|
warning(_("Coerced j from numeric to integer. Please pass integer for efficiency; e.g., 2L rather than 2")); |
I find this warning a bit annoying. It's very common in R to have double "just work" as integer. Why this of all places is there an efficiency concern?
I can't imagine the efficiency concern is real anyway -- unless we are coercing 1e7+ column numbers to integer I can't imagine there's a real performance implication.
I also see two cases where this might come up:
-
You are passing j as a variable to set():
setDT(DT, , col_idx, val)
In this case, the recommendation is to do as.integer(col_idx), which is what assign.c is doing implicitly. Is there really any difference in efficiency?
-
You are doing interactive work (as I just was when I came across this) and lazily omitting the L:
setDT(DT, , c(1, 4), NULL)
OK, I could write 1L/4L, and in package/production code, I probably should. But to throw a warning seems a bit heavy-handed. Users really under such use cases should be using lintr::implicit_integer_linter() to catch such cases.
Originally added a long time ago: e2cf7e2
Reactions are currently unavailable
data.table/src/assign.c
Line 428 in afa87e1
I find this warning a bit annoying. It's very common in R to have double "just work" as integer. Why this of all places is there an efficiency concern?
I can't imagine the efficiency concern is real anyway -- unless we are coercing 1e7+ column numbers to integer I can't imagine there's a real performance implication.
I also see two cases where this might come up:
You are passing j as a variable to set():
In this case, the recommendation is to do as.integer(col_idx), which is what assign.c is doing implicitly. Is there really any difference in efficiency?
You are doing interactive work (as I just was when I came across this) and lazily omitting the L:
OK, I could write 1L/4L, and in package/production code, I probably should. But to throw a warning seems a bit heavy-handed. Users really under such use cases should be using lintr::implicit_integer_linter() to catch such cases.
Originally added a long time ago: e2cf7e2