won't work for my real dataset.
Edit: In my real dataset, each salesman has more than two sales, and for each salesman I've stacked the bars so that the smallest bars in magnitude are closest to the x axis and the largest bars (i.e. biggest profit, biggest loss) are farthest from the x axis. For this reason, I need to call geom_bar() on both the profits dataframe and the losses dataframe. (I originally left this information out to try to avoid making my question too complex.)

Answer 1

The problem is the first plot call to geom_bar(losses dataset) only has two levels of salesman, hence the order is changed - that's why calling profits first still works (as there are still all levels). But your reordering works if you change the plot call
sales<-data.frame(ItemID=c(1,3,7,9,10),
Salesman=c("Bob","Sue","Jane","Bob","Sue"),
ProfitLoss=c(10.00,9.00,9.50,-7.50,-11.00))
#to order the bars
sales$Salesman<-reorder(sales$Salesman,-sales$ProfitLoss,FUN="sum")
# Changed plot call
ggplot(sales, aes(x = factor(Salesman), y = ProfitLoss)) +
geom_bar(stat = "identity",position="dodge",color="white")
-------------------------------------------------------------------------------
Following edit; Do you want the longest bars [ie the largest (profit + abs(losses))] furthest from the y-axis, rather than by descending revenue. You can do this by changing the reorder function. Apologies if i misunderstand.
I changed Jane's data so that it is the longest overall bar
sales<-data.frame(ItemID=c(1,3,7,9,10),
Salesmn=c("Bob","Sue","Jane","Bob","Sue"),
ProfitLoss=c(10.00,9.00,29.50,-7.50,-11.00))
sales$Salesman<-reorder(sales$Salesman,-sales$ProfitLoss,function(z) sum(abs(z)))
ggplot(sales, aes(x = factor(Salesman), y = ProfitLoss)) +
geom_bar(stat = "identity",position="dodge",color="white")