select documents with sub arrays that match some critieria - mongodb

Question

I have a collections with documents such as:
{
_id: "1234",
_class: "com.acme.classA",
a_collection: [
{
otherdata: 'somedata',
type: 'a'
},
{
otherdata: 'bar',
type: 'a'
},
{
otherdata: 'foo',
type: 'b'
}
],
lastChange: ISODate("2014-08-17T22:25:48.918Z")
}
I want to find all document by id and a subset of the sub array. for example I want to find all documents with id "1234" and a_collection.type is 'a' giving this result:
{
_id: "1234",
_class: "com.acme.classA",
a_collection: [
{
otherdata: 'somedata',
type: 'a'
},
{
otherdata: 'bar',
type: 'a'
}
],
lastChange: ISODate("2014-08-17T22:25:48.918Z")
}
I have tried this :
db.collection_name.aggregate({
$match: {
'a_collection.type': 'a'
}
},
{
$unwind: "$a_collection"
},
{
$match: {
"a_collection.type": 'a'
}
},
{
$group: {
_id: "$_id",
a_collection: {
$addToSet: "$a_collection"
},
}
}).pretty()
but this doesnt return other properties ( such as 'lastChange' )
what is the correct way to do this ?

Answer 1

Are you using PHP?
And is this the only way you can get the "text"?
maybe you can rewrite it that it is like an JSON element.
something like that:
{
"_id": "1234",
"_class": "com.acme.classA",
"a_collection": [
{
"otherdata": "somedata",
"type": "a"
},
{
"otherdata": "bar",
"type": "a"
},
{
"otherdata": "foo",
"type": "b"
}
]
}
Then you can use the json_decode() function from PHP to make an array and then you can search and return only the needed data.
Edit: I read read false. do you search for a funktion like this?
db.inventory.find( {
$or: [ { _id: "1234" }, { 'a_collection.type': 'a' }]
} )
[Here][1] I found the code ;) [1]: http://docs.mongodb.org/manual/tutorial/query-documents/

Answer 2

this is the correct query:
db.collection_name.aggregate({
$match: {
'a_collection.type': 'a'
}
},
{
$unwind: "$a_collection"
},
{
$match: {
"a_collection.type": 'a'
}
},
{
$group: {
_id: "$_id",
a_collection: {
$addToSet: "$a_collection"
},
lastChange : { $first : "$lastChange" }
}
}).pretty()

Answer 3

Something is very strange about your desired query (and your pipelines). First of all, _id is a reserved field with a unique index on it. The result of finding all documents with _id = "1234" can only be 0 or 1 documents. Second, to find documents with a_collection.type = "a" for some element of the array a_collection, you don't need the aggregation framework. You just need a find query:
> db.test.find({ "a_collection.type" : "a" })
So all the work here appears to be winnowing the subarray of one document down to just those elements with a_collection.type = "a". Why do you have these objects in the same document if most of what you do is split them up and eliminate some to find a result set? How common and how truly necessary is it to harvest just the array elements with a_collection.type = "a"? Perhaps you want to model your data differently so a query like
> db.test.find({ <some condition>, "a_collection.type" : "a" })
returns you the correct documents. I can't say how you can do it best with the given information, but I can say that your current approach strongly suggests revision is needed (and I'm happy to help with suggestions if you include further information or post a new question).

Answer 4

I would agree with the answer you have submitted yourself, but for that in MongoDB 2.6 and greater there is a better way to do this with $map and $setDifference. Which wer both introduced at that version. But where available, this is much faster in the approach:
db.collection.aggregate([
{ "$match": { "a_collection.type": "a" } },
{ "$project": {
"$setDifference": [
{ "$map": [
"input": "$a_collection",
"as": "el",
"in": {
"$cond": [
{ "$eq": [ "$$el.type", "a" ] },
"$$el",
false
]
}
]},
[false]
]
}}
])
So that has no "group" or initial "unwind" which both can be costly options, along with the $match stage. So MongoDB 2.6 does it better.