Principle of MongoDB geographic location index implementation

Geographic location index support is one of the highlights of MongoDB, which is also one of the reasons why Foursquare, the world's most popular lbs service, chooses MongoDB. We know that the general database index structure is B + tree, and how to convert the geographical location into a form of B + tree, which will be described below.

First, let's divide the entire map to be indexed into 16 × 16 squares, for example, (the lower left corner is coordinate 0, the upper right corner is coordinate 16, 16 ):

 

Data of [x, y] alone cannot be indexed. Therefore, when MongoDB creates an index, it will calculate a hash value that can be used for Indexing Based on the coordinates of the corresponding field, the value is geohash. The following uses the coordinates of [4, 6] on the map (the Red Cross position in the figure) as an example.

The first step is to divide the entire map into four equal-Size Blocks, for example:

After dividing the data into four blocks, we can define the values of these four blocks, as shown below (00 in the lower left, 01 in the upper left, 10 in the lower right, and 11 in the upper right ):

 

01	11
00	10

 

In this way, the geohash value of [4, 6] is currently 00

Then we cut each of the four small blocks as follows:

In this case, [4, 6] points are located in the upper right area and the value on the upper right is 11. In this way, the geohash value of [4, 6] points is changed to: 0011.

Continue to perform two splitting operations:

Finally, the geohash value of [4, 6] Is 00110100.

In this way, the points with similar points on the map can be converted into geohash values with the same prefix.

We can see that the accuracy of this geohash value is proportional to the number of Map Division times. The previous example divided the map four times. MongoDB is divided into 26 times by default. This value is controllable during index creation. The command to create a two-dimensional geographic location index is as follows:

DB. Map. ensureindex ({point: "2D"}, {min: 0, Max: 16, Bits: 4 })

The bits parameter is divided several times. The default value is 26.