Scan() does not populate has-many relationships

[naltimari]

In context of the 'has-many' example, if you do:

users := make([]User, 0)

db.NewSelect().
    Model((*User)(nil)).
    Relation("Profiles").
    Scan(ctx, &users)

users[k].Profiles will always be nil, regardless of the related Profiles they might have.

I don't think this is by design, as the documentation states that you can pass a struct as a 'destination', and no mention about the above behaviour.

[l-farrell]

I am also experiencing this. I think I found what is happening:

The value of model used here to find joins:

https://github.com/uptrace/bun/blob/master/query_select.go#L877

Will not have any, because its derived from dst earlier:

https://github.com/uptrace/bun/blob/master/query_select.go#L849

and getModel doesn't copy across the joins when dst is non-zero:

https://github.com/uptrace/bun/blob/master/query_base.go#L188

[mehow-ves]

This works for me, bun 1.1.8

// models.go
type Match struct {
	bun.BaseModel `bun:"table:match,alias:m"`

	Match_id      int       `json:"id" bun:",pk,autoincrement"`
	Creator       int       `json:"creator"`
}

type User struct {
	bun.BaseModel `bun:"table:user,alias:u"`

	User_id  int      `json:"id" bun:",pk,autoincrement"`
	Username string   `json:"username"`
	Matches  []*Match `json:"matches" bun:"rel:has-many,join:user_id=creator"`
}

// schema.resolvers.go
var user []*model.User
r.DB.NewSelect().Model(&user).Relation("Matches", func(q *bun.SelectQuery) *bun.SelectQuery {
     return q.Column("match_id", "creator")
}).Scan(ctx)

Although without a data loader it is inefficient as it executes a separate query for every Match for every User.

[amamrenko]

any update on that issue?

[steve-hb]

Same here, this behaviour took me about two days to figure out - imposter syndrome included.

[ijt]

I've been bitten by this as well, ended up doing crappy workarounds until I figured out what was wrong.

[RobertV17]

Hi all! I faced a similar issue but found a solution by passing only the context to the .Scan() method without the model variable. It wasn’t obvious at first, but here's an example (I used version v1.2.3):

type Account struct {
	bun.BaseModel `bun:"table:account"`
	ID            int64         `bun:"id,pk,autoincrement"`
	Username      string        `bun:"username,notnull"`
	Email         string        `bun:"email,notnull"`
	Password      string        `bun:"password,notnull"`
	CreatedAt     time.Time     `bun:"created_at,nullzero,default:current_timestamp"`
	PayAccounts   []*PayAccount `bun:"rel:has-many,join:id=account_id"`
}

type PayAccount struct {
	bun.BaseModel `bun:"table:pay_account"`
	ID            int64     `bun:"id,pk,autoincrement"`
	Username      string    `bun:"username,notnull"`
	Email         string    `bun:"email,notnull"`
	Password      string    `bun:"password,notnull"`
	AccountId     int64     `bun:"account_id,notnull"`
	CreatedAt     time.Time `bun:"created_at,nullzero,default:current_timestamp"`
}

func main() {
	postgresURL := &url.URL{
		Scheme: "postgres",
		User:   url.UserPassword("service", "service-pass"),
		Host:   fmt.Sprintf("%s:%s", "localhost", "5433"),
		Path:   "service-db",
	}
	connString := postgresURL.String()

	config, err := pgx.ParseConfig(connString)
	if err != nil {
		log.Panic().Err(err).Msg("Failed to parse PostgreSQL config")
	}

	sqldb := stdlib.OpenDB(*config)

	err = sqldb.Ping()

	if err != nil {
		log.Panic().Err(err).Msg("Failed to ping PostgreSQL")
	}

	db := bun.NewDB(sqldb, pgdialect.New())

	var accounts []*Account

	q := db.NewSelect().Model(&accounts).Relation("PayAccounts")

	err = q.Scan(context.Background())
	fmt.Println(accounts)
}