Why String.split uses the delim input as set of chars instead of Regex pattern or string delim?

[enigma]

Eg:

use "collections"
use "regex"

actor Main
  new create(env: Env) =>
    let base = "opinion"
    let delim = "pi"
    for i in base.split(delim).values() do
      env.out.write("actual[")
      env.out.write(i)
      env.out.print("]")
    end
    env.out.print("============")
    for i in alternative_split(base, delim).values() do
      env.out.write("expected[")
      env.out.write(i)
      env.out.print("]")
    end

  fun alternative_split(string: String, delim: String): Array[String] =>
    try
      return Regex(delim).split(string)
    end
    Array[String]

prints:

actual[o]
actual[]
actual[n]
actual[on]
============
expected[o]
expected[nion]

alternative_split is still not that great now as one has to escape regex special symbols. Also using Regex involves a lot of partial methods.
Given String.split takes a string I'd expect a behaviour more similar to what Python does:

>>> 'this is (not) true'.split(' (not) ')
['this is', 'true']

My understanding is that this doesn't matter for the default argument behaviour, ie split by whitespacy symbols.

Could the signature of split be:

fun split(delim: (String | None), n: USize = 0): Array[String] iso^ =>
   // if delim is None keep current behavior, splitting on any of " \t\v\f\r\n"

I think that could keep the default behaviour and make the specified one a bit less surprising (even though it's documented).
Also because of the influence Python had on me I think a common use case that might deserve to be the default is more similar to Regex("[\s\t\v\f\r\n]+").split. What do you think?

[jemc]

I haven't looked at this in detail yet, but it's possible we might want to consider this as a "principle of least surprise bug". I'm adding the "needs discussion during sync" label so we can discuss it at the next public weekly meeting.

[SeanTAllen]

When this came up on IRC, I was very surprised. I consider this a bug.

[SeanTAllen]

Discussed on sync. This is a "principle of least surprise bug".

[SeanTAllen]

@enigma do you want to take this on? If yes, I can add you as a read only contributor so you can be assigned to the issue.

[enigma]

@SeanTAllen yes please!

[SeanTAllen]

O wow. we can assign people who aren't team members now @jemc ?

[jemc]

@SeanTAllen - I assumed you had already added him.

But now that you mention it, I think maybe you can assign someone who is already participating in the ticket (or maybe the requirement is that they filed the ticket?).

[SeanTAllen]

@jemc that's a pretty cool change. i don't think it used to be that way.

[SeanTAllen]

@enigma shall i mark this as in progress?

[enigma]

@SeanTAllen Yes please. I don't think I can change any label.

[SeanTAllen]

Hi @enigma,

Any progress?

Need any assistance?

[SeanTAllen]

I'm unassigned @enigma and setting this back to ready for work.

[andersarpi]

Isn't the functionality requested here exactly what split_by does?

[enigma]

I didn't notice this was still open, my bad!

This should have been closed by f566596 - unless we want split_by to replace split for good.