Sorry, this change will break the following code:

mut x = 3
def a [] {
    $x = 4
}

It should raise error.

Sorry, this change will break the following code:

mut x = 3
def a [] {
    $x = 4
}

It should raise error.

Ah we definitely need a test for those scoping/mutability rules

@WindSoilder good catch! I added a boolean field to SyntaxShape::Block that makes it parse into an Expr::Closure. Hopefully that's alright?

Thanks! I think I get the idea behind these changes, but TBH I'm not sure if the solution is good. It seems too hacky.
Please give me some time to think about this :)

I also found it randomly break my script:

export def main [word: string] {
    if true {
        return "aa"
    } else {
        # query from web
        try {
            let result = http get $"http://url" -m 4 | str length
            $result
        } catch {|e|
            if true {
                let spell_check = "a"
                $spell_check
            } else {
                error make ($e | get raw)
            }
        }
    }
}

It gives me the following error:

Error: nu::parser::parse_mismatch_with_full_string_msg

  × Parse mismatch during operation.
    ╭─[entry #47:4:12]
  3 │             return "aa"
  4 │ ╭─▶     } else {
  5 │ │           # query from web
  6 │ │           try {
  7 │ │               let result = http get $"http://url" -m 4 | str length
  8 │ │               $result
  9 │ │           } catch {|e|
 10 │ │               if true {
 11 │ │                   let spell_check = "a"
 12 │ │                   $spell_check
 13 │ │               } else {
 14 │ │                   error make ($e | get raw)
 15 │ │               }
 16 │ │           }
 17 │ ├─▶     }
    · ╰──── expected one of a list of accepted shapes: [Block(true), Expression]
 18 │     }
    ╰────

Update: @WindSoilder that error should be fixed by #14420

That parse error seems to start with the max-time type change (d289c77). I'll work on parse_multispan_value's OneOf handling to try and fix this (reproducible on main):

if true { http get "" -m 4 } else { } # expected duration with valid units
if true { } else { http get "" -m 4 } # expected one of a list of accepted shapes: [Block, Expression]

TBH I'm not sure if the solution is good. It seems too hacky. Please give me some time to think about this :)

It's also a breaking change for plugins. That alone might call for a different approach.

where are we at on this?

I'll fix the conflict but don't know of a cleaner way to solve this without Block(bool) or a new SyntaxShape variant for blocks that can't capture mutable variables.

To be honest I still don't think it's the right way to go. Hopefully we can solve it in the new parser.
So I'd like to close the pr and leave that issue open.