String.split does not fully remove repeated patterns

[vanstee]

When splitting with a pattern that is repeated, empty strings are left in the resulting list:

iex(1)> String.split("foo!bar!!!", %r/!/)
["foo", "bar", "", "", ""]               

I would expect the result to simply be:

["foo", "bar"]

String.split/1 seems to remove repeated whitespace, which makes it seem inconsistent:

iex(1)> String.split("foo bar   ")
["foo", "bar"]

This does start diverging from the semantics of :binary.split. Although we should probably at least support the :trim option, but that would still leave empty strings if the pattern is repeated at the beginning of the string.

[josevalim]

Thanks @vanstee! We need for now to at least fix String.split/1 to not trim any result. Could you please send a pull request?

[vanstee]

@josevalim Sure!

Just wanted to double check. This change would make the following true:

iex(1)> String.split(" foo  bar ")
["", "foo", "", "bar", ""]

Is that the intent? It would then match binary:split but still seems a little inconvenient to work with.

[ericmj]

@vanstee That's how I would assume that String.split works. If you don't want empty strings you could always Enum.filter the result.

[josevalim]

It should actually be:

iex> String.split "  foo bar  ", " "
["", "", "foo", "bar", "", ""]

[vanstee]

Cool! I'll put this together.