MongoDB Map减少在大数据量中返回意外结果

Question

I am new to PHP as well as mongo DB and i have a data set of 80000 records and this is a local deployment.

My Data Structure is simple:

(
    [_id] => MongoId Object
        (
            [$id] => 53c146aebc7d867d058b94b3
        )

    [name] => Mark
    [txnType] => Borrowed
    [amount] => 5876
)

I am running a Map Reduce Job as defined below:

$map = new MongoCode("function ()
{
    { 
        emit({name:this.name,type:this.txnType},this.amount);
    }
}");
$reduce = new MongoCode("
    function (key, values)
    {
        var total=0;
        var count=0;
        for (var i in values) { 
            if (!isNaN(values[i])) {
                total+=values[i];
            };
            count++;
        }
        return {total:total, count:count};
    }
    ");

$sales =  $db->command(array(
    "mapreduce" => "data", 
    "map" => $map,
    "reduce" => $reduce,
    "out" => "sales"
    ));

The Concept is basically that there are 4 guys who may have transactions of type Borrowed, Sold, Purchase and Lent. Each record representing a txn.

I want to just create a data pivot getting the data as:

Name : Type : Total Amount : Count of Txns

Some how the data that is propping up is messed up. The counts when added up should add up to 80000, but instead its adding up to only 216.

I am not able to understand why this is happening.. Can anyone please help me. where am i going wrong and what to correct.

My need is to basically draw up analytic for the transaction.

Answer 1

The problem is that your emit is the outputting the same format as your reduce.

Here is what you emit for value:

this.amount

Here is what you return from reduce:

return {total:total, count:count};

In order for reduce to work correctly when it re-reduces (remember, reduce may be called zero, once or multiple times on the same key value) you must emit this format:

emit({name:this.name,type:this.txnType},{ total: this.amount, count: 1} );

And therefore your reduce function should now be:

    var total=0;
    var count=0;
    for (var i in values) { 
        if (!isNaN(values.total[i])) {
            total+=values.total[i];
        };
        count+=values.count;
    }
    return {total:total, count:count};

The two most important rules of mapReduce in MongoDB :

1. emit value in exactly the same format as your reduce function returns

2. structure reduce so that it can be called zero, once or multiple times for each key

Note that you can perform the same aggregation much more efficiently and faster with Aggregation Framework like so:

db.collection.aggregate( {$group: 
    { _id : {name: "$name", type: "$txnType"},
      total: {$sum: "$amount"},
      count: {$sum: 1}
    }
}