Helper to get single result from a M2M relation

[bookshelf]

Helper to get single result from a M2M relation

[tgriesser]

So this is probably what you want:

class MyModel extends Bookshelf.Model
  items: (id) ->
    @belongsToMany(Item)

new MyModel(id: 1).items().query('where', 'id', id).fetch().then (c) ->
  c.at(0)

The reason I didn't want to put finder methods in is because there's a million permutations of them, and once you start adding direct support for them the API can quickly become a bit of a mess.

By using query the way described above you can add them wherever you need... and by not doing query() but instead query('method', args*) you stay on the context of the current model/collection you're accessing rather than dropping down into the Knex builder.

Though I agree some way to fetch/return a single model from a collection could a good addition in this case (hadn't considered relations), I'll take a look into supporting it...

And yeah, a BaseModel is exactly what I'm thinking would be a good way to go... so you can build out the API you'd find most intuitive... it's a little more verbose the first time but it makes the library easier to wrap your head around.

Also, a get raw json as the result when what I would like is a Collection of Models.

Sorry... I'm not sure I follow... you do or do not want that?

[vjpr]

Sorry... I'm not sure I follow... you do or do not want that?

query().where() produces the json from the db. When I wanted a Model or a Collection of Models.

But you have covered this above:

you stay on the context of the current model/collection you're accessing rather than dropping down into the Knex builder.

[tgriesser]

Yeah, query('where', ... will do it.

[vjpr]

And yeah, a BaseModel is exactly what I'm thinking would be a good way to go... so you can build out the API you'd find most intuitive... it's a little more verbose the first time but it makes the library easier to wrap your head around.

I agree. The thing that drew me to Bookshelf was the fact that it was layered really nicely. In Sequelize and most ORMs if you find something you can't do you have to drop straight down to SQL, and extensibility is difficult.

[vjpr]

Yeah, query('where', ... will do it.

Would be nice if you could pass in a hash.

[tgriesser]

Would be nice if you could pass in a hash.

Interesting, so like query({where: {id: 1}, orWhere: {id: 2})

[vjpr]

Yep. That would be great.

Also, if a query is called on a collection I think it should always return the models as a collection of models.

If the user wants to run a raw query they should use Bookshelf.Knex.

[tgriesser]

Yeah... that's the idea, unless you call query() with no arguments which returns the underlying Knex builder, which I think I'll keep in there.

Take a look at the last commit for the object support.

[tgriesser]

So I think that I'm going to revisit the idea of fetching a single object from a collection at some point down the road... for now I'm going to have the object support for the query chain be the solution for this for now.