MongoDB/PyMongo: How to use dot notation in a Map

Question

I'm trying to count how many records I have located in each zip code.
In my MongoDB, zip code is embedded; using dot notation, it's located at a.res.z (a for address, res for residential, z for zip). For example, this works just fine:

db.NY.count({'a.res.z' : '14120'})

But when I try the map function (in python, because I'm using PyMongo):

map = Code("function () {"
           "    emit(this.a.res.z, 1);"
           "}")

I get this error when I call mapreduce:

pymongo.errors.OperationFailure: db assertion failure, assertion: 'map invoke failed: JS Error: TypeError: this.a has no properties nofile_b:0', assertionCode: 9014

Dot notation works at the top level - e.g. one dot - but I can't get it to work on embeddeds. What's the secret?

Answer 1

This error means that at least one object you're map-reducing is missing the res field of its a. See:

> db.NY.find({}, {_id: 0})                                             
{ "a" : { "res" : { "z" : 10011 } }, "name" : "alice" }
{ "a" : { "res" : { "z" : 10011 } }, "name" : "bob" }
{ "a" : { "res" : { "z" : 10012 } }, "name" : "carol" }
> m                                                                    
function () {
    emit(this.a.res.z, 1);
}
> r                                                                    
function (key, values) {
    var v = 0;
    values.forEach(function (obj) {v += obj;});
    return v;
}
> db.runCommand({mapreduce: "NY", map: m, reduce: r, out: {inline: 1}})
{
    "results" : [
        {
            "_id" : 10011,
            "value" : 2
        },
        {
            "_id" : 10012,
            "value" : 1
        }
    ],
    "timeMillis" : 0,
    "counts" : {
        "input" : 3,
        "emit" : 3,
        "output" : 2
    },
    "ok" : 1
}
> db.NY.insert({a: {}, name: "empty"})                                                       
> db.runCommand({mapreduce: "NY", map: m, reduce: r, out: {inline: 1}})
{
    "assertion" : "map invoke failed: JS Error: TypeError: this.a.res has no properties nofile_b:1",
    "assertionCode" : 9014,
    "errmsg" : "db assertion failure",
    "ok" : 0
}

You can use a query argument to map-reduce to operate only on those that have the fields that you want:

> db.runCommand({mapreduce: "NY", map: m, reduce: r, out: {inline: 1},
                 query: {"a.res.z": {$exists: true}}})
{
    "results" : [
        {
            "_id" : 10011,
            "value" : 2
        },
        {
            "_id" : 10012,
            "value" : 1
        }
    ],
    "timeMillis" : 1,
    "counts" : {
        "input" : 3,
        "emit" : 3,
        "output" : 2
    },
    "ok" : 1
}

To use the query argument from PyMongo, you can set it as a keyword argument to map_reduce(...)