Thanks for sending @michaellzc - I'd be happy to merge this.

I can offer a few options:

1. I can merge this as-is, without support for *time.Time - I'm content with doing that.
2. You can implement *time.Time using the generic Ptr approach, I would be happy to have that change and require the latest-and-greatest Go version in go.mod.

I'll let you decide which one you want to go for depending on how much time you want to spend here.

2. You can implement *time.Time using the generic Ptr approach, I would be happy to have that change and require the latest-and-greatest Go version in go.mod.

😛 I need *time.Time for my own use case, so I'll continue on option 2 with a twist:

// an exported func from autogold
func TimePtr(t time.Time) *time.Time {
	return &t
}

// render as
autogold.TimePtr(time.Date(...))

SomethingPtr is quite a common pattern in many Go projects, e.g., k8s.io/utils/pointer, github.com/aws/jsii-runtime-go.

it's not elegant, but we can avoid breaking backward compatibility unless we have to.

@michaellzc I'm not a fan of that TimePtr change - that feels awkward and a bit cluttery of the API space. I have zero qualms about breaking compatibility, so please do the Ptr approach unless there's something that makes that much more difficult

nice, let's do the Ptr then

Codecov Report

Merging #25 (1c77fe9) into main (5de3f44) will increase coverage by 1.02%.
The diff coverage is 97.14%.

❗ Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.

@@            Coverage Diff             @@
##             main      #25      +/-   ##
==========================================
+ Coverage   74.36%   75.38%   +1.02%     
==========================================
  Files           6        7       +1     
  Lines         745      780      +35     
==========================================
+ Hits          554      588      +34     
- Misses        144      145       +1     
  Partials       47       47              

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

@michaellzc not sure if you'd be up for it (fine if not) but maybe you could remove valast.Addr in favor of valast.Ptr? It's doing the same thing but results in uglier ASTs as it doesn't use generics:

No worries if not, just let me know what you decide & I'll merge either way.

@michaellzc not sure if you'd be up for it (fine if not) but maybe you could remove valast.Addr in favor of valast.Ptr? It's doing the same thing but results in uglier ASTs as it doesn't use generics:

valast/valast.go

Lines 151 to 167 in 5de3f44

	// Addr returns a pointer to the given value.
	//
	// It is the only way to create a reference to certain values within a Go expression,
	// for example since &"hello" is illegal, it can instead be written in a single expression as:
	//
	// valast.Addr("hello").(*string)
	func Addr(v interface{}) interface{} {
	vv := reflect.ValueOf(v)
	// Create a slice with v in it so that we have an addressable value.
	sliceType := reflect.SliceOf(vv.Type())
	slice := reflect.MakeSlice(sliceType, 1, 1)
	if v != nil {
	slice.Index(0).Set(vv)
	}
	return slice.Index(0).Addr().Interface()
	}

No worries if not, just let me know what you decide & I'll merge either way.

refactor from Addr to Ptr is done.

Trying to understand what is going in AddrInterface see if we can get rid of it as well

@michaellzc not sure if you'd be up for it (fine if not) but maybe you could remove valast.Addr in favor of valast.Ptr? It's doing the same thing but results in uglier ASTs as it doesn't use generics:

valast/valast.go

Lines 151 to 167 in 5de3f44

	// Addr returns a pointer to the given value.
	//
	// It is the only way to create a reference to certain values within a Go expression,
	// for example since &"hello" is illegal, it can instead be written in a single expression as:
	//
	// valast.Addr("hello").(*string)
	func Addr(v interface{}) interface{} {
	vv := reflect.ValueOf(v)
	// Create a slice with v in it so that we have an addressable value.
	sliceType := reflect.SliceOf(vv.Type())
	slice := reflect.MakeSlice(sliceType, 1, 1)
	if v != nil {
	slice.Index(0).Set(vv)
	}
	return slice.Index(0).Addr().Interface()
	}

No worries if not, just let me know what you decide & I'll merge either way.

refactor from Addr to Ptr is done.

Trying to understand what is going in AddrInterface see if we can get rid of it as well

possible solution for future reference:

// It doesn't work for all cases yet
func InterfacePtr[V any, I any](v *V, pointerToType *I) *I {
	sliceType := reflect.SliceOf(reflect.TypeOf(pointerToType).Elem())
	slice := reflect.MakeSlice(sliceType, 1, 1)
	if v != nil {
		slice.Index(0).Set(reflect.ValueOf(v))
	}
	return slice.Index(0).Addr().Interface().(*I)
}

I may find some existing problems with AddrInterface, so I am going to leave it as it is.

1. Many tests are not setting options to ExportedOnly, and the rendered code contains a private struct foo from internal/test pkg.

2. Rendered output of TestPointers/interface is not valid Go

// input
v := test.Bazer(test.NewBaz())
&v

// output
valast.AddrInterface(test.Bazer{&test.Baz{
	Bam:  (1.34 + 0i),
	zeta: &test.foo{bar: "hello"},
}},
	(*test.Bazer)(nil)).(*test.Bazer)

I don't think you can do test.Bazer{&test.Baz{}} with an interface in Go (Bazer is an interface)