R help - Sep 2024 - lattice xyplot with cumsum() function inside

This code gives unexpected result.

library(data.table)
library(lattice)

set.seed(123)
mydt <- data.table(date = seq.Date(as.IDate("2024-01-01"), by = 1,
length.out = 50), xgroup = "A", x = runif(50, 0, 1))
mydt <- rbindlist(list(mydt, data.table(date = mydt$date, xgroup =
"B", x = runif(50, 0, 3))))
mydt[, `:=`(xcumsum = cumsum(x)), by = .(xgroup)]
mydt[, lapply(.SD, sum), by = .(xgroup), .SDcols = c("x")]
#   xgroup        x
#   <char>    <num>
#1:      A 26.00455
#2:      B 71.55405

#For xgroup = "B", line starts at the sum of all previous x values
including xgroup = "A"
#Intended result is to separate cumsum(x) for groups "A" and
"B"
mydt[, xyplot(cumsum(x) ~ date, groups = xgroup, type = c("l",
"g"))]

mydf <- as.data.frame(mydt)
xyplot(cumsum(x) ~ date, groups = xgroup, type = c("l",
"g"), data mydf)
#Same graph

aggregate(x ~ xgroup, FUN = sum, data = mydf)
#  xgroup        x
#1      A 26.00455
#2      B 71.55405

#In graph, group "A" goes up to 26.  But group "B" goes up
to 26 + 71.55

#I can get intended graph in this way
mydt[, xyplot(xcumsum ~ date, groups = xgroup, type = c("l",
"g"),
auto.key = list(columns = 2, space = "bottom"))]

Is this a bug or incorrect use of function?

Thanks,
Naresh

? Sat, 28 Sep 2024 13:34:22 +0000
Naresh Gurbuxani <naresh_gurbuxani at hotmail.com> ?????:
> mydt[, lapply(.SD, sum), by = .(xgroup), .SDcols = c("x")]
> #   xgroup        x
> #   <char>    <num>
> #1:      A 26.00455
> #2:      B 71.55405
> 
> #For xgroup = "B", line starts at the sum of all previous x
values
> including xgroup = "A"
> #Intended result is to separate cumsum(x) for groups "A" and
"B"
> mydt[, xyplot(cumsum(x) ~ date, groups = xgroup, type = c("l",
"g"))]
> Is this a bug or incorrect use of function?
Thank you very much for providing a reproducible example! Lattice first
evaluates the formula and only then splits the results into groups.
This is a consequence of the design where grouping is handled by the
panel functions such as panel.superpose.

I think you can obtain the desired result by letting data.table group
the rows:

xyplot(
 x ~ date, groups = xgroup, type = c("l", "g"),
 data = mydt[,
  c(list(date=date), lapply(.SD, cumsum)),
  by = xgroup, .SDcols = 'x'
 ]
)

A more idiomatic solution might be:

mydt[,xsum := cumsum(x), by = xgroup]
xyplot(xsum ~ date, groups = xgroup, type = c("l", "g"),
data = mydt)

-- 
Best regards,
Ivan