v1.4.6 change has broken existing behaviour

[inhere]

Regression

This change has broken existing behaviour.

For eg: I have a JSON:

{
  "main1": {
    "child11": 1,
    "child12": 2
  },
  "main2": {
    "child21": 5,
    "child22": 6
  }
}

Now, as per validation rules, key main1 and main2 are optional. And their respective children keys should be validated only when main key exists.

The validation rules are defined as:

type Request struct {
	Main1 *struct {
		Child11     string `json:"child11" validate:"required"`
		Child12 string `json:"child12" validate:"required"`
	} `json:"main1"`
	Main2 *struct {
		Child21 string `json:"child21" validate:"required"`
		Child22 string `json:"child22" validate:"required"`
	} `json:"main2"`
}

Earlier, uptil v1.4.5, these definitions were working fine and children struct was validated only when the respective main key existed

Now, with the release v1.4.6 , this behaviour does not work anymore as nil pointer check is removed.

Originally posted by @madhurbhaiya in #191 (comment)

[lsh-0]

I thought I was going crazy! I'm new to go and gookit and have been trying to get 'child' fields of an embedded struct ignored for hours ... I've learnt quite a bit about gookit in the process but still, was getting very frustrated.

In my case I have:

type SubConfig1 struct {
	Val int `validate:"required|min:1|max:65535"`
}

type SubConfig2 struct {
	Name string `validate:"required|in:foobar"`
}

type Config struct {
	Sub1 *SubConfig1
	Sub2 *SubConfig2
}

and then ignoring the other sub-config structs depending on what the user passes in:

        // [...]
        user_command := "do2"
	idx := map[string]string{
		"Sub1": "do1",
		"Sub2": "do2",
	}
	for field_name, cmd := range idx {
		if user_command != cmd {
			v.AddRule(field_name, "safe")
		}
	}

I'm sure there is an easier way (that doesn't involve that idx), but in 1.4.6 this doesn't work and 1.4.5 it works as expected.

Thanks @inhere for saving my evening :)

[lsh-0]

Can I get some indication if this is a bug that will be fixed or if I was relying on undocumented behaviour?

If I was relying on undocumented behaviour and this isn't a bug to be fixed then I need to find some other way or library to achieve this validation.

[inhere]

hi @lsh-0 @madhurbhaiya @OscarVanL Add new validator optional for resolve this issue.

On struct, add optional on parent field, will be not validating sub-fields on parent is nil.

validate/issues_test.go

Lines 1267 to 1306 in 27b33f1

	func TestIssues_192(t *testing.T) {
	type Request struct {
	Main1 *struct {
	Child11 int `json:"child11" validate:"required"`
	Child12 int `json:"child12" validate:"required"`
	} `json:"main1" validate:"optional"` // optional - will not validate Main1.* on Main1 is nil
	Main2 *struct {
	Child21 int `json:"child21" validate:"required"`
	Child22 int `json:"child22" validate:"required"`
	} `json:"main2"`
	}
	jsonStr := `{
	"main1": {
	"child12": 2
	},
	"main2": {
	"child21": 5,
	"child22": 6
	}
	}`
	req := &Request{}
	err := jsonutil.DecodeString(jsonStr, req)
	assert.NoErr(t, err)
	t.Run("with child data", func(t *testing.T) {
	v := validate.Struct(req)
	assert.False(t, v.Validate())
	assert.StrContains(t, v.Errors.String(), "main1.child11 is required to not be empty")
	})
	// set main1 = nil, should not validate Main1.*
	t.Run("set main1 to nil", func(t *testing.T) {
	req.Main1 = nil // fields under Main1 will not be validated
	v := validate.Struct(req)
	assert.True(t, v.Validate())
	})
	}

[lsh-0]

thank you @inhere, using the master branch of gookit/validate with the new optional validator appears to fix the issue for me.

[inhere]

Please upgrade to https://github.com/gookit/validate/releases/tag/v1.5.2