Platinum Partner
news,architects,nosql,architecture,mongodb

Index Sub-Documents in MongoDB

Let’s say I have a collection of documents that all kind of look like this:

> db.foos.find().pretty()
{
    "_id" : ObjectId("511fe286777e76dfdddbf440"),
    "foo" : "bar",
    "timestamp" : ISODate("2013-02-16T20:02:39.417Z")
}

Then, all of a sudden I get a requirement handed to me that this collection needs to be searched for all documents with a timestamp within a given hour, on a given month, within a given year. Writing a query for that is going be a bit gnarly, so we create a sub-document in our documents that makes this query easier.

{
    "_id" : ObjectId("511fe286777e76dfdddbf440"),
    "foo" : "bar",
    "timestamp" : ISODate("2013-02-16T20:02:39.417Z"),
    "timefields" : {
        "y" : 2013,
        "mo" : 2,
        "d" : 16,
        "h" : 20,
        "mi" : 02,      
    }
}

So now my query will look like the following and all will be happy in the world. And yes, I could have put each of these elements at the top level, but bear with me for demonstration purposes…

> db.foos.find({"timefields.y":2013, "timefields.mo": 2, "timefields.h":20}).pretty()
{
    "_id" : ObjectId("511fe286777e76dfdddbf440"),
    "foo" : "bar",
    "timestamp" : ISODate("2013-02-16T20:02:39.417Z"),
    "timefields" : {
        "y" : 2013,
        "mo" : 2,
        "d" : 16,
        "h" : 20,
        "mi" : 2
    }
}

Ok, but what about that index that was (hypothetically) on timestamp? Because we are searching other fields we aren’t using indexes in our query and the search is happening in linear time. The horror!

Well, mongo is flexible if it isn’t anything else. You can just as easily index fields in a sub-document like you can any other field. For instance:

> var indices = { "timefields.h" : 1,
... "timefields.y" : 1,
... "timefields.mo" : 1,
... "timefields.d" : 1,
... "timefields.m" : 1}
> db.foos.ensureIndex(indices)

> db.foos.getIndices()
[
    ...
    {
        "v" : 1,
        "key" : {
            "timefields.h" : 1,
            "timefields.y" : 1,
            "timefields.mo" : 1,
            "timefields.d" : 1,
            "timefields.m" : 1
        },
        "ns" : "test.foos",
        "name" : "timefields.h_1_timefields.y_1_timefields.mo_1_timefields.d_1_timefields.m_1"
    }
]

And now when I run explain on the cursor using the same query I performed before, you can now see those indices are being used:

> db.foos.find({"timefields.y":2013, "timefields.mo": 2, "timefields.h":20}).pretty().explain()
{
    ...
    "indexBounds" : {
        "timefields.h" : [
            [
                20,
                20
            ]
        ],
        "timefields.y" : [
            [
                2013,
                2013
            ]
        ],
        "timefields.mo" : [
            [
                2,
                2
            ]
        ],
        "timefields.d" : [
            [
                {
                    "$minElement" : 1
                },
                {
                    "$maxElement" : 1
                }
            ]
        ],
        "timefields.m" : [
            [
                {
                    "$minElement" : 1
                },
                {
                    "$maxElement" : 1
                }
            ]
        ]
    },
}


 

Published at DZone with permission of {{ articles[0].authors[0].realName }}, DZone MVB. (source)

Opinions expressed by DZone contributors are their own.

{{ tag }}, {{tag}},

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}
{{ parent.authors[0].realName || parent.author}}

{{ parent.authors[0].tagline || parent.tagline }}

{{ parent.views }} ViewsClicks
Tweet

{{parent.nComments}}