proposal: Go 2: default values for named return parameters

[qexk]

Would you consider yourself a novice, intermediate, or experienced Go programmer?

Experienced.

What other languages do you have experience with?

Many functional programming languages, C, Perl, JS/TS, and lots of others.

Would this change make Go easier or harder to learn, and why?

This would add a bit of complexity, but it wouldn’t make Go any harder to learn because novice developers don’t use named return parameters that often.

Has this idea, or one like it, been proposed before?

I didn’t find anything.

Who does this proposal help, and why?

It helps developers and reviewers understand what value will be returned by a function most of the time.

What is the proposed change?

Go has the ability to give names to return parameters. It is a great feature, as it gives some kind of documentation to the reader, and allows developers to better clarify their intent if they’re writing a long function.

However, named return parameters are always zero initialised. Although this is great for consistency, I feel like an explicit default value can help in readability.

func Function() (state int := 2, err error) {
	...
}

Think of it as the default case of a switch. It could also represent a contract with a default value if the function succeeds, and that contract can be understood just by reading the function’s prototype.

Other syntax proposals:

• the parameter’s type could be inferred, but I find it harder to read

  func Function() (state := 2, err error) {
  	...
  }

• the operator used could be a simple =, like a const definition:

  func Function() (state int = 2, err error) {
  	...
  }

There should be restrictions on the type of expression you can write. This shouldn’t be allowed IMO:

func Function() (state int := func() int {
	return 12
}(), err error) {
	...
}

Maybe allow only const expressions or a single variable name?

Is this change backward compatible?

This change is backward compatible.

Show example code before and after the change.

Here’s the piece of code that gave me the idea for this proposal (this function represents a state of a parser):

func pRaw(p *parser) (nextState stateFn, err error) {
	switch p.item.Typ {
	case lexer.ItemRaw:
		p.raw.WriteString(p.item.Val)
		nextState = pRaw

	case lexer.ItemSep:
		p.pushRawIfAny()
		p.pushSeparator()
		nextState = pRaw

	case lexer.ItemLacc:
		p.pushRawIfAny()
		nextState = pExprBegin

	case lexer.ItemEOF:
		p.pushRawIfAny()
		nextState = nil

	default:
		err = UnimplementedError{p.item, "pRaw"}
	}
	return
}

This function could be rewritten as:

func pRaw(p *parser) (nextState stateFn := pRaw, err error) {
	switch p.item.Typ {
	case lexer.ItemRaw:
		p.raw.WriteString(p.item.Val)

	case lexer.ItemSep:
		p.pushRawIfAny()
		p.pushSeparator()

	case lexer.ItemLacc:
		p.pushRawIfAny()
		nextState = pExprBegin

	case lexer.ItemEOF:
		p.pushRawIfAny()
		nextState = nil

	default:
		err = UnimplementedError{p.item, "pRaw"}
	}
	return
}

Here, I understand that the next state of the parser is likely going to be pRaw.

What is the cost of this proposal? (Every language change has a cost).

This proposal would affect a lot of tools, but the implementation cost isn’t big.

The compile time cost should be negligible.

No runtime cost.

Can you describe a possible implementation?

I made a quick & dirty proof of concept here.

How would the language spec change?

The Signature section of the language spec should be the only one affected by this proposal.

Orthogonality: how does this change interact or overlap with existing features?

This extends the named return parameter feature.

---

Thank you for reading this proposal. This can seem useless, as one could just assign to the return parameter on the first line, but I think the readability is worth it.

[ianlancetaylor]

It seems hard to explain why result parameters can have default values but ordinary parameters can't.

Why not just make the first statements of the function assignments to the result parameters?

[ianlancetaylor]

Based on the (lack of) discussion above, and the emoji voting, this is a likely decline. Leaving open for four weeks for final comments.

[ianlancetaylor]

No further comments.