Closes #604. by=.EACHI is implemented for not joins.

X[ !Y, j, by=.EACHI] now works as intended.

R/data.table.R

@@ -457,8 +457,8 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) {
                             # Otherwise, types of i join columns are alyways promoted to match x's
                             # types (with warning or verbose)
             i = shallow(i)  # careful to only plonk syntax on i from now on (otherwise i would change)
-                            # TO DO: enforce via .internal.shallow attribute and expose shallow() to users
-                            # This is why shallow() is very importantly internal only, currently.
+                             # TO DO: enforce via .internal.shallow attribute and expose shallow() to users
+                             # This is why shallow() is very importantly internal only, currently.
             resetifactor = NULL  # Keep track of any factor to factor join cols (only time we keep orig)
             for (a in seq_along(leftcols)) {
                 # This loop is simply to support joining factor columns
@@ -510,6 +510,16 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) {
                     set(i,j=lc,value=newval)
                 }
             }
+            # Implementation for not-join along with by=.EACHI, #604
+            if (notjoin && byjoin) {
+                notjoin = FALSE
+                if (verbose) {last.started.at=proc.time()[3];cat("not-join called with 'by=.EACHI'; Replacing !i with i=setdiff(x,i) ...");flush.console()}
+                i = setdiff_(x, i, rightcols, leftcols) # part of #547
+                if (verbose) {cat("done in",round(proc.time()[3]-last.started.at,3),"secs\n");flush.console}
+                setnames(i, names(origi)[leftcols])
+                setattr(i, 'sorted', names(i)) # since 'x' has key set, this'll always be sorted
+                origi = i
+            }
             f__ = integer(nrow(i))   # these could be returned as a list from bmerge?
             len__ = integer(nrow(i))
             allLen1 = logical(1)

README.md

@@ -154,15 +154,20 @@ We moved from R-Forge to GitHub on 9 June 2014, including history.
   26. `.N` is now available in `i`, [FR#724](https://github.com/Rdatatable/data.table/issues/724). Thanks to newbie indirectly [here](http://stackoverflow.com/a/24649115/403310) and Farrel directly [here](http://stackoverflow.com/questions/24685421/how-do-you-extract-a-few-random-rows-from-a-data-table-on-the-fly).
 
   27. `.()` can now be used in `j` and is identical to `list()`, for consistency with `i`.
-  ```R
-      DT[,list(MySum=sum(B)),by=...]
-      DT[,.(MySum=sum(B)),by=...]     # same
-      DT[,list(colB,colC,colD)]
-      DT[,.(colB,colC,colD)]          # same
-  ```
+    ```R
+    DT[,list(MySum=sum(B)),by=...]
+    DT[,.(MySum=sum(B)),by=...]     # same
+    DT[,list(colB,colC,colD)]
+    DT[,.(colB,colC,colD)]          # same
+    ```
   Similarly, `by=.()` is now a shortcut for `by=list()`, for consistency with `i` and `j`.
 
-
+  28. `by=.EACHI` is now implemented for *not-joins* as well. Closes [#604](https://github.com/Rdatatable/data.table/issues/604). Thanks to Garrett See for filing the FR. As an example:
+    ```R
+    DT = data.table(x=c(1,1,1,1,2,2,3,4,4,4), y=1:10, key="x")
+    DT[!J(c(1,4)), sum(y), by=.EACHI] # is equivalent to DT[J(c(2,3)), sum(y), by=.EACHI]
+    ```
+
 #### BUG FIXES
 
   1. `fread()`:

inst/tests/tests.Rraw

@@ -5001,6 +5001,19 @@ test(1364.15, setdiff_(X, Y, "c", "a"), error="When x's column ('c') is characte
 test(1364.16, setdiff_(X, Y), error="setdiff(x,y) requires same number of columns for both x and y")
 test(1364.17, setdiff_(X[, list(a)], Y[, list(a)]), data.table(a=c(1,2)))
 
+# not join along with by=.EACHI, #604
+DT <- data.table(A=c(1,1,1,2,2,2,2,3,3,4,5,5))[, `:=`(B=as.integer(A), C=c("c", "e", "a", "d"), D=factor(c("c", "e", "a", "d")), E=1:12)]
+setkey(DT, A)
+test(1365.1, DT[!J(c(2,5)), sum(E), by=.EACHI], DT[J(c(1,3,4)), sum(E), by=.EACHI])
+setkey(DT, B)
+test(1365.2, DT[!J(c(4:5)), list(.N, sum(E)), by=.EACHI], DT[J(1:3), list(.N, sum(E)), by=.EACHI])
+setkey(DT, C)
+test(1365.3, copy(DT)[!"c", f := .N, by=.EACHI], copy(DT)[c("a", "d", "e"), f := .N, by=.EACHI])
+setkey(DT, D)
+test(1365.4, DT[!J(factor("c")), .N, by=.EACHI], DT[J(factor(c("a", "d", "e"))), .N, by=.EACHI])
+test(1365.5, DT[!"c", lapply(.SD, sum), by=.EACHI, .SDcols=c("B", "E")], DT[c("a", "d", "e"), lapply(.SD, sum), by=.EACHI, .SDcols=c("B", "E")])
+
+
 ##########################