str_split creates empty leading "" when splitting on "" about stringr HOT 6 CLOSED

The behaviour of strsplit confused me a bit. Is this consistent?

strsplit("", "")
# [[1]]
character(0)
strsplit(" ", " ")
# [[1]]
# [1] ""

from stringr.

After reading the algorithm in ?strsplit, I found this is consistent but not typical. Usually (i.e., in the other languages), the second command should return 2 empty characters such as ["", ""].

from stringr.

The latter seems correct to me, since it should behave the same as

> strsplit("a","a")
[[1]]
[1] ""

But I think once you've decided that splitting on "" means to 'split into individual characters', whether strsplit("","") should return "" or character(0) seems like a judgement call to me.

My issue with the behavior of str_split is that pulling only the leading zero-length string seems very arbitrary. One could argue that there are infinitely many instances of "" in every string! Or at least one between every single character. So why shouldn't it return something like:

str_split("abc","")
[[1]]
[1] "" "a" "" "b" "" "c" ""

That's why I think 'split into individual characters' should generally mean 'split into non-zero length characters' with the exception of when you're splitting something that is already empty.

from stringr.

Ok, that helps. Unfortunately I don't have time to write a fix at the moment, but tests/patches would be appreciated! Does this problem also affect str_split_fixed?

from stringr.

It does effect str_split_fixed, in the sense that it also assumes that every string has a leading empty string:

str_split_fixed("abc","",4)
     [,1] [,2] [,3] [,4]
[1,] ""   "a"  "b"  "c" 

Again, I think this is mainly confusing because ?str_split_fixed also says that splitting on "" will split into single characters. At the moment, str_split_fixed behaves like this:

> str_split_fixed("abc","",1)
     [,1] 
[1,] "abc"
> str_split_fixed("abc","",2)
     [,1] [,2] 
[1,] ""   "abc"
> str_split_fixed("abc","",3)
     [,1] [,2] [,3]
[1,] ""   "a"  "bc"
> str_split_fixed("abc","",4)
     [,1] [,2] [,3] [,4]
[1,] ""   "a"  "b"  "c" 

where I would expect either that pattern = "" would override n completely and always force a split into "a" "b" "c". Or possibly something more subtle like this:

> str_split_fixed("abc","",1)
     [,1] 
[1,] "abc"
> str_split_fixed("abc","",2)
     [,1] [,2] 
[1,] "a"   "bc"
> str_split_fixed("abc","",3)
     [,1] [,2] [,3]
[1,] "a"   "b"  "c"
> str_split_fixed("abc","",4)
     [,1] [,2] [,3] [,4]
[1,] "a"   "b"  "c"  "" 

from stringr.

Fixed in stringi branch

from stringr.