MongoDB map reduce count giving more results than a query - mongodb

Question

I have a collection users in Mongo and I execute this map reduce which I believe is the equivalent of a COUNT(*) GROUP BY origin:
> m = function() { for (i in this.membership) {
... emit( this.membership[i].platform_profile.origin, 1 );
... } }
function () {
for (i in this.membership) {
emit(this.membership[i].platform_profile.origin, 1);
}
}
> r = function( id, values ) { var result = 0;
... for ( var i = 0; i < values.length; i ++ ) { result += values[i]; }
... return result; }
function (id, values) {
var result = 0;
for (var i = 0; i < values.length; i++) {
result += values[i];
}
return result;
}
> db.users.mapReduce(m, r, {out : { inline: 1}});
{
"results" : [
{
"_id" : 0,
"value" : 15
},
{
"_id" : 1,
"value" : 449
},
...
}
But if I try to count how many documents have this field set to a specific value like 1, I get fewer results:
db.users.count({"membership.platform_profile.origin": 1});
424
What am I missing?

Answer 1

Are your count queries using a sparse index by any chance? My only guess there would be if some other query criteria resulted in documents absent from from index to be ignored from the count.
I recreated your schema with some fixture data and the results between map/reduce and simple count queries are in agreement:
db.users.drop();
var map = function() {
for (i in this.membership) {
emit(this.membership[i].platform_profile.origin, 1);
}
};
var reduce = function(id, values ) {
var result = 0;
for (var i = 0; i < values.length; i++) {
result += values[i];
}
return result;
}
var origins = {1: "a", 2: "b", 3: "c", 4: "d"};
for (var i = 0; i < 1000; ++i) {
var membership = [];
for (var o in origins) {
if (0 == i % o) {
membership.push({ platform_profile: { origin: origins[o] }});
}
}
db.users.save({ membership: membership });
}
db.users.mapReduce(map, reduce, {out: {inline: 1}}).results.forEach(function(result){
print(result["_id"] + ": " + result["value"]);
});
for (var o in origins) {
print(origins[o] + ": " + db.users.count({"membership.platform_profile.origin": origins[o]}));
}
Here's the output:
$ mongo --quiet mr_count.js
a: 1000
b: 500
c: 334
d: 250
a: 1000
b: 500
c: 334
d: 250

Answer 2

You can use the following map/reduce for the equivalent of COUNT(*) GROUP BY origin
Map/Reduce Functions :
map = function() {
if(!this.membership) return;
for (i in this.membership) {
if(!this.membership[i].platform_profile || !this.membership[i].platform_profile.origin) return;
emit(this.membership[i].platform_profile.origin, 1);
}
}
reduce = function(key, values) {
var count = 0;
for (v in values) {
count += values[v];
}
return count;
}
result = db.runCommand({
"mapreduce" : "users",
"map" : map,
"reduce" : reduce,
"out" : "users_count"
});

Answer 3

I had the same issue. I replaced x.length by Array.sum(x) in the reduce function (assuming you emit 1 in the map function) and it works. I agree x.length should work too, but I cannot explain why it does not.