Download the PHP package eelcol/laravel-meilisearch without Composer

Laravel Meilisearch with QueryBuilder

When you want to use Meilisearch in your Laravel application, you can use Laravel Scout. This is an easy way to sync your models to Meilisearch and quickly search models using Meilisearch. However, sometimes Laravel Scout is not enough. For example if you want:

• More control over your Meilisearch database: do not only save models for example.
• Set searchable, filterable or sortable attributes.
• Perform more complex queries to Meilisearch, for example with multiple filters.
• Use a query builder to fetch data from the Meilisearch database.
• To use some functionalities that are not available in Meilisearch out-of-the-box. For example, displaying documents in random order or displaying facets that are not available in the filtered documents.

This package deals with these kind of situations. You decide which information to send to Meilisearch, and which information you want back. The query builder specifically built for Meilisearch helps to build more complex queries.

When using this package, you should determine yourself when and which data you sent to Meilisearch. So if you automatically want to sent models to Meilisearch after a model is saved or created, Laravel Scout might be a better solution.

Compatibility with Meilisearch

Currently, this package supports Meilisearch up to version 0.27. Version 0.28 of Meilisearch introduced some breaking changes. A new version of this package compatible with 0.28 will be released soon.

Installation

Determine which version you need based on the Meilisearch version:

So for example, when using Meilisearch 1.0.2, use the following command:

Setup .env

Change your .env to include the following variables:

When not using a Meilisearch key, the .env variable MEILISEARCH_KEY can be any value.

Publish assets

Getting started

Create an index

First you need to create an index to save documents to. For example, you need an index to save our products catalogue to. So the following command can be used:

This command will create a file database/meilisearch/products.php. In this file, you can adjust settings for this index. This is not required, however it is highly recommended. If you leave the standard settings, Meilisearch will use all columns of your data to search on. To achieve this, Meilisearch must index all columns of your data. This will take a longer time, and uses more server resources. That's why it is recommended to specify which columns should be searchable, filterable and sortable.

Everytime you want to change something to the settings, simply change this file. After the changes, run the command below.

Migrate the index to the Meilisearch database

Now the index has to be actually created. To achieve this, run the following command:

Compare this to the database migrations of Laravel. First you have to create a database migration, next you have to run the migration to actually create the table, or make the adjustment.

Run this command every time you make changes to the database/meilisearch/products.php file. Also, run this command on every deployment, so you have an up-to-date Meilisearch instance in production.

When you want to set the index settings on a different Meilisearch installation, you can use the --mshost and --mskey options:

Master data in Meilisearch

All the functionalities that are mentioned in the Meilisearch docs are available in this package. The most important functionalities are listed below:

Insert data

To insert a document in the index products, you can do 1 of the following:

You can also directly insert a model or collection. A model gets converted to an array. In order to do this, the package check if the following methods exists on the object, in the following order:

A Product model for example, can look like the following:

The model can be inserted like this:

A collection can also be directly inserted:

Retrieve data

Documents of an index can be retrieved using the getDocuments method. When you want to apply filters, using the query builder is advised. The data is automaticly paginated.

Delete documents

Documents can be deleted by ID using the deleteDocument or deleteDocuments methods. Both methods return a MeilisearchTask object.

Documents can also be deleted using the query builder, see below.

Retrieve facet values

You can retrieve all available values for a given facet. For example, the code below will search for all brands with the letter 'a' in it.

Use the query builder

If you want to apply filtering or sorting, I recommend using the query builder. You can take a look in the tests folder to see some examples. A few simple examples are listed below.

Filter on attribute

Simple filtering can be done using the where method:

Multiple wheres can also be combined:

Filter with 'or'

Currently, it is not possible to filter with 'OR' on the top-level. If you want to filter with 'OR', you have to create a 'where-group' first. The following call will generate an error:

The following code will work however:

This is because of the way Meilisearch filters work, and how this package renders the filters. It also prevents possible issues when combining 'AND' and 'OR' statements. For example, the following query could return unexpected results:

Should this query be:

So for now, when using an 'OR' statement, you should start a where-group first.

Where in

This works best with arrays. For example, you have a product with multiple categories:

This data can be queried:

The whereIn method will check if at least 1 of the values is present on the model. So the query above, will return all documents.

Where matches

The whereIn method will check if at least 1 of the values is present on the model. The whereMatches method will check if ALL values are present on the model:

Empty data

whereIsEmpty or whereNotEmpty can be used to select documents that have empty values for the given attribute. This matches the following JSON values: "", [], {}.

whereNull or whereNotNull can be used to select documents that have NULL values for the given attribute.

Using facets

Columns that are attributed as filterable can be used in facets. The querybuilder will return these facets with a product-count attached to it. The facets can be defined by using the setFacets or addFacet methods:

Delete documents using the querybuilder

You can use filters on a query to delete documents. Other elements on the query, such as pagination or sorting will not be applied.

The query above will delete all products with the category phones.

The query above is exactly the same as the other query! Remember: when deleting documents using the query builder, only filters will be applied. Limits, ordering and other manipulations will not be applied.

Specify which attributes to search on

You can specify per query the attributes to search on. By default, a query searches on all attributes you have specified in the settings file. But if you want to search on another set of attributes, you can use the searchOnAttributes method. The attributes must be a subset of the attributes you are searching on by default.

For example, when you have setup an index that searches on title, description and brand by default, you can make a query that only searches on the title:

This query only finds products with Nike in the title. If a product does not have Nike in the title, but in the description, the above query will not return this product.

Disjunctive facets distribution

In the current version of Meilisearch, facets of an attribute are not returned when you are filtering on that specific attribute. See the following discussion: https://github.com/meilisearch/product/discussions/187

For example, when you run the above query, the colors grey, silver, gold, yellow are returned. Next, you only want to display the products with a yellow color. So you apply a filter:

However when you do this, the facet color will now only return yellow. That makes it more difficult to display all the possible colors to the end-user. Thats why this package has a keepFacetsInMetadata method. You can apply filters inside this method, which will not be applied when fetching metadata.

Starting from Meilisearch version 1.1, this problem can be solved using the multi-search endpoint. This is how I solved this issue in this package. The package will make an extra query for every filter used in the query. But all queries are combined in a single request, to reduce the number of resources needed. Take a look at the following example:

This query will make the following requests:

• Get all brands, with products that are in the phones category and match color=yellow and size=XL
• Get all colors, with products that are in the phones category and match size=XL
• Get all sizes, with products that are in the phones category and match color=yellow

Since version 1.1 of Meilisearch, this is the recommended way of using multi-select facets.

Limits and offsets

Limits and offsets can easily be added to the query. The following query will return 10 results, starting from the 20th result:

Paginate the results

Just like using the Laravel querybuilder for the database, you can paginate the results coming from Meilisearch. Simply use the paginate method. When using this method, earlier calls to limit and offset are ignored.

Optionally supply the name of the query-parameter to use to fetch the current page. 'page' is used by default.

Sort the results

In random order

Out of the box, Meilisearch does not offer the option to randomly order documents. However, sometimes you want to display a few random products. To make this possible, this package adds this functionality. Be aware that the package will make a query to your Meilisearch database for every random element, plus 1 extra query. So if you want to fetch 100 documents in random order, there will be 101 queries made. Meilisearch queries are very fast, however when you make this kind of number of queries, it can still become slow. So I recommend to use this method only with a low number of documents (less than 10), or for example cache the results.