Mongo aggregation combines $match steps, resulting in slow query - mongodb

Question

This question is a follow up to Query with $in and $nin doesn't use index. I've tried using aggregation to declare the order of steps.
db.assets.aggregate([
{
"$match": {
"tags": {
"$in": ["blah"]
}
}
},
{
"$match": {
"tags": {
"$nin": ["test"]
}
}
}
], {"explain": true})
You'd think Mongo would now understand that we want to filter by $in first. Well, you'd be surprised.
{
"stages" : [
{
"$cursor" : {
"query" : {
"$and" : [
{
"tags" : {
"$in" : [
"blah"
]
}
},
{
"tags" : {
"$nin" : [
"test"
]
}
}
]
},
"planError" : "InternalError No plan available to provide stats"
}
}
],
"ok" : 1
}
The planner doesn't even know what to do. It turns out it actually combines both $matches into one query, and then runs into the same problem as Query with $in and $nin doesn't use index, eventually returning the results in about 2-3 seconds (which corresponds to the 2331ms on the linked question).

Answer 1

It looks you can trick the aggregator by inserting an empty skip step:
db.assets.aggregate([
{
"$match": {
"tags": {
"$in": ["blah"]
}
}
},
{
"$skip": 0
},
{
"$match": {
"tags": {
"$nin": ["test"]
}
}
}
], {"explain": true})
With that, the planner will use the index and the results are returned immediately.
{
"stages" : [
{
"$cursor" : {
"query" : {
"tags" : {
"$in" : [
"blah"
]
}
},
"plan" : {
"cursor" : "BtreeCursor ",
"isMultiKey" : false,
"scanAndOrder" : false,
"indexBounds" : {
"tags" : [
[ "blah", "blah" ]
]
},
"allPlans" : [
{
"cursor" : "BtreeCursor ",
"isMultiKey" : false,
"scanAndOrder" : false,
"indexBounds" : {
"tags" : [
[ "blah", "blah" ]
]
}
}
]
}
}
},
{
"$skip" : NumberLong(0)
},
{
"$match" : {
"tags" : {
"$nin" : [
"test"
]
}
}
}
],
"ok" : 1
}