Closes Rdatatable#1257, on=.() syntax is now possible.

R/data.table.R

@@ -597,11 +597,22 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) {
                 stop("When i is a data.table (or character vector), the columns to join by must be specified either using 'on=' argument (see ?data.table) or by keying x (i.e. sorted, and, marked as sorted, see ?setkey). Keyed joins might have further speed benefits on very large data due to x being sorted in RAM.")
             }
             if (!missing(on)) {
-                if (!is.character(on))
-                    stop("'on' argument should be a named atomic vector of column names indicating which columns in 'i' should be joined with which columns in 'x'.")
-                parse_on <- function(on) {
+                # on = .() is now possible, #1257
+                parse_on <- function(onsub) {
                     ops = c("==", "<=", "<", ">=", ">", "!=")
                     pat = paste("(", ops, ")", sep = "", collapse = "|")
+                    if (is.call(onsub) && onsub[[1L]] == "eval") {
+                        onsub = eval(onsub[[2L]], parent.frame(2L), parent.frame(2L))
+                        if (is.call(onsub) && onsub[[1L]] == "eval") onsub = onsub[[2L]]
+                    }
+                    if (is.call(onsub) && as.character(onsub[[1L]]) %in% c("list", ".")) {
+                        spat = paste("[ ]+(", pat, ")[ ]+", sep="")
+                        onsub = lapply(as.list(onsub)[-1L], function(x) gsub(spat, "\\1", deparse(x, width.cutoff=500L)))
+                        onsub = as.call(c(quote(c), onsub))
+                    }
+                    on = eval(onsub, parent.frame(2L), parent.frame(2L))
+                    if (!is.character(on))
+                        stop("'on' argument should be a named atomic vector of column names indicating which columns in 'i' should be joined with which columns in 'x'.")
                     this_op = regmatches(on, gregexpr(pat, on))
                     idx = (vapply(this_op, length, 0L) == 0L)
                     this_op[idx] = "=="
@@ -620,7 +631,7 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) {
                         names(on)[empty_idx] = on[empty_idx]
                     list(on = on, ops = idx_op)
                 }
-                on_ops = parse_on(on)
+                on_ops = parse_on(substitute(on))
                 on = on_ops[[1L]]
                 ops = on_ops[[2L]]
                 # TODO: collect all '==' ops first to speeden up Cnestedid

README.md

@@ -78,6 +78,8 @@
 
   30. x's columns can be referred to in `j` using the prefix `x.` at all times. This is particularly useful when it is necessary to x's column that is *also a join column*. This is a patch addressing [#1615](https://github.com/Rdatatable/data.table/issues/1615).
 
+  31. `on=.()` syntax is now posible, e.g., `X[Y, on=.(x==a, y==b)]`, [#1257](https://github.com/Rdatatable/data.table/issues/1257). Thanks @dselivanov.
+
 #### BUG FIXES
 
   1. Now compiles and runs on IBM AIX gcc. Thanks to Vinh Nguyen for investigation and testing, [#1351](https://github.com/Rdatatable/data.table/issues/1351).

inst/tests/tests.Rraw

@@ -8651,6 +8651,15 @@ dt1 = data.table(x=sample(letters[1:2], 10, TRUE), y=sample(c(1L,5L,7L), 10, TRU
 dt2 = data.table(x=c("b", "a"), y=c(1L, 9L))
 test(1654, dt1[dt2, on="x>x"], error="Only '==' operator")
 
+# on= with .() syntax, #1257
+dt1 = data.table(x=sample(letters[1:2], 10, TRUE), y=sample(c(1L,5L,7L), 10, TRUE), z=1:10, k=11:20)
+dt2 = data.table(x=c("b", "a"), y=c(1L, 9L))
+test(1655.1, dt1[dt2, on=.(x)], dt1[dt2, on="x"])
+test(1655.2, dt1[dt2, on=.(x==x)], dt1[dt2, on=c("x==x")])
+test(1655.3, dt1[dt2, on=.(x==x)], dt1[dt2, on=c("x"="x")])
+test(1655.4, dt1[dt2, on=.(y>=y)], dt1[dt2, on=c("y>=y")])
+test(1655.4, dt1[dt2, on=.(x==x, y>=y)], dt1[dt2, on=c("x==x", "y>=y")])
+
 ##########################
 
 # TODO: Tests involving GForce functions needs to be run with optimisation level 1 and 2, so that both functions are tested all the time.