R lazy evaluation ifelse logical condition

Question

I have the following data.table:

dtable <- data.table(column1 = c(1, 2, 3, 5, 6, 7, 8),
                     column2 = c(1, 1, 1, 5, 5, 6, 8),
                     column3 = c(7, 8, 9, 0, 9, 2, 3))

I would like to make something like the following functionality, but in a parametrized function:

dtable %>% 
  dplyr::group_by(column1) %>% 
  dplyr::summarise(Result = ifelse(column1 == column2, "A", "B"))

To do this, I've created the following function:

Test <- function(.df, .columnName, .columnToGroup){
  res <- .df %>%
    # This line is interpreted correctly
    dplyr::group_by_(lazyeval::interp(.columnToGroup, .columnToGroup = as.name(.columnToGroup))) %>%

    # This line does not interpret the == condition as a logical one
    dplyr::summarise_(Result = ifelse((lazyeval::interp(.columnToGroup == .columnName,
                                                        .columnToGroup = as.name(.columnToGroup),
                                                        .columnName = as.name(.columnName))),
                                      "A", "B"))
  return(res)
}

I'm using the Non-Standard Evaluation functions (group_by_ and summarise_) in combination with the lazyeval::interp function, but the == condition is not interpreted the right way, and I get the following exeception:

Test(dtable, "column1", "column2")

 Error in UseMethod("interp") : 
  no applicable method for 'interp' applied to an object of class "logical"

I've tried many different combinations (quote, expr_env, as.lazy, etc.) with no luck. Thanks to this great Non-standard evaluation guide I was able to use these lazyeval functions to evaluate arithmetic expressions before, but I can't find the way to make them to interpret a logical condition in this piece of code.

Any help would be greatly appreciated.

Answer 1

With ifelse, we can try (using @docendodiscimus use of list in comments)

Test <- function(.df, .columnName, .columnToGroup){
 .df %>%   
   dplyr::group_by_(.dots = .columnToGroup )%>%
            dplyr::summarise_(.dots =

                            setNames(list(lazyeval::interp(quote(ifelse(colGrp == colName,
                               "A", "B")), .values = list(colGrp = as.name(.columnToGroup),
                                                    colName = as.name(.columnName)))),
                                  "Result"))

 }

res2 <- Test(dtable, "column2", "column1")
identical(res1, res2)
#[1] TRUE

where 'res1' is

res1 <- dtable %>% 
          dplyr::group_by(column1) %>% 
          dplyr::summarise(Result = ifelse(column1 == column2, "A", "B"))

Update

With the new version of dplyr i.e. 0.6.0 (awaiting release on April 2017), we can also unquote within group_by and summarise. The enquo function does similar job as substitute from base R by taking up the input argument can creating a quosure and this gets unquoted (!! or UQ) within group_by and summarise for evaluation

Test1 <- function(df, colN, colGrp){
      colN <- enquo(colN)
      colGrp <- enquo(colGrp)

      df %>% 
         group_by(!!colGrp) %>%
         summarise(Result = if_else((!!colGrp) == (!!colN), "A", "B"))
}

res3 <- Test1(dtable, column2, column1)
identical(res2, res3)
#[1] TRUE