[R-Forge #203] Allow 2 column to specify range in i instead of %between%

[arunsrinivasan]

Submitted by: Matt Dowle; Assigned to: Nobody; R-Forge link

%between% is currently a vector scan.
Instead allow DT[J(id,DT(from,to)),...]
where the i is a data.list, from and to are equi-joined, and the range between them returned. Depends on non vectors being allowed in a data.table.

[arunsrinivasan]

From Lucas' email - here for reference as this is closely related. Refer to the email exchange with Xavier/Lucas for more info:

require(data.table)

set.seed(45L)
DT_USER <- unique(data.table(UID=1:10, UTIME=sample(10, 30, TRUE)))
DT_USER <- DT_USER[order(UID, UTIME)]

DT_TRANSACTION <- data.table(TIME=1:10, AMOUNT=sample(100, 10, TRUE))

# This is what i really want, join user and transaction,
# but faster than doing the full cartesian product and cosuming less memory too
DT_USER[, DUMMY := 1]
DT_TRANSACTION[, DUMMY := 1]

DT_RESULT <- merge(DT_USER, DT_TRANSACTION, by="DUMMY",
                   allow.cartesian=TRUE)
DT_RESULT[, DUMMY := NULL]
DT_RESULT <- DT_RESULT[TIME <= UTIME]

This requires a full join followed by a subset, which of course can and should be avoided! Thanks a lot Lucas.

[arunsrinivasan]

Lucas' Q can now be done with:

DT_TRANSACTION[DT_USER, on="TIME<=UTIME", allow.cartesian=TRUE]
# or
DT_TRANSACTION[DT_USER, on=.(TIME<=UTIME), allow.cartesian=TRUE]