Updating multiple database records (with independent data) in a single query

In the world of SQL Databases, it is pretty easy to update one or many records in a table, you just need to execute an UPDATE query with some column-value bindings and some filters if needed, this will help you either update single record, or update many records with the same data.

But what if we want to update many records with independent data? šŸ¤”

1. The Project

Letā€™s imagine we have a project where we keep track of the following customer data:

  • Name
  • Birthday

And our project as an API that allows us to perform basic CRUD operations over such data, but letā€™s focus on an easy to consume update endpoint:

POST /api/customers/{id}
Content-Type: application/json

{
  "name": "<Insert name>",
  "birthday": "<Insert birthday>"
}

When implementing this server-side, we could use the following SQL query to update our record:

UPDATE "customers"
SET "name" = :name, "birthday" = :birthday
WHERE "id" = :id

Where :id would be extracted from the endpoint route, and :name and :birthday from the request body.

2. The issue

With our imaginary project deployed, there comes a necessity to update the information of many customers in a single call, and while the end client could just fire many API calls, it is not desirable due to the unnecessary resource consumption of processing many API calls, and the latency between our servers and the client sending such API requests.

To solve this issue, we introduce an batch endpoint under our update endpoint, which will allow our API to resolve the data of many customers in a single API request:

POST /api/update/batch
Content-Type: application/json

{
  "<id>": {
    "name": "...",
    "birthday": "...",
  },
  //...
}

This way, we could just iterate over the values of the request and fire up the necessary UPDATE queries to update our records in our database:

foreach ($request_data as $id => $data) {
  DB::table('customers')
    ->where('id', '=', $id)
    ->update($data)
}

// PS: Don't forget to sanitize your data!

With this, we can update the data of many customers in a single API call, while keeping the data for each customer scoped, and we can even filter our columns that are not updated, so our database only processes the information that was actually requested to be updated.

Problem solved, right? āœØ

Wellā€¦ we just missed one little detailā€¦

As you may already know, firing multiple database queries adds up in waiting time before your server can return a response to the end client, and when performing bulk actions in a single call, the response time increases according to the number of queries you need to execute, the complexity, and the latency between your API and the Database ā³.

So, for small sets of data, this is not a problem, but what if, suddenly, our administrators need to update the information of ~1k customers? The response may not be that fast as when we were only batching 5 customers for updates.

3. Multiple records, with independent data, in a single query

We need to optimize the way our batch endpoint works, but how? Thereā€™s no query available in our SQL database that can handle the update of many records with individual sets of data, and letā€™s not talk about dynamically selecting columns to updateā€¦ We can only:

  • UPDATE each record individually, with its own column-value bindings, or
  • UPDATE all records, setting the same value for all the columns.

In order to achieve this, we can make use of the CASE SQL statement, which combined with the UPDATE statement, can open the doors to:

  1. Update multiple records.
  2. Each with independent column-value bindings.
  3. In a single query.

Interesting, right? šŸ’” Well, letā€™s see how the actual query may look like:

UPDATE "customers"
SET
    "name" = CASE
        WHEN "id" = :id_1 THEN :name_1
        WHEN "id" = :id_2 THEN :name_2
        /* ... */
        WHEN "id" = :id_n THEN :name_n
        ELSE "id" /* If current record's id is not the statement list, then fallback to it's own value */
    END,
    "birthday" = CASE
        WHEN "id" = :id_1 THEN :birthday_1
        WHEN "id" = :id_2 THEN :birthday_2
        /* ... */
        WHEN "id" = :id_n THEN :birthday_n
        ELSE "id" /* If current record's id is not the statement list, then fallback to it's own value */
    END
WHERE "id" in :ids

The bindings for this list should be recollected beforehand in order to build the query properly, and when properly implemented, this pattern can not only help you optimize your batch API endpoint, it could also help in situations where:

  • You need to update many items conditionally when deploying a new version of your project.
  • When you are working with an ORM and you need to ā€œsaveā€ the changes of many models at once.

Hope you like this little, but spicy, SQL trick šŸŒ¶.

šŸŽ Bonus: Laravel Mass-Update Package

You may not be aware, but Iā€™m actually a huge Laravel fan, and so, I couldnā€™t miss the opportunity to build a package which wraps the implementation details on this post into an easy to use model trait. It is called Laravel Mass-Update.

If youā€™re not into Laravel or PHP, you can simply check out the source code of this package and port it to your framework, ORM or language of choice.

If youā€™re into Laravel, well, this package is pretty easy to set up, just:

  1. Install the package: composer install iksaku/laravel-mass-update
  2. Import the trait into your model of choice:
use Illuminate\Database\Eloquent\Model;
use Iksaku\Laravel\MassUpdate\MassUpdatable;

class Customer extends Model
{
    use MassUpdatable;

    // ...
}
  1. Make use of itā€™s scoped massUpdate() method:
Customer::query()->massUpdate(values: [...], uniqueBy: [...]);

Please take some time to read the packageā€™s README to see some potential use cases in which you may use this query.