Wakatta!

Like Eureka!, only cooler

Seven Databases in Seven Weeks MongoDB Day 1

It has been … a little over a week since I closed with HBase so it is time to move to the next database in the program: MongoDB. It is a so-called document oriented database, relying on Javascript and JSON like Riak but, MongoDB, unlike Riak, has a built-in support for querying, and can also perform partial updates.

Fittingly, this first day is about CRUD and queries.

CRUD in MongoDB

Unlike Riak, MongoDB makes a difference between Create and Update operations. It is not possible to insert documents with the same id twice in the same collection:

1
2
3
4
5

> db.test.insert( { _id: 'one', value: 1 } )
> db.test.find()
{ "_id" : "one", "value" : 1 }
> db.test.insert( { _id: 'one', value: 2 } )
E11000 duplicate key error index: book.test.$_id_  dup key: { : "one" }

Like in SQL, the Read, Update and Delete operations all operate on the result of a search. And because the criteria query for a search is just a JSON document, it is possible to reuse it in different contexts. Combined with JavaScript as a query/shell language, this makes a very flexible combination.

Reaching into objects

As documents are naturally nested, it is important to know how to refer to nested attributes and how to use them in queries.

Using the JSON nesting notation is possible but does not have the expected semantic (unless you change your expectations):

1
2

> db.towns.find( { mayor: { name: 'Sam Adams' } }, {name: 1} )
>

The query above returns nothing, despite the fact that there is a town whose mayor is named ‘Sam Adams’. The problem here is that this specific notation will match documents in the collection whose mayor attribute has the exact value { name: 'Sam Adams' }, rather than merely those whose mayor attribute has a sub-attribute name with the Sam Adams value (and possible other sub-attributes, a possibility that the first query denies).

The correct form for the query is

1
2
3

> db.towns.find( { 'mayor.name' : 'Sam Adams' }, {name: 1})
{ "_id" : ObjectId("4eed55b0ef971f5317e68e91"), "name" : "Portland" }
>

This documentation from the official website was very useful to me in clarifying this aspect.

The $not operator

I first got curious about this operator because the example from the book did not work. It generated an error in MongoDB 2.0.2 (the latest at the time of writing).

1
2
3
4
5
6
7
8

db.countries.find(
    {
        'exports.foods' : {
            $not : { name : 'burgers' }
        }
    },
    { _id : 0, name : 1 }
)

The error is

1

error: { "$err" : "invalid use of $not", "code" : 13034 }

From the source code, it seems $not cannot be applied to a (implicit) equal operator. So the book code should be using $ne:

1

db.countries.find( { 'exports.foods.name': { $ne: 'burgers' } } )

It does not help that the online shell (accessible from the ‘Try it out’ link on http://www.mongodb.org/) does not return an error for the book query. It does not return anything at all, so it is still not correct.

Still, even with 2.0.2, this query:

1
2

> db.countries.find( {$not: { name: 'United States' } } )
>

is accepted but does not return every country but US. Yet again it returns nothing.

As far as I can tell, this query is actually incorrect (the $not operator should be applied to an operator, not a match), so its value is irrelevant, but it is worrying that the parser allows this code.

The general problem is that the semantic of $not is not exactly the same as the similar operator from logic. So basic logic reasoning cannot help.

I suppose this is just something to keep in mind: be wary of using $not.

Exercises

The online MongoDB documentation

is here.

Query using regular expressions

As explained here, MongoDB supports two notations: a direct JavaScript regexp object using /.../ and slightly more verbose but more flexible $regexp operator. The syntax for the regular expression in both cases is the same.

MongoDB drivers

The list of drivers.

Print a JSON document containing { "hello" : "world" }

The tojson function can display (and pretty-print) JSON documents:

1

tojson({ "hello": "world" });

Find a town by case insensitive regular expression search

This would be the i option to the regular expression:

1

db.towns.find({name: /new/i})

Find all cities whose names contain an ‘e’, and are famous for food or beer.

A good use for the $in operator:

1
2
3
4

db.towns.find(
    { name : /e/,
      famous_for :
      { $in: ['food', 'beer'] } } )

or the equivalent but verbose:

1
2
3
4

db.towns.find(
    { name: /e/,
      $or: [{famous_for : 'food'},
            {famous_for: 'beer'} ] } )

Find all countries that do not export tasty bacon

To find countries that export tasty bacon, one has to use $elemMatch. It should not come as a surprise that it is used here as well, but with the dreaded $not:

1
2
3
4
5
6

db.countries.find(
    {'exports.foods':
     { $not:
       {$elemMatch:
        { name: 'bacon',
          tasty: true } } } }, {name: 1} )

produces

1
2

{ "_id" : "ca", "name" : "Canada" }
{ "_id" : "mx", "name" : "Mexico" }

(I had not deleted Canada from the database).

The seemingly equivalent

1
2
3
4
5
6

db.countries.find(
    { $not:
      {'exports.foods':
       { $elemMatch:
         { name: 'bacon',
           tasty: true } } } })

will not work. It does not return anything. See above my findings on $not.

Create a database blogger with a collection of articles

A new database:

Creating a new database is extremely easy: just name it when you launch the mongo shell command:

1

mongo blogger

A new collection:

Like for the database, a collection is created just by inserting something in it:

1
2
3
4

db.articles.insert(
    { name: 'Frederic', email: 'my@email.org',
      creation: new Date(),
      test: 'Neque porro quisquam est qui dolorem ipsum quia dolor sit amet, consectetur, adipisci velit...' } )

Update the article with an array of comments

With the operator $set, I can specify just what I need to update, not the whole document as in Riak:

1
2
3
4
5
6
7

db.articles.update(
    {"_id" : ObjectId("4ef3c78c7d67f191b06653e4") },
    {$set: { comments:
             [ {author: 'Alice', text: 'Me too'},
               {author: 'Bob', text: 'Me three'},
               {author: 'Carol', text: 'I know all your secret' } ]
           } } )

I can check the inserted value with db.articles.findOne() (without arguments it returns the first element in the collection, and pretty prints it):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

{
        "_id" : ObjectId("4ef3c78c7d67f191b06653e4"),
        "comments" : [
                {
                        "author" : "Alice",
                        "text" : "Me too"
                },
                {
                        "author" : "Bob",
                        "text" : "Me three"
                },
                {
                        "author" : "Carol",
                        "text" : "I know all your secrets"
                }
        ],
        "creation" : ISODate("2011-12-23T00:13:00.636Z"),
        "email" : "my@email.org",
        "name" : "Frederic",
        "test" : "Neque porro quisquam est qui dolorem ipsum quia dolor sit amet, consectetur, adipisci velit..."
}

Tomorrow will cover indexing, and more advanced uses of the data (such as MapReduce).

Comments