data.table j使用temp变量和by列的操作

Question

可以在两行中获得所需的行为，但是一行失败（参见下面的代码）。怀疑与j中的temp变量有关，或者在fail-1中使用by和j中的列，尽管fail-2也提供了NA。

MWE和输出（首先是所需的）如下所示：

library(data.table)
foostr <- "ab"
foodt <- data.table(v1 = letters[1:3], v2 = 1:3)

# desired output
bardt <- foodt[, .(v3 = sapply(.SD, grepl, x = foostr)), by = c("v1", "v2"), .SDcols = "v1"]
bardt[, v4 := .SD$v2*.SD$v3, by = "v1"]
print(bardt)
# v1 v2    v3 v4
# 1:  a  1  TRUE  1
# 2:  b  2  TRUE  2
# 3:  c  3 FALSE  0

# fail-1: NAs?
bardt <- foodt[, .(v3 = {foov3 = sapply(.SD, grepl, x = foostr)}, v4 = .SD$v2*foov3), by = c("v1", "v2"), .SDcols = "v1"]
print(bardt)
# v1 v2    v3 v4
# 1:  a  1  TRUE NA
# 2:  b  2  TRUE NA
# 3:  c  3 FALSE NA

# fail-2: NAs?
bardt <- foodt[, .(v3 = {foov3 = sapply(.SD, grepl, x = foostr)}, v4 = .SD$v2*foov3), by = c("v1"), .SDcols = "v1"]
print(bardt)
# v1    v3 v4
# 1:  a  TRUE NA
# 2:  b  TRUE NA
# 3:  c FALSE NA