Convert columns of arbitrary class to the class of matching columns in another data.table

Question

Question:

I\'m working in R. I want the shared columns of 2 data.tables (shared meaning same column name) to have matching classes. I\'m struggling

Answer 1

Based on the discussion in this question, and comments in this answer, I'm thinking I may have had it right, and just landed on an odd exception.

Note that the class doesn't change, but the technicality is that it doesn't matter (for my particular use-case that prompted the question). Below I show my "failed approach", but by following through to the merge, and the classes of the columns in the merged data.table, we can see why the approach works: integers will just get promoted.

s2c <- function (x, type = "list") 
{
    as.call(lapply(c(type, x), as.symbol))
}

# In this case, I can assume all columns of A can be found in B
# I am also able to assume that the desired conversion is possible
B.class <- sapply(B[,eval(s2c(names(A)))], class)
for(col in names(A)){
    set(A, j=col, value=as(A[[col]], B.class[col]))
}

# Below here is new from what I tried in question
AB <- data.table:::merge.data.table(A, B, all=T, by=c("stratum","year"))

sapply(AB, class)
  stratum      year        bt        yr 
"integer" "numeric" "numeric" "numeric" 

Although the problem in the question isn't solved by this answer, I figured I'd post to point out that the failure to convert "integer" to "numeric" might not be a problem in many situations, so this is a straightforward, albeit circumstantial, solution.