MongoDB , how to choose the right design? - mongodb

Question

I have to create a collection of document and I have a doubt about the right design.
Each document is an "identity"; each Identity has a list of "partner Data"; each partner data are defined by an ID and a set of Data.
One approach can be (1):
{
_id: ...
partners: [
{
id: partner1,
data: {
}
},
{
id: partner2,
data: {
}
},
]
}
Another approach can be (2)
{
_id: ...
partners: {
partner1: {
data: {
}
},
partner2: {
data: {
}
},
]
}
I prefer the first one, but considering that I could have million of these identities, which could be the most performed schema?
A typical query can be: "how many identities have partner with ID N".
With the second example, a query can be:
db.identities.find({partner.partnerName: {$exists:true}})
With first approach, how can I get this count?
The second solution is more easy to handle Server Side; each document will have a list where each KEY is the partner ID, so instead of scan all document, I can simply get partner data by key...
What do you think about these solutions? I prefer the first one but the second I think that is more "usable"...
Thanks

Answer 1

I prefer the first one, but considering that I could have million of these identities, wich could be the most performed schema?
If you going to have millions of identities, then both approaches
are not really scalable .
Each document in mongo has a size limit (16MB) (read about it here)
In case you are going to have really lot's of identities ,
the scalable approach would be to create a different collection,
only for the relations and partnership data.
Now , I also want you to consider how you treat "partnership",
if I'm a user and I got you on my partners list , will you see me as a partner on your list ?
In case we both see each other as partners , then mongo-db may not be the best solution. graph db's are more appropriate for dealing with relations of this type.
All solutions within mongo for two-ways relations will be built on double updates (Your id on my partner's list , My id on your partner's list).
(In SQL you could add an extra condition for joining but not in mongo) ,
so you don't need to save twice partnerships . (me and you , you and me)
just you and me.
Do you see where is going this way ?
If you need to go only in one way,
Then just create a second collection , "partnerships" ,
{
_id: should be uniqe,
user_id: 'your_id',
partner_id: 'his_id'
data: {} or just flatten the fields into the root object.
}
Please notice that you create a row for each partnership !
Then you could use $lookup in order to query for a user with all of his
partners .
something like:
db.getCollection('partners').aggregate([
{
$lookup: {
from: 'parterships',
localField: '_id',
foreignField: 'user_id',
as: 'partners'
}
},
{
$project: {
name: 1,
partners: 1,
num_partners: { $size: "$partners" }
}
}
])
Read more about the aggregation stages here.
In case you are not going to have lot's of partnership's Then please continue with
your first approach which is good .
The second approach will make most queries to this collection pretty weird and you will always have to write code in order to query this table .
It won't be "straight forward" mongo queries .