non-joins with data.tables

Question

I have a question on the data.table idiom for "non-joins", inspired from Iterator's question. Here is an example:

library(data.table)

dt1 <- data.table(A1=letters[1:10], B1=sample(1:5,10, replace=TRUE))
dt2 <- data.table(A2=letters[c(1:5, 11:15)], B2=sample(1:5,10, replace=TRUE))

setkey(dt1, A1)
setkey(dt2, A2)

The data.tables look like this

> dt1               > dt2
      A1 B1               A2 B2
 [1,]  a  1          [1,]  a  2
 [2,]  b  4          [2,]  b  5
 [3,]  c  2          [3,]  c  2
 [4,]  d  5          [4,]  d  1
 [5,]  e  1          [5,]  e  1
 [6,]  f  2          [6,]  k  5
 [7,]  g  3          [7,]  l  2
 [8,]  h  3          [8,]  m  4
 [9,]  i  2          [9,]  n  1
[10,]  j  4         [10,]  o  1

To find which rows in dt2 have the same key in dt1, set the which option to TRUE:

> dt1[dt2, which=TRUE]
[1]  1  2  3  4  5 NA NA NA NA NA

Matthew suggested in this answer, that a "non join" idiom

dt1[-dt1[dt2, which=TRUE]]

to subset dt1 to those rows that have indexes that don't appear in dt2. On my machine with data.table v1.7.1 I get an error:

Error in `[.default`(x[[s]], irows): only 0's may be mixed with negative subscripts

Instead, with the option nomatch=0, the "non join" works

> dt1[-dt1[dt2, which=TRUE, nomatch=0]]
     A1 B1
[1,]  f  2
[2,]  g  3
[3,]  h  3
[4,]  i  2
[5,]  j  4

Is this intended behavior?

Answer 1

New in v1.8.3 :

A new "!" prefix on i signals 'not-join' (a.k.a. 'not-where'), #1384.
  DT[-DT["a", which=TRUE, nomatch=0]]   # old not-join idiom, still works
  DT[!"a"]                              # same result, now preferred.
  DT[!J(6),...]                         # !J == not-join
  DT[!2:3,...]                          # ! on all types of i
  DT[colA!=6L | colB!=23L,...]          # multiple vector scanning approach
  DT[!J(6L,23L)]                        # same result, faster binary search
'!' has been used rather than '-' :
  * to match the 'not-join' and 'not-where' nomenclature
  * with '-', DT[-0] would return DT rather than DT[0] and not be backwards
    compatibile. With '!', DT[!0] returns DT both before (since !0 is TRUE in
    base R) and after this new feature.
  * to leave DT[+...] and DT[-...] available for future use

Answer 2

As far as I know, this is a part of base R.

# This works
(1:4)[c(-2,-3)]

# But this gives you the same error you described above
(1:4)[c(-2, -3, NA)]
# Error in (1:4)[c(-2, -3, NA)] : 
#   only 0's may be mixed with negative subscripts

The textual error message indicates that it is intended behavior.

Here's my best guess as to why that is the intended behavior:

From the way they treat NA's elsewhere (e.g. typically defaulting to na.rm=FALSE), it seems that R's designers view NA's as carrying important information, and are loath to drop that without some explicit instruction to do so. (Fortunately, setting nomatch=0 gives you a clean way to pass that instruction along!)

In this context, the designers' preference probably explains why NA's are accepted for positive indexing, but not for negative indexing:

# Positive indexing: works, because the return value retains info about NA's
(1:4)[c(2,3,NA)]

# Negative indexing: doesn't work, because it can't easily retain such info
(1:4)[c(-2,-3,NA)]