shabushabu / laravel-paradedb-search
Integrates the pg_search extension by ParadeDB into Laravel
Fund package maintenance!
boris-glumpler
Requires
- php: ^8.2
- illuminate/contracts: ^11.0
- spatie/laravel-package-tools: ^1.16
Requires (Dev)
- larastan/larastan: ^2.9
- laravel/pint: ^1.14
- nunomaduro/collision: ^8.0
- orchestra/testbench: ^9.0
- pestphp/pest: ^2.34
- pestphp/pest-plugin-arch: ^2.7
- pestphp/pest-plugin-laravel: ^2.3
- pestphp/pest-plugin-type-coverage: ^2.8
- phpstan/extension-installer: ^1.3
- phpstan/phpstan-deprecation-rules: ^1.1
- phpstan/phpstan-phpunit: ^1.3
- roave/security-advisories: dev-latest
- tpetry/laravel-postgresql-enhanced: ^0.39.0
- tpetry/laravel-query-expressions: ^1.3
Suggests
- tpetry/laravel-postgresql-enhanced: Adds vector operators to use in regular Eloquent where statements
- tpetry/laravel-query-expressions: Provides useful expressions for use in ParadeDB search queries
README
ParadeDB Search for Laravel
Integrates the pg_search
Postgres extension by ParadeDB into Laravel
Supported minimum versions
Installation
Caution
Please note that this is a new package and, even though it is well tested, it should be considered pre-release software
Before installing the package you should install and enable the pg_search extension.
You can then install the package via composer:
composer require shabushabu/laravel-paradedb-search
You can also publish the config file:
php artisan vendor:publish --tag="laravel-paradedb-search-config"
These are the contents of the published config file:
return [ 'index_suffix' => '_idx', ];
Usage
Add a bm25 index
Each model that you want to be searchable needs a corresponding bm25
index. These can be generated within a migration like so:
use ShabuShabu\ParadeDB\Indices\Bm25; return new class extends Migration { public function up(): void { Schema::create('products', static function (Blueprint $table) { // all your product fields }); Bm25::index('products') ->addNumericFields(['amount']) ->addBooleanFields(['is_available']) ->addDateFields(['created_at', 'deleted_at']) ->addJsonFields(['options']) ->addTextFields([ 'name', 'currency', 'description' => [ 'tokenizer' => [ 'type' => 'default', ], ], ]) ->create(drop: true); } public function down(): void { Bm25::index('products')->drop(); } };
Preparing your model
Just add the Searchable
trait to your model to enable search:
use Illuminate\Database\Eloquent\Model; use ShabuShabu\ParadeDB\Concerns\Searchable; class Product extends Model { use Searchable; // the rest of the model... }
ParadeQL
ParadeDB Search for Laravel comes with a fluent builder for ParadeQL, a simple query language.
This builder can be passed as a condition to a search where
method or used within the various ParadeDB expressions.
Basic query
use ShabuShabu\ParadeDB\ParadeQL\Builder; Builder::make()->where('description', 'keyboard')->get(); // results in: description:keyboard
Add an IN condition
Builder::make() ->where('description', ['keyboard', 'toy']) ->get(); // results in: description:IN [keyboard, toy]
Add an AND NOT condition
Builder::make() ->where('category', 'electronics') ->whereNot('description', 'keyboard') ->get(); // results in: category:electronics AND NOT description:keyboard
Boost a condition
Builder::make()->where('description', 'keyboard', boost: 1)->get(); // results in: description:keyboard^1
Apply the slop operator
Builder::make()->where('description', 'ergonomic keyboard', slop: 1)->get(); // results in: description:"ergonomic keyboard"~1
More complex example with a sub condition
Builder::make() ->where('description', ['keyboard', 'toy']) ->where( fn (Builder $builder) => $builder ->where('category', 'electronics') ->orWhere('tag', 'office') ) ->get(); // results in: description:IN [keyboard, toy] AND (category:electronics OR tag:office)
Apply a simple filter
use ShabuShabu\ParadeDB\ParadeQL\Operators\Filter; Builder::make()->whereFilter('rating', Filter::equals, 4)->get(); // results in: rating:4
Apply a boolean filter
use ShabuShabu\ParadeDB\ParadeQL\Operators\Filter; Builder::make()->whereFilter('is_available', '=', false)->get(); // results in: is_available:false
Apply a basic range filter
use ShabuShabu\ParadeDB\ParadeQL\Operators\Filter; Builder::make()->whereFilter('rating', '>', 4)->get(); // results in: rating:>4
Apply an inclusive range filter
use ShabuShabu\ParadeDB\ParadeQL\Operators\Range; Builder::make()->whereFilter('rating', Range::includeAll, [2, 5])->get(); // results in: rating:[2 TO 5]
Apply an exclusive range filter
use ShabuShabu\ParadeDB\ParadeQL\Operators\Range; Builder::make()->whereFilter('rating', Range::excludeAll, [2, 5])->get(); // results in: rating:{2 TO 5}
ParadeDB functions
For more complex operations, it will be necessary to use some of the provided ParadeDB functions, all of which have corresponding query expressions:
Get all the records
use ShabuShabu\ParadeDB\Query\Expressions\All; Product::search()->where(new All())->get();
Get none of the records
use ShabuShabu\ParadeDB\Query\Expressions\Blank; Product::search()->where(new Blank())->get();
Boost a query
use ShabuShabu\ParadeDB\Query\Expressions\All; use ShabuShabu\ParadeDB\Query\Expressions\Boost; Product::search()->where(new Boost(new All(), 3.9))->get();
Add a constant score
use ShabuShabu\ParadeDB\Query\Expressions\All; use ShabuShabu\ParadeDB\Query\Expressions\ConstScore; Product::search()->where(new ConstScore(new All(), 3.9))->get();
Perform a disjunction max query
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\DisjunctionMax; Product::search()->where( new DisjunctionMax(Builder::make()->where('description', 'keyboard')) )->get();
The DisjunctionMax
constructor also accepts an array of queries, so using the fluid interface might be more convenient for multiple queries:
Product::search()->where( DisjunctionMax::query() ->add(Builder::make()->where('description', 'keyboard')) ->add('description:blue') ->tieBreaker(1.2) )->get();
This also allows you to conditionally add queries:
Product::search()->where( DisjunctionMax::query() ->add(Builder::make()->where('description', 'keyboard')) ->add('description:blue', when: false) )->get();
Search for a fuzzy term
use ShabuShabu\ParadeDB\Query\Expressions\FuzzyTerm; Product::search()->where(new FuzzyTerm('description', 'keyboard'))->get();
Highlight search terms
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\Highlight; use ShabuShabu\ParadeDB\Query\Expressions\DisjunctionMax; Product::search() ->select(['*', new Highlight('id', 'name')]) ->where(new DisjunctionMax(Builder::make()->where('description', 'keyboard'))) ->get();
Search for a phrase
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\Phrase; Product::search() ->where(new Phrase('description', ['robot', 'building', 'kits'])) ->get();
Perform a phrase prefix query
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\PhrasePrefix; Product::search() ->where(new PhrasePrefix('description', ['robot', 'building', 'kits', 'am'])) ->get();
Search within a given range
use ShabuShabu\ParadeDB\Query\Expressions\Range; use ShabuShabu\ParadeDB\Query\Expressions\Ranges\Int4; use ShabuShabu\ParadeDB\Query\Expressions\Ranges\Bounds; Product::search() ->stableSort() ->where(new Range('rating', new Int4(1, 3, Bounds::includeStartExcludeEnd))) ->get();
Here are the supported range types (all within the ShabuShabu\ParadeDB\Query\Expressions\Ranges
namespace), plus their corresponding Postgres type:
Int4::class;
orint4range
Int8::class;
orint8range
Numeric::class;
ornumrange
Date::class;
ordaterange
Timestamp::class;
ortsrange
TimestampTz::class;
ortstzrange
Perform a regex query
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\Regex; Product::search() ->where(new Regex('description', '(team|kits|blabla)')) ->get();
Search for a term
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\Term; Product::search() ->where(new Term('description', 'building')) ->get();
Search for a set of terms
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\Term; use ShabuShabu\ParadeDB\Query\Expressions\TermSet; Product::search() ->where(new TermSet([ new Term('description', 'building'), new Term('description', 'things'), ])) ->get();
The above query can also be written in a fluid manner:
Product::search()->where( TermSet::query() ->add(new Term('description', 'building')) ->add(new Term('description', 'things')) )->get();
The term
method allows you to conditionally add terms:
$when = false; Product::search()->where( TermSet::query()->add(new Term('description', 'things'), $when) )->get();
Perform a complex boolean query
use App\Models\Product; use ShabuShabu\ParadeDB\Query\Expressions\Range; use ShabuShabu\ParadeDB\Query\Expressions\Boolean; use ShabuShabu\ParadeDB\Query\Expressions\FuzzyTerm; use ShabuShabu\ParadeDB\Query\Expressions\Ranges\TimestampTz; Product::search() ->where(new Boolean( must: [ new Range('created_at', new TimestampTz(null, now())), ], should: [ new Boost(new FuzzyTerm('name', 'keyboard'), 2), new FuzzyTerm('description', 'keyboard'), ], mustNot: [ new Range('deleted_at', new TimestampTz(null, now())), ], )) ->get();
Boolean queries can also be constructed in a fluid manner:
Product::search()->where( Boolean::query() ->must(new Range('created_at', new TimestampTz(null, now()))) ->should(new Boost(new FuzzyTerm('name', 'keyboard'), 2)) ->should(new FuzzyTerm('description', 'keyboard')) ->mustNot(new Range('deleted_at', new TimestampTz(null, now()))) )->get();
The two queries above are identical. The fluent methods allow you to conditionally add queries, though:
$when = false; Product::search()->where( Boolean::query() ->must(new Range('created_at', new TimestampTz(null, now()))) ->should(new Boost(new FuzzyTerm('name', 'keyboard'), 2), $when) )->get();
Sort by rank
use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\Term; use ShabuShabu\ParadeDB\Query\Expressions\Rank; Product::search() ->addSelect(new Rank('id')) ->where(new Term('description', 'building')) ->get();
Pagination
It's also possible to paginate the results. Both the paginate
and simplePaginate
methods use the underlying limit
& offset
functionality, so will be more performant:
use App\Models\Product; use App\Models\Product; use ShabuShabu\ParadeDB\ParadeQL\Builder; Product::search() ->where(Builder::make()->where('description', 'keyboard')) ->paginate(20);
Search parameters
The ParadeDB search
function allows you to set a variety of parameters to fine-tune your search. All of these can be set here as well:
use App\Models\Product; use App\Models\Product; use ShabuShabu\ParadeDB\ParadeQL\Builder; Product::search() ->where(Builder::make()->where('description', 'keyboard')) ->alias('alias') ->stableSort() ->limit(12) ->offset(24) ->get();
Hybrid search
Whenever a similarity query is provided, the package will automatically perform a hybrid search. Please note that a ParadeDB query is still required!
use App\Models\Product; use ShabuShabu\ParadeDB\ParadeQL\Builder; use ShabuShabu\ParadeDB\Query\Expressions\Distance; use ShabuShabu\ParadeDB\Query\Expressions\Similarity; Product::search() ->where(Builder::make()->where('description', 'keyboard')) ->where(new Similarity('embedding', Distance::l2, [1, 2, 3])) ->get();
Search parameters
Similarly to the full-text search, there are also parameters you can set for a hybrid search:
Product::search() ->where(Builder::make()->where('description', 'keyboard')) ->where(new Similarity('embedding', Distance::l2, [1, 2, 3])) ->bm25Limit(100) ->bm25Weight(0.5) ->similarityLimit(100) ->similarityWeight(0.5) ->get();
Modifying the underlying query
Occasionally, it will be necessary to modify the base query, for example to eager-load some relationships. This can be accomplished like so:
use App\Models\Product; use Illuminate\Database\Eloquent; use ShabuShabu\ParadeDB\ParadeQL\Builder; Product::search() ->modifyQueryUsing(fn (Eloquent\Builder $builder) => $builder->with('tags')) ->where(Builder::make()->where('description', 'keyboard')) ->get();
A word of caution
While it is possible to combine ParadeDB queries with regular Eloquent queries, you will incur some performance penalties.
For optimal performance it is recommended to let the bm25
index do as much work as possible, so wherever possible you should use the built-in filters as well as limit & offset!
Getting help
If your issue has something to do with this package, then please use the issues and discussions!
If your issue is related to pg_search
, tho, then please create a discussion in the ParadeDB repo.
To make this a bit easier, you can use the paradedb:help
command that ships with this package:
php artisan paradedb:help
Please note that this command is just an implementation of the paradedb.help()
function. Please use this command wisely!
Testing
The tests require a PostgreSQL database, which can easily be set up by running the following script:
composer testdb
Warning
Please note that both pg_search and pgvector extensions need to be available already.
Then run the tests:
composer test
Or with test coverage:
composer test-coverage
Or with type coverage:
composer type-coverage
Or run PHPStan:
composer analyse
ParadeDB test table
There is also a command that allows you to create and drop the built-in test table
php artisan paradedb:test-table create
Changelog
Please see CHANGELOG for more information on what has changed recently.
Contributing
Please see CONTRIBUTING for details.
Security Vulnerabilities
Please review our security policy on how to report security vulnerabilities.
Credits
- Taylor Otwell for creating Laravel
- ParadeDB for creating
pg_search
- ShabuShabu
- All Contributors
Disclaimer
This is a 3rd party package and ShabuShabu is not affiliated with either Laravel or ParadeDB.
License
The MIT License (MIT). Please see License File for more information.