go/types: add Func.Signature method

[adonovan]

Proposal Details

The expression f.Type().(*types.Signature), where f is a *types.Func, appears over a hundred times in x/tools. Each time I read it, I am momentarily compelled to prove a little theorem that the type assertion is sound.

For brevity, convenience, and simplicity, I propose that we add this method to go/types:

// Signature returns the signature (type) of the function or method.
func (f *Func) Signature() *Signature {
    return f.Type().(*types.Signature)
}

[gopherbot]

Change https://go.dev/cl/565035 mentions this issue: x/tools: add explicit Unalias operations

[findleyr]

Thanks for proposing this. "All Funcs have Signature type" is one of those non-obvious things that must be repeatedly recalled (why can't they have Typ[Invalid]...). This API will significantly improve readability for even experienced readers.

[gopherbot]

Change https://go.dev/cl/565375 mentions this issue: go/types: add Func.Signature method

[mvdan]

Would this API ever return nil or panic? Presumably not, but what if I called NewFunc with a nil signature, which seems to be allowed?

[adonovan]

Would this API ever return nil or panic? Presumably not, but what if I called NewFunc with a nil signature, which seems to be allowed?

Good question. I don't know why sig=nil was ever permitted, but it was, and it appears to be used (e.g. to construct special-purpose func symbols with no recv/tparams/params/results). I think we should interpret nil as the trivial signature, and make sig!=nil an invariant.

[mvdan]

I would agree with that. Perhaps nil was used to avoid allocations, in which case you could reuse a global empty signature, akin to other globals like the types.Typ table. Documenting how NewFunc understands a nil signature would also be nice.

[rsc]

Based on the discussion above, this proposal seems like a likely accept.
— rsc for the proposal review group

The proposal is to add:

// Signature returns the signature (type) of the function or method.
func (f *Func) Signature() *Signature

[aclements]

Based on #65772 (comment), should we specify in the doc comment what happens with a nil type? Or, @adonovan, are you proposing a change to NewFunc if sig==nil (that seems like it wouldn't be backwards compatible)?

[adonovan]

Based on #65772 (comment), should we specify in the doc comment what happens with a nil type? Or, @adonovan, are you proposing a change to NewFunc if sig==nil (that seems like it wouldn't be backwards compatible)?

It was never intended that nil was a valid argument to NewFunc, but it may have been exploited. We should change NewFunc to replace nil with a trivial non-nil Signature to preserve the invariant.

[rsc]

No change in consensus, so accepted. 🎉
This issue now tracks the work of implementing the proposal.
— rsc for the proposal review group

The proposal is to add:

// Signature returns the signature (type) of the function or method.
func (f *Func) Signature() *Signature