将文档的 MongoDB 字段合并为一个文档

Question

I'm using MongoDB 2.6.6

I have these documents in a MongoDB collection and here is an example:

{ ..., "field3" : { "one" : [ ISODate("2014-03-18T05:47:33Z"),ISODate("2014-06-02T20:00:25Z") ] }, ...}
{ ..., "field3" : { "two" : [ ISODate("2014-03-18T05:47:33Z"),ISODate("2014-06-02T20:00:25Z") ] }, ...}
{ ..., "field3" : { "three" : [ ISODate("2014-03-18T05:47:39Z"),ISODate("2014-03-19T20:18:38Z") ] }, ... }

I would like the merge these documents in one field. For an example, I would like the new result to be as follows:

{ "field3", : { "all" : [ ISODate("2014-03-18T05:47:39Z"),ISODate("2014-03-19T20:18:38Z"),......  ] },}

I'm just not sure any more how to have that result!

Answer 1

Doesn't really leave much to go on here but you can arguably get the kind of merged result with mapReduce:

db.collection.mapReduce(
  function() {
    var field = this.field3;

    Object.keys(field).forEach(function(key) {
      field[key].forEach(function(date) {
        emit( "field3", { "all": [date] } )
      });
    });
  },
  function (key,values) {

    var result  = { "all": [] };

    values.forEach(function(value) {
      value.all.forEach(function(date) {
        result.all.push( date );
      });
    });

    result.all.sort(function(a,b) { return a.valueOf()-b.valueOf() });

    return result;

  },
  { "out": { "inline": 1 } }
)

Which being mapReduce is not exactly in the same output format given it's own restrictions for doing things:

{
    "results" : [
            {
                    "_id" : "field3",
                    "value" : {
                            "all" : [
                                    ISODate("2014-03-18T05:47:33Z"),
                                    ISODate("2014-03-18T05:47:33Z"),
                                    ISODate("2014-03-18T05:47:39Z"),
                                    ISODate("2014-03-19T20:18:38Z"),
                                    ISODate("2014-06-02T20:00:25Z"),
                                    ISODate("2014-06-02T20:00:25Z")
                            ]
                    }
            }
    ],
    "timeMillis" : 86,
    "counts" : {
            "input" : 3,
            "emit" : 6,
            "reduce" : 1,
            "output" : 1
    },
    "ok" : 1
}

Since the aggregation here into a single document is fairly arbitrary you could pretty much argue that you simply take the same kind of approach in client code.

At any rate this is only going to be useful over a relatively small set of data with next to the same sort of restrictions on the client processing. More than the 16MB BSON limit for MongoDB, but certainly limited by memory to be consumed.

So I presume you would need to add a "query" argument but it's not really clear from your question. Either using mapReduce or your client code, you are basically going to need to follow this sort of process to "mash" the arrays together.

I would personally go with the client code here.