There several possible ways to approach this from writing code in your favorite language of choice using appropriate APIs to select data, transform it and then insert it into MongoDB.
You can also do it using SQL, MongoDB query language and the shell. One straightforward way is to select flat data via SQL, dump it into CSV file, import it into MongoDB and use aggregation framework to transform it into the format you want.
If you are lucky enough to use a database that supports arrays or other ways of grouping rows into single list types, then you can make a single select and turn it into JSON or MongoDB insert statement.
For these examples, I'm going to assume you want the format equivalent to a document for each city:
{
State:"Orissa",
City:{
Name:"Phulbani",
Area:[
"Phulbani","Phulbani2","Pokali","Madira"
]
}
}
Sample data in RDBMS:
asya=# select * from states;
id | name
----+---------------
1 | California
2 | New York
3 | Massachusetts
(3 rows)
asya=# select * from cities;
id | name | states_id
----+---------------+-----------
1 | Los Angeles | 1
2 | San Francisco | 1
3 | San Diego | 1
4 | New York | 2
5 | Brooklyn | 2
6 | Buffalo | 2
7 | Boston | 3
(7 rows)
asya=# select * from cityarea;
id | name | city_id
----+--------------------+---------
1 | Beacon Hill | 7
2 | Backbay | 7
3 | Brookline | 7
4 | Park Slope | 5
5 | Little Italy | 4
6 | SOHO | 4
7 | Harlem | 4
8 | West Village | 4
9 | SoMa | 2
10 | South Beach | 2
11 | Haight Ashbury | 2
12 | Cole Valley | 2
13 | Bunker Hill | 1
14 | Skid Row | 1
15 | Fashion District | 1
16 | Financial District | 1
(16 rows)
The easy way with arrays:
SELECT 'db.cities.insert({ state:"' || states.name || '", city: { name: "' || cities.name || '", areas : [ ' || array_to_string(array_agg('"' || cityarea.name || '"'),',') || ']}});'
FROM states JOIN cities ON (states.id=cities.states_id) LEFT OUTER JOIN cityarea ON (cities.id=cityarea.city_id) GROUP BY states.name, cities.name;
gives you output that can go straight into MongoDB shell:
db.cities.insert({ state:"California", city: { name: "Los Angeles", areas : [ "Financial District","Fashion District","Skid Row","Bunker Hill"]}});
db.cities.insert({ state:"California", city: { name: "San Diego", areas : [ ]}});
db.cities.insert({ state:"California", city: { name: "San Francisco", areas : [ "Haight Ashbury","South Beach","SoMa","Cole Valley"]}});
db.cities.insert({ state:"Massachusetts", city: { name: "Boston", areas : [ "Beacon Hill","Brookline","Backbay"]}});
db.cities.insert({ state:"New York", city: { name: "Brooklyn", areas : [ "Park Slope"]}});
db.cities.insert({ state:"New York", city: { name: "Buffalo", areas : [ ]}});
db.cities.insert({ state:"New York", city: { name: "New York", areas : [ "Little Italy","West Village","Harlem","SOHO"]}});
The longer way if you don't have support for array or list types is to select joined data:
asya=# SELECT states.name as state, cities.name as city, cityarea.name as area
FROM states JOIN cities ON (states.id=cities.states_id)
LEFT OUTER JOIN cityarea ON (cities.id=cityarea.city_id);
state | city | area
---------------+---------------+--------------------
California | Los Angeles | Financial District
California | Los Angeles | Fashion District
California | Los Angeles | Skid Row
California | Los Angeles | Bunker Hill
California | San Francisco | Cole Valley
California | San Francisco | Haight Ashbury
California | San Francisco | South Beach
California | San Francisco | SoMa
California | San Diego |
New York | New York | West Village
New York | New York | Harlem
New York | New York | SOHO
New York | New York | Little Italy
New York | Brooklyn | Park Slope
New York | Buffalo |
Massachusetts | Boston | Brookline
Massachusetts | Boston | Backbay
Massachusetts | Boston | Beacon Hill
(18 rows)
I used a left outer join on cityarea because in my sample data I had a city without any areas listed but I wanted to get all state, city pairs even if there was not an area listed for it.
You can dump this out interactively or via a command line (use appropriate syntax for your RDBMS). I'll do it interactively:
asya=# \a
Output format is unaligned.
asya=# \f
Field separator is "|".
asya=# \f ,
Field separator is ",".
asya=# \t
Showing only tuples.
asya=# \o dump.txt
asya=# SELECT states.name as state, cities.name as city, cityarea.name as area
FROM states JOIN cities ON (states.id=cities.states_id)
LEFT OUTER JOIN cityarea ON (cities.id=cityarea.city_id);
asya=# \q
I now have a comma separated file with state, city and area as the three fields. I can load it into MongoDB via mongoimport utility:
asya$ mongoimport -d sample -c tmpcities --type csv --fields state,city,area < dump.txt
connected to: 127.0.0.1
2014-08-05T07:41:36.744-0700 check 9 18
2014-08-05T07:41:36.744-0700 imported 18 objects
Now to transform to format I want, I use aggregation:
mongo sample
MongoDB shell version: 2.6.4
connecting to: sample1
> db.tmpcities.aggregate(
{$group:{_id:"$city", state:{$first:"$state"}, areas:{$push:"$area"}}},
{$project:{state:1,_id:0,city:{name:"$_id", areas:"$areas"}}},
{$out:'cities'})
> db.cities.find({},{_id:0})
{ "_id" : "Boston", "state" : "Massachusetts", "areas" : [ "Brookline", "Backbay", "Beacon Hill" ] }
{ "_id" : "New York", "state" : "New York", "areas" : [ "West Village", "Harlem", "SOHO", "Little Italy" ] }
{ "_id" : "Buffalo", "state" : "New York", "areas" : [ "" ] }
{ "_id" : "Brooklyn", "state" : "New York", "areas" : [ "Park Slope" ] }
{ "_id" : "San Diego", "state" : "California", "areas" : [ "" ] }
{ "_id" : "San Francisco", "state" : "California", "areas" : [ "Cole Valley", "Haight Ashbury", "South Beach", "SoMa" ] }
{ "_id" : "Los Angeles", "state" : "California", "areas" : [ "Financial District", "Fashion District", "Skid Row", "Bunker Hill" ] }
[ ]) not objects ({ }) for Cities and CityArea. – Gamophyllous