Account for roll='nearest'

Author: MichaelChirico

NEWS.md

@@ -78,7 +78,7 @@
 
 17. A data.table with a column of class `vctrs_list_of` (from package {vctrs}) prints as expected, [#5948](https://github.com/Rdatatable/data.table/issues/5948). Before, they could be printed messily, e.g. printing every entry in a nested data.frame. Thanks @jesse-smith for the report, @DavisVaughan and @r2evans for contributing, and @MichaelChirico for the PR.
 
-18. Fixed incorrect sorting of merges where the first column of a key is a factor with non-`sort()`-ed levels (e.g. `factor(1:2, 2:1)` and it is joined to a character column, [#5361](https://github.com/Rdatatable/data.table/issues/5361). Thanks to @gbrunick for the report and Benjamin Schwendinger for the fix.
+18. Fixed incorrect sorting of merges where the first column of a key is a factor with non-`sort()`-ed levels (e.g. `factor(1:2, 2:1)` and it is joined to a character column, [#5361](https://github.com/Rdatatable/data.table/issues/5361). Thanks to @gbrunick for the report, Benjamin Schwendinger for the fix, and @MichaelChirico for a follow-up fix caught by revdep testing.
 
 19. Spurious warnings from internal code in `cube()`, `rollup()`, and `groupingsets()` are no longer surfaced to the caller, [#6964](https://github.com/Rdatatable/data.table/issues/6964). Thanks @ferenci-tamas for the report and @venom1204 for the fix.
  

R/data.table.R

@@ -2048,8 +2048,9 @@ replace_dot_alias = function(e) {
   if (!.Call(CisOrderedSubset, irows, nrow(x)))
     return(NULL)
 
-  # see #1010. don't set key when i has no key, but irows is ordered and !roll
-  if (roll && length(irows) != 1L)
+  # see #1010. don't set key when i has no key, but irows is ordered and isFALSE(roll)
+  #   NB: roll could still be a string like 'nearest', #7146
+  if (!isFALSE(roll) && length(irows) != 1L)
     return(NULL)
 
   new_key <- head(x_key, key_length)

inst/tests/tests.Rraw

@@ -7123,6 +7123,11 @@ dt3 <- merge(dt1, dt2, by="x")
 test(1483.81, key(dt3), "x")
 test(1483.82, nrow(dt3[x %in% "c", ]), 2L)
 
+# avoid regression of roll='nearest' join of int->double, #7146
+DT1 <- data.table(a=1L, key='a')
+DT2 <- data.table(a=2.0, key='a')
+test(1483.83, DT1[DT2, roll='nearest'], data.table(a=2L, key='a'))
+
 # NULL items should be removed when making data.table from list, #842
 # Original fix for #842 added a branch in as.data.table.list() using point()
 # Then PR#3471 moved logic from data.table() into as.data.table.list() and now removes NULL items up front, so longer need for the branch