fix eager creation for associations by adding where clause

[xixinjie]

Problem

When try to use db.Eager().ValidateAndCreate(user) to create a user and underlying has_many books, it succeeded for the first time when books table is empty, but failed to save the books in following creations.

How to fix

Add where clause when check if the books exist as shown in the PR.

exists, errE := Q(c).Where("id = ?", id).Exists(i)

Samples

type User struct {
	ID string `json:"id" db:"id"`
	Name string `json:"name" db:"name"`
	Books []Book `json:"books" has_many:"books" fk_id:"user_id"`
	CreatedAt time.Time `json:"created_at" db:"created_at"`
	UpdatedAt time.Time `json:"updated_at" db:"updated_at"`
}

type Book struct {
	ID string `json:"id" db:"id"`
	Name string `json:"name" db:"name"`

	UserID string `json:"user_id" db:"user_id"`
	User *User `json:"user" belongs_to:"users"`

	CreatedAt time.Time `json:"created_at" db:"created_at"`
	UpdatedAt time.Time `json:"updated_at" db:"updated_at"`
}

	u := &User{
		ID:        "1",
		Name:      "Foo",
		Books:     []Book{{
			ID:        "1",
			Name:      "Hello World!",
		}},
	}

	vErrs, err := db.Eager().ValidateAndCreate(u)
	if err != nil {
		panic(fmt.Sprintf("failed to create user: %s", err))
	}

	if vErrs.HasAny() {
		panic(fmt.Sprintf("failed to match validation: %s", vErrs))
	}

	log.Printf("created a new user: %s", u.ID)

SQL logs:

sql - INSERT INTO `users` (`created_at`, `id`, `name`, `updated_at`) VALUES (:created_at, :id, :name, :updated_at)
sql - SELECT EXISTS (SELECT books.created_at, books.id, books.name, books.user_id, books.updated_at FROM books AS books WHERE id = ?) | ["1"]
INSERT INTO `books` (`created_at`, `id`, `name`, `user_id`, `updated_at`) VALUES (:created_at, :id, :name, :user_id, :updated_at)

[xixinjie]

@stanislas-m @markbates any chance to review or any suggestion?

[aeneasr]

Thank you, could you please add a failing test case first which shows the behaviour you describe - first to verify your findings and second to verify your fix!

[aeneasr]

The ID column is variable depending on the struct and should therefore not be hardcoded. m has a function whereID which should be used instead.

[sio4]

Hello @xixinjie,

Thank you for reporting this issue and contribution. I took a look at the issue and definitely the original usage was problematic. Please resolve the review comment, then I will take care of the next steps along with test cases.

I am not sure if you still have an interest in this issue. If you do not respond, I will take over this issue.

[sio4]

LGTM

minor fix will be on another PR.