Change range semantic to behaves more like rust.

[WindSoilder]

Description

We have decided to make range to behave like rust, that is:

> 1..3
╭───┬───╮
│ 0 │ 1 │
│ 1 │ 2 │
╰───┴───╯
> 1..=3
╭───┬───╮
│ 0 │ 1 │
│ 1 │ 2 │
│ 2 │ 3 │
╰───┴───╯
> 1..<3
Error: nu::shell::external_command

  × External command failed
   ╭─[entry #2:1:1]
 1 │ 1..<3
   · ──┬──
   ·   ╰── executable was not found
   ╰────
  help: No such file or directory (os error 2)

This pr also:

1. removes 1..<3 syntax. The main change is happened in crates/nu-parser/src/parser.rs:parse_range
2. change range semantic in these commands due to changes in crates/nu-cmd-base/src/util.rs

• detect_columns (for -c flag)
• str substring
• str index-of (for -r flag)
• bytes at
  Take str substring as example:

Before

❯ "abcd" | str substring 1..=3
bc

After

❯ "abcd" | str substring 1..=3
bcd

The result of "abcd" | str substring 1..3 remains unchanged because the semantic of 1..3 is changed from [1,2,3] to [1,2], so it returns "bc".

Apart from changes in crates/nu-parser/src/parser.rs:parse_range and crates/nu-cmd-base/src/util.rs, remaining changes are most about adjusting test code to follow new semantic

Fixes: #7761

User-Facing Changes

It's a big breaking change, because the semantic of 1..3 is changed from [1,2,3] to [1,2]

Tests + Formatting

Adjusted all tests

After Submitting

Need to update documentation about ranges

[WindSoilder]

Make as draft because the decision is not finally made

[NotTheDr01ds]

Coming to this based on #13077 - As a user, I think the existing semantic is much more logical. I know that various languages and libraries are split on this:

• In PHP (not necessarily a language I want to mimic, but ...) range(1,3) results in 1, 2, and 3.
• In JavaScript, the d3 library does the same, but Lodash results in 1 and 2 only.
• Python range(1, 3) results in 1 and 2 only

Personally, I see 1..3 as more logically 1..=3, with 1..<3 being available if you need the alternative semantic.

On the flip side, I fully understand that slicing operations should (and must) be based on the rule "0 is the position before the first element, count up from there", which means that length + 1 is the position after the last character. E.g.:

> [ 1 2 3 4 ] | range 0..4
[ 1 2 3 ]

... and the same, as you mentioned, for str substring.

[WindSoilder]

@NotTheDr01ds Thanks for your input! Happy to see there are some languages return inclusive pair on something like 1..3(PHP in your example)
I'm ok with keeping current semantic.

So I will close the pr

[stevenxxiu]

I support this PR. Rust and Python are very popular languages, with Nushell being written in Rust too. It'll be more familiar for people coming from those, especially when manipulating strings.