join_keys[i, j] or join_keys[[i]][[j]]

[gogonzo]

follow up to: #78

join_keys is an S3 class extending a "list". I've heard voices concerned about unintuitive usage of [[ when setting and getting join-key-columnames of i and j. Below two equivalent examples showing current and desired look.

jk <- default_cdisc_join_keys

# current
jk[["ADSL"]][["ADRS"]]
jk[["ADSL"]][["ADRS"]] <- c("USUBJID", "STUDYID")

# desirable
jk["ADSL", "ADRS"]
jk["ADSL", "ADRS"] <- c("USUBJID", "STUDYID")

Things to consider:

• jk[["ADTTE"]][["ADRS"]] signals that it is a list so methods like c(...) or modifyList are more natural.
• what about jk$ADTTE$ADRS?
• jk["ADSL", "ADRS"] is visibly easier, even more than a jk$ADSL$ADRS.
• jk["ADTTE", "ADRS", ...] allows for more arguments like infer_parent_keys - if keys keys between "ADTTE" and "ADRS" are not specified, than they are given (optionally) based on the common relationship with "ADSL"
• jk["ADTTE", "ADRS"] returns a vector and as opposed to scalar in case of an array (matrix)
• [.join_keys <- function(i, j) restricts visibly that object is two dimensional (not more). With a list it is not so strict, as one might try jk[[1]][[1]][[1]].