Merge pull request apache#65 from concretevitamin/parallelize-fix

Fix that collect(parallelize(sc,1:72,15)) drops elements.

pkg/R/context.R

@@ -62,7 +62,7 @@ parallelize <- function(sc, coll, numSlices = 1) {
   if (numSlices > length(coll))
     numSlices <- length(coll)
 
-  sliceLen <- length(coll) %/% numSlices
+  sliceLen <- ceiling(length(coll) / numSlices)
   slices <- split(coll, rep(1:(numSlices + 1), each = sliceLen)[1:length(coll)])
 
   # Serialize each slice: obtain a list of raws, or a list of lists (slices) of

pkg/inst/tests/test_parallelize_collect.R

@@ -67,6 +67,17 @@ test_that("collect(), following a parallelize(), gives back the original collect
   expect_equal(collect(strListRDD2), as.list(strList))
 })
 
+test_that("regression: collect() following a parallelize() does not drop elements", {
+  lapply(1:72,
+         function(collLen) {
+           lapply(1:15, function(numPart) {
+             expected <- runif(collLen)
+             actual <- collect(parallelize(jsc, expected, numPart))
+             expect_equal(actual, as.list(expected))
+           })
+         })
+})
+
 test_that("parallelize() and collect() work for lists of pairs (pairwise data)", {
   # use the pairwise logical to indicate pairwise data
   numPairsRDDD1 <- parallelize(jsc, numPairs, 1)