find and rfind differ on empty needle

[n5m]

find and rfind return different values when the string that is being searched for has length 0.

Example

import strutils

doAssert "abc".find("") == 0
doAssert "abc".rfind("") == -1

Current Output

Both assertions pass.

Expected Output

One of the following:

1. find and rfind both return 0
2. find and rfind both return -1
3. find returns low("abc"), and rfind returns high("abc")

Which one is the best?

Additional Information

git hash cfe1924

$ nim -v
Nim Compiler Version 1.5.1 [Linux: amd64]
Compiled at 2021-05-29
Copyright (c) 2006-2021 by Andreas Rumpf

git hash: cfe19247e856950f7facaf7adbc50bca8dec3990
active boot switches: -d:release

[Clyybber]

Option 3 is the most intuitive and logical to me.

[timotheecour]

One of the following:

none of the following. instead it should be:
4: find returns low("abc"), and rfind returns len("abc")

eg:

doAssert "abc".find("") == 0
doAssert "abc".rfind("") == 3
doAssert "abc".rfind("c") == 2
doAssert "abc".rfind("bc") == 1
doAssert "abc".rfind("abc") == 0

[kaushalmodi]

@timotheecour How would you distinguish between

doAssert "abc".find("a") == 0 # expected and current behavior

and

doAssert "abc".find("") == 0 # current and proposed behavior

[timotheecour]

@kaushalmodi just replying to your comment here: #18128 (comment)

I don't understand this comment; my proposal would change "abc".rfind(""), not "abc".find("")

both currently and under my proposal, you'd still have:

doAssert "abc".find("") == 0
doAssert "abc".find("a") == 0
doAssert "abc".find("ab") == 0
doAssert "abc".find("abc") == 0

which makes sense

[kaushalmodi]

my proposal would change "abc".rfind(""), not "abc".find("")

Yes, I understand. I brought it up as this issue is about fixing the find and rfind for "".

---

doAssert "abc".find("") == 0
doAssert "abc".find("a") == 0

which makes sense

It feels strange to me that the first occurrence of both "" and "a" is at index 0 of the "abc" string. Technically, "abc"[0] == 'a'.

[kaushalmodi]

Thinking more on this, "abc".find("") == 0 makes sense if I think in this manner:

doAssert "abc".find("") == 0
doAssert "abc".find("" & "a") == 0
doAssert "abc".find("" & "ab") == 0
doAssert "abc".find("" & "abc") == 0

---

Summary

No objection with "abc".find("") == 0 and "abc".rfind("") == 3 👍