Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

MarshalJSON without having all objects in memory at once

Tags:

go

marshalling

I want to use json.Encoder to encode a large stream of data without loading all of it into memory at once.

// I want to marshal this
t := struct {
    Foo string

    // Bar is a stream of objects 
    // I don't want it all to be in memory at the same time.
    Bar chan string 
}{
    Foo: "Hello World",
    Bar: make(chan string),
}

// long stream of data
go func() {
    for _, x := range []string{"one", "two", "three"} {
        t.Bar <- x
    }
    close(t.Bar)
}()

I thought maybe the json package had this functionality build in, but that's not the case.

playground

// error: json: unsupported type: chan string
if err := json.NewEncoder(os.Stdout).Encode(&t); err != nil {
    log.Fatal(err)
}

I'm currently just building the json string myself.

playground

w := os.Stdout
w.WriteString(`{ "Foo": "` + t.Foo + `", "Bar": [`)

for x := range t.Bar {
    _ = json.NewEncoder(w).Encode(x)
    w.WriteString(`,`)
}

w.WriteString(`]}`)

Is there a better way to do this?

If the json.Marshaler was like this is would be trivial.

type Marshaler interface {
    MarshalJSON(io.Writer) error
}
like image 360
Ilia Choly Avatar asked Oct 22 '22 03:10

Ilia Choly


1 Answers

Unfortunately the encoding/json package doesn't have a way to do this yet. What you're doing now (manually) is the best way to do it, without modifying the built-in package.

If you were to patch encoding/json, you could modify the reflectValueQuoted function in encoding/json/encode.go

You would want to focus on the Array case (Slice has a fallthrough):

// Inside switch:
case reflect.Array:
    e.WriteByte('[')
    n := v.Len()
    for i := 0; i < n; i++ {
        if i > 0 {
            e.WriteByte(',')
        }
        e.reflectValue(v.Index(i))
    }
    e.WriteByte(']')

I'm assuming you'd want to treat channel the same way. It would look something like this:

// Inside switch:
case reflect.Chan:
    e.WriteByte('[')
    i := 0
    for {
        x, ok := v.Recv()
        if !ok {
            break
        }
        if i > 0 {
            e.WriteByte(',')
        }
        e.reflectValue(x)
        i++
    }
    e.WriteByte(']')

I haven't done much with channels in reflect, so the above may need other checks.

If you do end up going this route, you could always submit a patch.

like image 184
Luke Avatar answered Oct 24 '22 03:10

Luke