Use the Approximation Pattern

On this page

About this Task

Steps
Insert sample data
Implement the approximation pattern
Results
Learn More

Use the approximation pattern when you have values that change frequently, but users don't need to know precise values. Instead of updating values every time data changes, the approximation pattern updates data based on a larger granularity, which results in fewer updates and a lower application workload.

About this Task

The approximation pattern is useful when values don't need to be reported exactly. For example:

City population
Website visits
Airline travelers

The preceding measurements are typically useful when approximated. The application can save time and resources by updating stored values by hundreds or thousands, depending on the scale of the data.

Steps

In this example, an application displays population data for a city of roughly 40,000 people. Application users are primarily looking for overall trends and don't need to know the exact city population.

Insert sample data

db.population.insertOne( {
   city: "New Perth",
   population: 40000,
   date: ISODate("2022-09-15")
} )

Implement the approximation pattern

The actual population value changes many times in a single day. Rather than updating the population value with each change, use application logic to insert a new document every time the population changes by 100.

For example, your application logic might resemble the following:

let population = 40000
function updateStoredPopulation(curr_population, new_population) {
   let population_change = Math.abs(curr_population - new_population)
   if (population_change >= 100) {
      db.population.insertOne(
         {
            city: "New Perth",
            population: new_population,
            date: Date()
         }
      )
      population = new_population
   }
}

Note

The preceding example is only illustrative and does not use accurate syntax. To learn the correct syntax for your application, refer to your corresponding driver documentation.

Results

The preceding application logic might result in these documents:

db.population.insertMany( [
   {
      city: "New Perth",
      population: 40100,
      date: ISODate("2024-09-20")
   },
   {
      city: "New Perth",
      population: 40200,
      date: ISODate("2024-10-01")
   },
   {
      city: "New Perth",
      population: 40300,
      date: ISODate("2024-10-09")
   },
] )

Note

How the updated values are gathered depends on your scenario. In this example, updated population values can be gathered from census reports.

By updating the population with a granularity of 100, the approximation pattern reduces the number of updates to 1% of the updates that would be required to track individual population changes.

Users can see the population increasing over time, which meets their needs of seeing high-level trends.

Learn More

Back

Computed Data

Group Data