[Fix #483] add_duration_to_date with leading NA

* Fix code that had wrong test for date in ops-addition.r
* Added tests to test-ops-additions.R
* Update NEWS.md

NEWS.md

@@ -10,7 +10,7 @@ Version 1.6.0.9000
 * [#472](https://github.com/hadley/lubridate/pull/472) Printing method for duration doesn't throw format error on fractional seconds.
 * [#475](https://github.com/hadley/lubridate/pull/475) character<> comparisons is no longer slow.
 * [#486](https://github.com/hadley/lubridate/issues/486) ceiling_date handles `NA` properly.
-
+* [#483](https://github.com/hadley/lubridate/pull/483) Fix add_duration_to_date error when duration first element is NA.
 
 Version 1.6.0
 =============

R/ops-addition.r

@@ -16,8 +16,16 @@ add_duration_to_date <- function(dur, date) {
   if(is.Date(date)){
     date <- as.POSIXct(date)
     ans <- with_tz(date + dur@.Data, "UTC")
-    if (hour(ans) == 0 && minute(ans) == 0 && second(ans) == 0)
+
+    if ( all(is.na(ans) ) ) return(as.Date(ans))  # ALL NAs
+
+    if ( all( hour(na.omit(ans)) == 0  &
+              minute(na.omit(ans)) == 0 &
+              second(na.omit(ans)) == 0 )
+    ) {
       return(as.Date(ans))
+    }
+
     return(ans)
   }
   new <- date + dur@.Data

tests/testthat/test-ops-addition.R

@@ -316,3 +316,26 @@ test_that("addition with period months and years returns NA when appropriate", {
   expect_equal(leap - years(0:4), yrs2)
 
 })
+
+test_that("addition with durations containing NA", {
+
+  dt <- ymd(20161018)
+  dt_1 <- ymd(20161019)
+
+  dd_1na  <- ddays(c(1,NA))
+  dd_na1  <- ddays(c(NA,1))
+  dd_nana <- ddays(c(NA,NA))
+
+  ans_1na  <- add_duration_to_date( dd_1na, dt )
+  ans_na1  <- add_duration_to_date( dd_na1, dt )
+  ans_nana <- add_duration_to_date( dd_nana, dt )
+
+  expect_is( ans_1na, "Date")
+  expect_is( ans_na1, "Date")
+  expect_is( ans_nana, "Date")
+
+  expect_equal( ans_1na, c(dt_1,NA) )
+  expect_equal( ans_na1, rev(c(dt_1,NA)) )
+  expect_equal( ans_nana, as.Date(c(NA,NA)) )
+
+})