[yurishkuro]

This is a follow-up to PR #42 that was trying to fix issue #37. The fix included a change to set the input to an empty map

		var mapVal map[string]interface{}
		inputVal = reflect.MakeMap(reflect.TypeOf(mapVal))

However, it was breaking some of the tests in OTEL PR because the map was not expected for non-map target types:

'boolean' expected type 'bool', got unconvertible type 'map[string]interface {}', value: 'map[]'

Changes:

• add a unit test that catches the above error
• when target type is map/struct or slice/array, set input to a nil pointer to map / slice respectively. Otherwise set it to zero value of the target type
• do not proceed to decoding if the input is nil even after the hook is run
• fix error reporting from one of the decode function - the error itself was panicking by calling .Type() on zero value. The same issue likely exists in all other error formatting functions, but I didn't want to bundle too many fixes into this PR.

Tests:

$ go test -race -shuffle=on ./...
?   	github.com/go-viper/mapstructure/v2/internal/errors	[no test files]
ok  	github.com/go-viper/mapstructure/v2	1.236s

Lints:

$ ../../jaegertracing/jaeger/.tools/golangci-lint run

(no script in the repo, I used a linter from Jaeger, which uses v1.61.0)

[yurishkuro]

This test would panic on main, or if the error formatting is fixed it would return 'boolean' expected type 'bool', got unconvertible type 'map[string]interface {}', value: 'map[]'

[yurishkuro]

@sagikazarmark one more fix for DecodeNil option, PTAL

[sagikazarmark]

Thanks @yurishkuro !

I'm going to release this as 2.2.1 since it's really a bugfix for that previous PR. Good catch by the way!

[yurishkuro]

Thanks! 🎉