Wrong result of unary `-` operation

[taichi-ishitani]

Pkl will return the wrong result for the unary - operation like below.

// foo.pkl
a = --1

The result of this operation should be 1 but got -1.

[bioball]

Thanks for the report; definitely looks like bug. Haven't looked into it yet, but this looks parser related. Unary minus otherwise works as expected:

negativeOne = -1
a = -negativeOne

Produces:

negativeOne = -1
a = 1

[madrob]

This is fun!

[DFrenkel]

Double unary minus could be easily confused with a decrement.

Java, JavaScript, Golang, .NET all fail to compile --1 expression either expecting parentheses around (-1) or expecting a variable to be decremented instead of a literal. Increment/decrement ++ and -- are deprecated in Swift 2.2 and removed in Swift 3. Python, however, handles x = --1 correctly.

I think pkl should follow the simplest solution and expect ++a or --a to fail with a clear message . An alternative is to start supporting prefix (and postfix for symmetry?) increments and decrements:

• Expect --a to fail when a is numerical literal
• Expect --a to mean "decrease a by 1 and return the new value of a as result" when a is a variable

What do you think?

[HT154]

This bug extends through parentheses as well:

x = -(-1)
// x = -1

A decrement operator of any kind seems like the wrong choice for Pkl as it implies some form of imperative mutation that doesn't exist in the language.

[bioball]

Good point, @DFrenkel. I'm sympathetic to the idea that --a should hard-fail, because it might be confused with decrement.