What is the correct way to search by the key of a nested MongoDB document?

Question

MongoDB

Given:

{
   _id : 1,
   attrib : {
         a : {obj1},
         b : {obj2}
   }
}

Task: check for the existence of objects (a, b).
The method proposed in the manual is slow and does not use indexes:

db.coll.find({'attrib.a' : {$exists : 1} }) // медленно, индексы не используются

db.coll.find({attrib : {a : {$exists : 1}} }) // нигде не предложенный способ, который к тому же не работает ))

db.coll.ensureIndex({attrib : 1}) // Нет эффекта для поиску по индексу
db.coll.ensureIndex({'attrib.a' : 1}) // Нет эффекта для поиску по индексу

How to properly arrange indexes so that the search for the existence of keys (a, b) is fast?

Answer 1

directly with $exists: true apparently nothing will work.
But you can try to dodge in several ways:
1. Add a service field, which is indexed. This can also have a good effect on the size of the index, especially if a and b are large, then indexing them entirely for the sake of $exists is wasteful.
2. If a and b are always of the same type, you can try using a query like this db.coll.find({'attrib.a': {'$type': 3}})
At least according to explain, the index he uses.

Answer 2

The essence of the question is not entirely clear, if you mean getting all documents containing attrib.a. Then you can use sparse indexes:

> db.ttt.find(); // документы коллекции
{ "_id" : ObjectId("50377f098537d96af9ea28ae"), "attrib" : { "a" : { }, "b" : { } } }
{ "_id" : ObjectId("50377f0c8537d96af9ea28af"), "attrib" : { "b" : { } } }
{ "_id" : ObjectId("50377f0f8537d96af9ea28b0"), "attrib" : { "a" : { } } }
{ "_id" : ObjectId("50377f138537d96af9ea28b1"), "attrib" : { } }

> db.ttt.ensureIndex({'attrib.a':1}, {sparse : true}); // разряженный индекс
> db.ttt.find().sort({'attrib.a':1}); // в результате будут только документы содержащие attrib.a
{ "_id" : ObjectId("50377f0f8537d96af9ea28b0"), "attrib" : { "a" : { } } }
{ "_id" : ObjectId("50377f098537d96af9ea28ae"), "attrib" : { "a" : { }, "b" : { } } }

> db.ttt.find().hint({'attrib.a':1}); // явно указан индекс
{ "_id" : ObjectId("50377f0f8537d96af9ea28b0"), "attrib" : { "a" : { } } }
{ "_id" : ObjectId("50377f098537d96af9ea28ae"), "attrib" : { "a" : { }, "b" : { } } }

You can immediately select documents with a certain attrib.a value. If you need to check for the presence of objects in a particular document, then this option is most likely not suitable.