problem with chunk.tapply

[s-u]

problem with chunk.tapply

[s-u]

Chunking creates chunks that are guaranteed to not split keys, but they are not creating chunks that have only one key - there can be one or more keys. To illustrate on your example:

> res <- chunk.tapply(input = "iris.csv", function(x) {
+   df <- dstrsplit(x, c(list(""), rep(list(1), 4)), sep=',')
+   cat("-- chunk --\n")
+   print(table(df[[1]]))
+ }, sep=",", CH.MERGE=c)
-- chunk --

    "setosa" "versicolor" 
          50           50 
-- chunk --

"virginica" 
         50 
> res
    "setosa" "versicolor"  "virginica" 
          50           50           50 

so the first chunk contains both "setosa" and "versicolor" which the last chunk has only "virginica". Since you only printed the number of rows, you didn't see the two different keys in the first chunk.

Note that this is typically desirable since the chunks are driven by data size and not the number of keys, so it is still efficient regardless of the number of keys.