Unintuitive output from a rolling join

When performing a rolling join by different column names, it seems like the first column name is selected by default regardless where the values are coming from. This is very confusing to me. Is this by design? (sorry flying tonight in didn't have time to test on the GH version too)

library(data.table) # V 1.9.6+
set.seed(123)
DT <- data.table(X = 1:3)
DT2 <- data.table(Y = 3:5, Z = sample(3))
DT[DT2, on = c(X = "Y"), roll = -Inf]
#    X Z <~~~ The column name is X but values are from Y
#1: 3 1
#2: 4 2
#3: 5 3

This. I couldn't agree more. I constantly work around this issue.

This issue is already covered in the linked one, it is also well explained there.
Current recommendation below. It is unlikely to be changed any time soon, feel free to post your proposal into linked issue.

DT[DT2, .(Y, Z), on = c(X = "Y"), roll = -Inf]
# when column names are overlapping
DT[DT2, .(Y=i.Y, Z=i.Z), on = c(X = "Y"), roll = -Inf]
# now also 'x.' prefix should work the same as 'i.' before giving better control

If I misinterpret this issue please re-open providing better example, which cannot be addressed by x. and i. prefixes when listing columns in j.

Navigation Menu

Repository navigation

Unintuitive output from a rolling join #1469

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Unintuitive output from a rolling join #1469

Description

Activity

ben519 commented on Nov 15, 2016

jangorecki commented on Nov 15, 2016

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions