foverlaps issue with datetimes < 1970-01-01 (#3382)

NEWS.md

@@ -16,6 +16,8 @@
 
 4. Grouping by unusual column names such as `by='string_with_\\'` and `keyby="x y"` could fail, [#3319](https://github.com/Rdatatable/data.table/issues/3319) and [#3378](https://github.com/Rdatatable/data.table/issues/3378). Thanks to @HughParsonage for reporting and @MichaelChirico for the fixes.
 
+5. `foverlaps()` returned incorrect results overlapping on POSIXct objects which were <= `1970-01-01`, i.e., datetime values that were represented internally as -ve numeric values. This is now fixed. Closes [#3349](https://github.com/Rdatatable/data.table/issues/3349). Thanks to @lux5 for reporting.
+
 #### NOTES
 
 1. When upgrading to 1.12.0 some Windows users might have seen `CdllVersion not found` in some circumstances. We found a way to catch that so the [helpful message](https://twitter.com/MattDowle/status/1084528873549705217) now occurs for those upgrading from versions prior to 1.12.0 too, as well as those upgrading from 1.12.0 to a later version. See item 1 in notes section of 1.12.0 below for more background.

R/foverlaps.R

@@ -70,10 +70,9 @@ foverlaps <- function(x, y, by.x=if (!is.null(key(x))) key(x) else key(y), by.y=
       bits = floor(log2(.Machine$double.eps))
       2 ^ (bits + (getNumericRounding() * 8L))
     }
-    incr = 1 + dt_eps()
     isdouble = TRUE
     isposix = "POSIXct" %chin% yclass
-  } else incr = 1L # integer or Date class for example
+  }
   ## hopefully all checks are over. Now onto the actual task at hand.
   origx = x; x = shallow(x, by.x)
   origy = y; y = shallow(y, by.y)
@@ -90,15 +89,16 @@ foverlaps <- function(x, y, by.x=if (!is.null(key(x))) key(x) else key(y), by.y=
     mcall = make_call(mcols, quote(c))
     if (type %chin% c("within", "any")) {
       mcall[[3L]] = substitute(
-        if (isposix) unclass(val)*incr # incr is okay since this won't be negative
+        # datetimes before 1970-01-01 are represented as -ve numerics, #3349
+        if (isposix) unclass(val)*(1 + sign(unclass(val))*dt_eps())
         else if (isdouble) {
           # fix for #1006 - 0.0 occurs in both start and end
           # better fix for 0.0, and other -ves. can't use 'incr'
           # hopefully this doesn't open another can of worms
           (val+dt_eps())*(1 + sign(val)*dt_eps())
         }
-        else val+incr,
-        list(val = mcall[[3L]], incr = incr))
+        else val+1L, # +1L is for integer/IDate/Date class, for examples
+        list(val = mcall[[3L]]))
     }
     make_call(c(icall, pos=mcall), quote(list))
   }

inst/tests/tests.Rraw

@@ -13414,6 +13414,12 @@ DT = data.table(a = c(1:3, 3:1))
 test(1984.38, rowidv(DT, prefix = 5L), error='must be NULL or a character vector')
 test(1984.39, rowidv(DT, prefix = c('hey','you')), error='must be NULL or a character vector')
 
+# Test for #3349, foverlaps returned spurious results with POSIXct objects < 1970-01-01
+x <- data.table(val=178.41,s=as.POSIXct("1968-04-25 04:20:00"),e=as.POSIXct("1968-04-25 04:20:00"))
+y <- data.table(event="#1",s=as.POSIXct("1968-04-19 15:20:00"),e=as.POSIXct("1968-04-24 07:20:00"))
+setkey(y, s, e)
+# x$s and x$e are identical (i.e., range is actually a point, but that's okay). The point is to ensure that 'x' is not within 'y'. In older versions, this will be a match because of 'incr' value being 1 + dt_eps() instead of 1-dt_eps() (because the date here is < 1970-01-01 which is a -ve numeric value internally). 1+dt_eps() should be only for +ve numerics.
+test(1985.1, nrow(foverlaps(x, y, by.x=c("s", "e"), type="within", nomatch=0L)), 0L)
 
 ###################################
 #  Add new tests above this line  #