mongodb fast tags query - mongodb

Question

I have a very large collection ( more than 800k ) and I need to implement a query for auto-complete ( based on word beginnings only ) functionality based on tags. my documents look like this:
{
"_id": "theid",
"somefield": "some value",
"tags": [
{
"name": "abc tag1",
"vote": 5
},
{
"name": "hij tag2",
"vote": 22
},
{
"name": "abc tag3",
"vote": 5
},
{
"name": "hij tag4",
"vote": 77
}
]
}
if for example my query would be for all tags that start with "ab" and has a "somefield" that is "some value" the result would be "abc tag1","abc tag3" ( only names ).
I care about the speed of the queries much more than the speed of the inserts and updates.
I assume that the aggregation framework would be the right way to go here, but what would be the best pipeline and indexes for very fast querying ?
the documents are not 'tag' documents they are documents representing a client object, they contain much more data fields that I left out for simplicity, each client has several tags and another field ( I changed its name so it wont be confused with the tags array ). I need to get a set without duplicates of all tags that a group of clients have.

Answer 1

Your document structure doesn't make sense - I'm assuming tags is an array and not an object. Try queries like this
db.tags.find({ "somefield" : "some value", "tags.name" : /^abc/ })
with an index on { "maintag" : 1, "tags.name" : 1 }. MongoDB optimizes left-anchored regex queries into range queries, which can be fulfilled efficiently using an index (see the $regex docs).
You can get just the tags from this document structure using an aggregation pipeline:
db.tags.aggregate([
{ "$match" : { "somefield" : "some value", "tags.name" : /^abc/ } },
{ "$unwind" : "$tags" },
{ "$match" : { "tags.name" : /^abc/ } },
{ "$project" : { "_id" : 0, "tag_name" : "$tags.name" } }
])
Index only helps for first $match, so same indexes for the pipeline as for the query.