golang working effectively with Null* types

Question

If one is working with a database, a Null* type is useful for most scenarios as one typically does not want a "zero" value going through, you want the NOT NULL constraints etc. to kick up and remind you that you haven't passed in all the data necessary.

So you create a structure like the following:

type Role struct {
    Id   sql.NullInt64
    Code sql.NullString
}

Thats great, but now you cannot get direct access to the properties and have to use Role.Id.Value to both get and set, this is going to get pretty old in a large app when you have to have the extra step every single time you want access to the properties.

It would be nice if you could assign directly eg. Role.Code = "Fsfs", and be able to do something like Role.Code.IsNull when you are interested in null checking. Is such a thing possible?

Answer 1

Is using intermediate pointer value(s) an option?

package main

import "fmt"

type tmp struct {
    Value int
}

func getInt() *int {
    i := 123

    return &i
}

func main() {
    // Re
    var v *int

    v = nil

    fmt.Printf("%T / %v\n", v, v)

    if v == nil {
        println("nil...")
    }

    v = getInt()

    fmt.Printf("%T / %v\n", v, *v)

    if v != nil {
        println("not nil...")
    }

    s := tmp{*v}

    fmt.Printf("%T / %v\n", s, s)
}

http://play.golang.org/p/lBrwTKh6-v

Answer 2

You can access Role.Code like that:

var r *Role
r.Code = *code

You can check for null like this:

fmt.Println(r.Code, r.Code.Valid)

If you change the value of r.Code manually without using an sql.Scanner a Setter could be helpful:

func (r *Role) SetCode(code string) {
  r.Code.String = code
  r.Code.Valid = true
}
func main() {
var r *Role 
r.SetCode("mi")
if r.Code.Valid {
  fmt.Println(r.Code)
}

I tried this out here: https://play.golang.org/p/faxQUm-2lr

Answer 3

Keep app and database code separate.

// Role belongs to app code, no compromises.

type Role struct {
    Id int64
    Code string
}

Model the database.

// Database has tables with columns.  

type Table struct {
    Name string
    Columns []string
}

var RoleTable = Table{
    Name: "roles",
    Columns: []string{
        "id",
        "code",
    },
}

Write code once to convert between model and database row.

// Database package needs to make it work.

// Write a model to database row. 
type Writer struct {
    Role
}

func (w *Writer) Write() []interface{} {
    return []interface{}{
        w.Role.Id,
        sql.NullString{
            Valid: len(w.Role.Code) > 0,
            String: w.Role.String,
        },
    }
}

// Read a database row into model. 
type Reader struct {
    Id int64
    Code sql.NullString
}

func (r *Reader) Scan(row *sql.Row) error {
    return row.Scan(
        &r.Id,
        &r.Code,
    )
}

func (r *Reader) Read() Role {
    return Role{
        Id: r.Id,
        Code: r.Code.String,
    }
}

Your schema is decoupled from app model. You can flatten structures like user contact details when saving or loading.

// Nested struct in app code. 
type User struct {
    TwitterProfile struct {
        Id string
        ScreenName string
    }
}

// Database row is normalized flat. 
var UserTable = Table{
    Name: "users",
    Columns: []string{
        "twitter_id",
        "twitter_screen_name",
    },
}

It's flexible. You can even scan join rows without intermediate structs.

type RowMux struct {
    vs []interface{}
}

func (mux *RowMux) Scan(vs ...interface{}) error {
    mux.vs = append(mux.vs, vs...)
    return nil
}

func (mux *RowMux) Mux(row *sql.Row) error {
    return row.Scan(mux.vs...)
}

// Scan join rows!
row := db.QueryRow(`
    SELECT users.*, roles.*
    FROM users
    JOIN roles ON users.id = roles.user_id
    WHERE users.twitter_id = "123"
`)
mux := &RowMux{}
userReader := &UserReader{}
userReader.Scan(mux)
roleReader := &RoleReader{}
roleReader.Scan(mux)
if err := mux.Mux(row); err != nil {
    panic(err)
}
user := userReader.Read()
role := roleReader.Read()