Hi !

We have a collection which is roughly based on a schema similar to :

{
  _id : GUID
  StartTime: DateTime
  // Other Information
  Counters: [  
    {
         Name : string,
        // Others informations...
    }
   ]
}

The majority of our aggregate query starts with a $match phase which is always similar to:

db.collections.aggregate([
{
  $match : 
  {
      "Counters.Name" : 'Name',
      "StartTime": {$gte : ISODate(....), $lte: ISODate(....)}
  }
}
])

And I have the two following indexes on the collection :

{'StartTime: 1}
{'Counters.Name' : 1, 'StartTime' : 1}

I've noticed two different behavior depending on the filters that are applied to the date.

1) If I'm only using $gte: ISODate() the Counters.Name_1_StartTime_1 index gets used.
2) If I'm using the $gte and $lte the StartTime_1 index get used.

Logically, it should use the Counters.Name_1_StartTime_1 one as it would restrict much more the scope of documents.

I've also tried with an $elemMatch and can't get the bounded query to use the one containing the Counters.Name and StartDate