PHP Mongo Aggregation仅返回_id

Question

I am trying to return a collection of messages grouped by in_reply_to field, I have this code:

$result = $this->db->Message->aggregate(
            array(
                array(
                    '$project' => array('message' => 1, 'in_reply_to'=> 1, 'to_user' => 1, 'from_user' => 1)
                ),
                array(
                    '$group' => array('_id' => '$in_reply_to'),
                ),
            )
        );
        print_r($result);exit;

the result is:

Array ( 
    [result] => Array ( 
        [0] => Array ( 
            [_id] => MongoId Object ( 
                [$id] => 53a03d43b3f7e236470041a8 
            ) 
        ) 
        [1] => Array ( 
            [_id] => MongoId Object ( 
                [$id] => 53a03cbdb3f7e2e8350041bb
            ) 
        ) 
   ) 
   [ok] => 1 
)

Ideally I'd like the entire Message object, but I did think that $project would be used to specify returns fields, even so, I dont get the fields I'm specifying.

Any help is greatly appreciated

Answer 1

In order to get all the messages in the thread you basically want to $push

$result = $this->db->Message->aggregate(
    array(
        array(
            '$group' => array(
                '_id' => '$in_reply_to',
                'messages' => array(
                    '$push' => array(
                        '_id' => '$_id',
                        'message' => '$message',
                        'to_user' => '$to_user',
                        'from_user' =>'$from_user'
                    )
                )
            )
        )
    )
);

MongoDB 2.6 you have the $$ROOT variable that shortens this:

$result = $this->db->Message->aggregate(
    array(
        array(
            '$group' => array(
                '_id' => '$in_reply_to',
                'messages' => array(
                    '$push' => '$$ROOT'
                )
            )
        )
    )
);

So that puts all of the related messages inside the "messages" array tied to that key.

Just as side note, while you can do this you may as well just sort the results by your "in_reply_to" field and process them that way looking for changes in the value to indicate a new thread.

Sorting with a find would be the fastest way to process, even if it does not conveniently put everything right under the one key.