Chapel: string slice expressions s[..0] and s[0..0] returning empty string

[ben-albrecht]

Summary of Problem

String slices s[..0] and s[0..0] return an empty string rather than s[0].

Steps to Reproduce

Source Code:

var s = 'string indexing is "fun"';

// These should all be true
writeln(s[0] == s[..0]);    // false
writeln(s[0] == s[0..0]);   // false

// Some other checks
writeln(s[0..1] == s[..1]); // true
writeln(s[1] == s[1..1]);   // true
writeln(s[1] == s[1..1]);   // true

writeln(s[..5] == s[0..5]); // true
writeln(s[2..2] == s[2]);   // true

// Sanity check to confirm it's only strings
var A = [1,2,3,4];

writeln(A[0] == A[..0]);   // true
writeln(A[0] == A[0..0]);  // true

Configuration Information

• Output of chpl --version: chpl version 1.23.0 pre-release (dcb3f0e574)

[ben-albrecht]

Thanks @cassella for helping identify this one.

[bradcray]

To be even more precise, s[..0] and s[0..0] are returning empty strings where they should be returning s.

[bradcray]

Happily, these same failures do not seem to occur in the bytes type.

[cassella]

This was discovered in, and the cause of, #15522.

[e-kayrakli]

There may be a quick fix based on the 0 vs 1-based indexing changes. But the potentially responsible part of the string implementation is problematic in different respects. I hope to resolve this as part of a refactor of that code as early as next week.

[ram-nad]

I just checked that, changing

1445: if cp_high > 0 {

to

1445: if cp_high >= 0 {

works.

[e-kayrakli]

Thanks @ram-nad, that's the quick fix that I was suspecting. I am hoping to make bigger changes to that code starting today or tomorrow, your observation will help :)

[ben-albrecht]

A note so we don't forget -- the following work-around in mason can be reverted when this is fixed:

https://github.com/chapel-lang/chapel/pull/15529/files#diff-71a91c1fdf3c3d0e5a204b343fafd8d6R89