How can I make this object mapping more dry and reusable in Go?

Question

I have created an object mapping in Go that is not relational, it is very simple.

I have several structs that looks like this:

type Message struct {
    Id       int64
    Message  string
    ReplyTo  sql.NullInt64 `db:"reply_to"`
    FromId   int64         `db:"from_id"`
    ToId     int64         `db:"to_id"`
    IsActive bool          `db:"is_active"`
    SentTime int64         `db:"sent_time"`
    IsViewed bool          `db:"is_viewed"`

    Method   string `db:"-"`
    AppendTo int64  `db:"-"`
}

To create a new message I just run this function:

func New() *Message {
    return &Message{
        IsActive: true,
        SentTime: time.Now().Unix(),
        Method:   "new",
    }
}

And then I have a message_crud.go file for this struct that looks like this:

To find a message by a unique column (for example by id) I run this function:

func ByUnique(column string, value interface{}) (*Message, error) {

    query := fmt.Sprintf(`
        SELECT *
        FROM message
        WHERE %s = ?
        LIMIT 1;
    `, column)

    message := &Message{}
    err := sql.DB.QueryRowx(query, value).StructScan(message)
    if err != nil {
        return nil, err
    }
    return message, nil
}

And to save a message (insert or update in the database) I run this method:

func (this *Message) save() error {
    s := ""
    if this.Id == 0 {
        s = "INSERT INTO message SET %s;"
    } else {
        s = "UPDATE message SET %s WHERE id=:id;"
    }
    query := fmt.Sprintf(s, sql.PlaceholderPairs(this))

    nstmt, err := sql.DB.PrepareNamed(query)
    if err != nil {
        return err
    }

    res, err := nstmt.Exec(*this)
    if err != nil {
        return err
    }

    if this.Id == 0 {
        lastId, err := res.LastInsertId()
        if err != nil {
            return err
        }
        this.Id = lastId
    }

    return nil
}

The sql.PlaceholderPairs() function looks like this:

func PlaceholderPairs(i interface{}) string {

    s := ""
    val := reflect.ValueOf(i).Elem()
    count := val.NumField()

    for i := 0; i < count; i++ {
        typeField := val.Type().Field(i)
        tag := typeField.Tag

        fname := strings.ToLower(typeField.Name)

        if fname == "id" {
            continue
        }

        if t := tag.Get("db"); t == "-" {
            continue
        } else if t != "" {
            s += t + "=:" + t
        } else {
            s += fname + "=:" + fname
        }
        s += ", "
    }
    s = s[:len(s)-2]
    return s
}

But every time I create a new struct, for example a User struct I have to copy paste the "crud section" above and create a user_crud.go file and replace the words "Message" with "User", and the words "message" with "user". I repeat alot of code and it is not very dry. Is there something I could do to not repeat this code for things I would reuse? I always have a save() method, and always have a function ByUnique() where I can return a struct and search by a unique column.

In PHP this was easy because PHP is not statically typed.

Is this possible to do in Go?

Answer 1

You probably want to use an ORM. They eliminate a lot of the boilerplate code you're describing.

See this question for "What is an ORM?"

Here is a list of ORMs for go: https://github.com/avelino/awesome-go#orm

I have never used one myself, so I can't recommend any. The main reason is that an ORM takes a lot of control from the developer and introduces a non-negligible performance overhead. You need to see for yourself if they fit your use-case and/or if you are comfortable with the "magic" that's going on in those libraries.

Answer 2

I don't recommend doing this, i personally would prefer being explicit about scanning into structs and creating queries.

But if you really want to stick to reflection you could do:

func ByUnique(obj interface{}, column string, value interface{}) error {
    // ...
    return sql.DB.QueryRowx(query, value).StructScan(obj)
}

// Call with
message := &Message{}
ByUnique(message, ...)

And for your save:

type Identifiable interface {
    Id() int64
}

// Implement Identifiable for message, etc.

func Save(obj Identifiable) error {
    // ...
}

// Call with
Save(message)

The approach i use and would recommend to you:

type Redirect struct {
    ID        string
    URL       string
    CreatedAt time.Time
}

func FindByID(db *sql.DB, id string) (*Redirect, error) {
    var redirect Redirect

    err := db.QueryRow(
        `SELECT "id", "url", "created_at" FROM "redirect" WHERE "id" = $1`, id).
        Scan(&redirect.ID, &redirect.URL, &redirect.CreatedAt)

    switch {
    case err == sql.ErrNoRows:
        return nil, nil
    case err != nil:
        return nil, err
    }

    return &redirect, nil
}

func Save(db *sql.DB, redirect *Redirect) error {
    redirect.CreatedAt = time.Now()

    _, err := db.Exec(
        `INSERT INTO "redirect" ("id", "url", "created_at") VALUES ($1, $2, $3)`,
        redirect.ID, redirect.URL, redirect.CreatedAt)

    return err
}

This has the advantage of using the type system and mapping only things it should actually map.

Answer 3

Your ByUnique is almost generic already. Just pull out the piece that varies (the table and destination):

func ByUnique(table string, column string, value interface{}, dest interface{}) error {
    query := fmt.Sprintf(`
            SELECT *
            FROM %s
            WHERE %s = ?
            LIMIT 1;
        `, table, column)

    return sql.DB.QueryRowx(query, value).StructScan(dest)
}

func ByUniqueMessage(column string, value interface{}) (*Message, error) {
    message := &Message{}
    if err := ByUnique("message", column, value, &message); err != nil {
        return nil, err
    }
    return message, error
}

Your save is very similar. You just need to make a generic save function along the lines of:

func Save(table string, identifier int64, source interface{}) { ... }

Then inside of (*Message)save, you'd just call the general Save() function. Looks pretty straightforward.

Side notes: do not use this as the name of the object inside a method. See the link from @OneOfOne for more on that. And do not get obsessed about DRY. It is not a goal in itself. Go focuses on code being simple, clear, and reliable. Do not create something complicated and fragile just to avoid typing a simple line of error handling. This doesn't mean that you shouldn't extract duplicated code. It just means that in Go it is usually better to repeat simple code a little bit rather than create complicated code to avoid it.

---

EDIT: If you want to implement Save using an interface, that's no problem. Just create an Identifier interface.

type Ider interface {
    Id() int64
    SetId(newId int64)
}

func (msg *Message) Id() int64 {
    return msg.Id
}

func (msg *Message) SetId(newId int64) {
    msg.Id = newId
}

func Save(table string, source Ider) error {
    s := ""
    if source.Id() == 0 {
        s = fmt.Sprintf("INSERT INTO %s SET %%s;", table)
    } else {
        s = fmt.Sprintf("UPDATE %s SET %%s WHERE id=:id;", table)
    }
    query := fmt.Sprintf(s, sql.PlaceholderPairs(source))

    nstmt, err := sql.DB.PrepareNamed(query)
    if err != nil {
        return err
    }

    res, err := nstmt.Exec(source)
    if err != nil {
        return err
    }

    if source.Id() == 0 {
        lastId, err := res.LastInsertId()
        if err != nil {
            return err
        }
        source.SetId(lastId)
    }

    return nil
}

func (msg *Message) save() error {
    return Save("message", msg)
}

The one piece that might blow up with this is the call to Exec. I don't know what package you're using, and it's possible that Exec won't work correctly if you pass it an interface rather than the actual struct, but it probably will work. That said, I'd probably just pass the identifier rather than adding this overhead.