Mysql to MongoDB and Query Equivalent to this schema - mysql

Question

I have a complex location database with the following schema :
table States
id : INT PK AutoIncrement
name : VarChar 50 UNIQUE
table Counties
id: INT PK AutoIncrement
stateID : INT ForeignKey ->States(id)
name : VARCHAR(50)
table Towns :
id: INT PK AutoIncrement
stateID : INT ForeignKey ->States(id)
countyID : INT ForeignKey ->Counties(id)
name : VARCHAR(50)
table listings
id : INT PK autoincrement
name: varchar(50)
stateID: INT
countyID: INT
townID: INT
When I want to display some statistic data about geographical repartition in a tree form like this :
state1 (105 results)
county 1 (50 results)
county 2 (55 results)
Town 1 ( 20 results)_
Town 2 ( 35 results)
state2 (200 results)
ect...
In mySQL I would had done this kind of queries :
**1st level : **
select count(*) as nb, S.namem, S.id as stateID from listings L INNER JOIN States S ON S.id=L.stateID GROUP BY S.id;
**2d level : **
foreach(results as $result){
$sql = "select count(*) as nb, from listings L INNER JOIN Counties C ON C.id=L.countyID WHERE L.stateID=".$result['stateID'];
});
and so on...There is a way to do that in a unique long query in MySQL too.
This is a trivial query and it is very fast on a SSD disk in Mysql.
I am starting to learn mongoDB and I want to know what kindof schema I should use to store my location data to optimize this $count() and $group() operations.
And which mongo query would do the job?

Answer 1

Store the documents with a structure like the listings table:
{
"name" : "listing0",
"state" : "Maryland",
"county" : "Washington",
"town" : "Faketown"
}
Then just find the number of listings per (state, country, town) triple with the aggregation pipeline
> db.listings.aggregate([
// hopefully an initial match stage to select a subset of search results or something
{ "$group" : { "_id" : { "state" : "$state", "county" : "$county", "town" : "$town" }, "count" : { "$sum" : 1 } } }
])
From here you can compute the numbers for the higher level of the tree by iterating over the result cursor, or you can run analogous pipelines to compute the numbers at the higher level of the tree. For example, for the county numbers in a specific state
> db.listings.aggregate([
// hopefully an initial match stage to select a subset of search results or something
{ "$match" : { "state" : "Oregon" } },
{ "$group" : { "_id" : { "state" : "$state", "county" : "$county" }, "count" : { "$sum" : 1 } } }
])