Golang和结构数组的结构

Question

I have the following structs defined for movies and TV shows:

type Movie struct {
    ID      string `json:"id"`
    Viewers int    `json:"count"`
}
type TVShow struct {
    ID      string `json:"id"`
    Season  int    `json:"season"`
    Episode int    `json:"episode"`
    Viewers int    `json:"count"`
}

Then I have the following structs that contain several movies or TV shows by country:

type Movies struct {
    PenultimateMonth map[string][]Movie
    LastMonth        map[string][]Movie
}
type TVShows struct {
    PenultimateMonth map[string][]TVShow
    LastMonth        map[string][]TVShow
}

Finally, I have a data struct that holds everything:

type Data struct {
    Movies  Movies
    Seasons Seasons
}

What I need to do is collect all IDs of all movies and TV shows from penultimate and last month.

I figured out that I can use reflection for that, but I only managed to iterate over each Data element individually instead of all:

func GetIDs(data *Data, country string) []string {
    var ids []string

    movies := reflect.ValueOf(data.Movies).Elem()
    tvShows := reflect.ValueOf(data.TVShows).Elem()

    for i := 0; i < movies.NumField(); i++ {
        moviesSubset := movies.Field(i).Interface().(map[string][]Movie)
        for _, movie := range moviesSubset[country] {
            ids = append(ids, movie.ID)
        }
    }

    for i := 0; i < tvShows.NumField(); i++ {
        tvShowsSubset := tvShows.Field(i).Interface().(map[string][]TVShow)
        for _, tvShow := range tvShowsSubset[country] {
            ids = append(ids, tvShow.ID)
        }
    }

    return ids
}

Is it possible to simplify the GetIDs function so that I don't need the two separate blocks for movies and TV shows, but only one to collect all IDs?

Answer 1

Use an interface:

type identifiable interface {
    GetID() string
}

type identifiables interface {
    GetIDs() []string
}

Which you can implement like:

func (m Movie) GetID() string { return m.ID }

And use to collect IDs polymorphically.

You could write it for each type, or make the maps store the interface and implement it once.

type identifiable interface {
    GetID() string
}

type identifiables interface {
    GetIDs() []string
}

func (m Movie) GetID() string { return m.ID }

type movies []Movie
type moviesByCountry map[string]movies

func (m movies) GetIDs() (ret []string) {
    for _, movie := range m {
        ret = append(ret, movie.GetID())
    }
    return
}

func (m moviesByCountry) GetIDs() (ret []string) {
    for _, slice := range m {
        ret = append(ret, slice.GetIDs()...)
    }
    return
}

func (d Data) GetCountryIDs(country string) []string {
    return gatherIDs(d.TVShows[country], d.Movies[country])
}

func gatherIDs(collections ...identifiables) (ret []string) {
    for _, collection := range collections {
        ret = append(ret, collection.GetIDs()...)
    }
    return
}

Here's a working playground example.

Not very efficient but simple and consistent IMHO. If that's a problem it can be optimized by passing in an accumulator slice, but I would suggest an interface with a private method of you go that route.

I think there's a reasonable case to be made for

type identifiableByCountry interface {
    GetCountryIDs(string) []string
}

since both Data and the map(s) could implement that.

If you prefer working with lambdas, you could also use something like:

type collection interface {
    walk(func(identifiable))
}

and implement that on the collection types, using it as

c.walk(func(o identifiable) {
    ids = append(ids, o.GetID())
})