README

Fork from https://github.com/hammerstonedev/fast-paginate

About

This is a fast limit/offset pagination macro for Hyperf. It can be used in place of the standard paginate methods.

This package uses a SQL method similar to a "deferred join" to achieve this speedup. A deferred join is a technique that defers access to requested columns until after the offset and limit have been applied.

In our case we don't actually do a join, but rather a where in with a subquery. Using this technique we create a subquery that can be optimized with specific indexes for maximum speed and then use those results to fetch the full rows.

The SQL looks something like this:

select * from contacts              -- The full data that you want to show your users.
    where contacts.id in (          -- The "deferred join" or subquery, in our case.
        select id from contacts     -- The pagination, accessing as little data as possible - ID only.
        limit 15 offset 150000
    )

You might get an error trying to run the query above! Something like This version of MySQL doesn't yet support 'LIMIT & IN/ALL/ANY/SOME subquery. In this package, we run them as two separate queries to get around that!

The benefits can vary based on your dataset, but this method allows the database to examine as little data as possible to satisfy the user's intent.

It's unlikely that this method will ever perform worse than traditional offset / limit, although it is possible, so be sure to test on your data!

If you want to read 3,000 words on the theory of this package, you can head over to aaronfrancis.com/2022/efficient-pagination-using-deferred-joins.

Installation

This package supports Hyperf 3.0+.

To install, require the package via composer:

composer require friendsofhyperf/fast-paginate

There is nothing further you need to do. The service provider will be loaded automatically by Hyperf.

Usage

Anywhere you would use Model::query()->paginate(), you can use Model::query()->fastPaginate()! That's it! The method signature is the same.

Relationships are supported as well:

User::first()->posts()->fastPaginate();

A Favor

If this helps you, please start me with before and after times! I'd love to know :D

Some community results so far:

• 30 seconds --> 250ms
• 28 seconds --> 2 seconds
• 7.5x faster
• 1.1 seconds --> 0.1 seconds
• 20 seconds --> 2 seconds
• 2 seconds --> .2 seconds

Contact

• Twitter
• Gmail

License

MIT