The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.

NAME

Elasticsearch::Model

DESCRIPTION

This module is meant to be used very much like ElasticSearchX::Model, before it was deprecated. In fact, most of the code is borrowed from that module.

All the changes I have made are aimed at two things:

  1. Compatibility with Elasticsearch 6+

  2. Mapping, period. Classes to index is all that this distribution does. We do not take data from elasticsearch and wrap it in the document type classes. We do not handle index aliasing, either, which can be quite complicated. For aliasing, however, deploy_to will allow you to specify the name of the index you will be creating, so that you can make the alias more directly.

SYNOPSIS

This is the simplest use case:

my $model = MyApplication::Elasticsearch::Model->new;

$model->deploy;

This is probably the actual use case, in which you have an alias for a versioned index name:

my $document_types = {
    order => 'Order',
    user  => 'User',
    note  => 'Note',
};

my $model = MyApplication::Elasticsearch::Model->new;

for my $doc_type (keys %$document_types) {
    my $index_object = $model->index($doc_type);
    my $versioned_index_name = $doc_type . "_" . time(); # to get "order_12345859245"
    $index_object->deploy_to(index_name => $versioned_index_name);

    # Get index currently point to by an alias
    my $old_index_name = $model->es->cat->aliases(name => $doc_type, format => 'json')->[0]->{index};

    # Switch the alias to the new index, and remove the alias from the old one.
    $model->es->indices->update_aliases(
        body => {
            actions => [
                {add => {alias => $doc_type, index => $new_index_name}},
                (
                    $old_index_name
                    ? ({remove => {alias => $doc_type, index => $old_index_name}})
                    : ()
                ),
            ],
        },
    );
}

To do either of the above, you start with a model.

package MyApplication::Elasticsearch::Model;

use Moose;
use Elasticsearch::Model;

my $document_types = {
    order => 'Order',
    user  => 'User',
    note  => 'Note',
};

my $index_settings = {
    refresh_interval   => '2s',
    number_of_shards   => 4,
    number_of_replicas => 3,
};

for my $doc_type (keys %$document_types) {
    index $doc_type => (
        namespace      => 'private',
        type           => $document_types->{$doc_type},
        traits         => ["MyApplication::Elasticsearch::DocTrait"],
        index_settings => $index_settings,
    );
}

normalizer normie => (
    type   => "custom",
    filter => ["lowercase"],
);

1;

Elasticsearch 6+ more or less requires one document type per index, so this makes things simple for us. We define one document class per document type, e.g., type user gets document class User.

We will, by default, search for these document classes underneath the model class; in this case, we would search for MyApplication::Elasticsearch::Model::User. You can, however, define a document_namespace in your model class as follows:

...

has_document_namespace "MyApplication::ESDocumentClasses";
...

This would mean that we look for MyApplication::ESDocumentClasses::User.

You use index to define the indexes belonging to your model. Here is an example of the Note document class.

package MyApplication::Elasticsearch::Model::Note;

use Moose;
use Elasticsearch::Model::Document;

has text => (
    is   => 'ro',
    isa  => 'Maybe[Str]',
    type => 'text',
);

has author_id => (
    is   => 'ro',
    isa  => 'Maybe[Int]',
    type => 'integer',
);

has_non_attribute_mapping {
    _source => {
        enabled => "false",
    },
};

1;

You define your attributes as you would expect, setting Moose type attributes, and also using the type attribute to tell elasticsearch what the elasticsearch type is. You can set non_attribute_mappings here as well, as shown in the above example, and these will be folded into your index mapping.

PUBLIC METHODS

index

This is an overloaded method that has 2 modes of operation:

1 Adds one single index to the model's metaclass.

This is a compile-time metaclass method that is called when you have the following in your model class:

index addresses => (
    type => 'Address',
);
2 Retrieves a single index object from the model instance.

This is a run-time instance method that is called when you do the following:

my $index_object = $model->index($doc_type);

analyzer

This is a compile-time metaclass method that enables you to declare a custom analyzer using the following syntax in your model class:

analyzer electro => (
    type      => "custom",
    tokenizer => "standard",
    filter    => ["lowercase", "filtration"],
);

tokenizer

This is a compile-time metaclass method that enables you to declare a custom tokenizer using the following syntax in your model class:

tokenizer splat => (
    type    => "simple_pattern",
    pattern => "[0123456789]{3}",
);

normalizer

This is a compile-time metaclass method that enables you to declare a custom normalizer using the following syntax in your model class:

normalizer normie => (
    type   => "custom",
    filter => ["lowercase"],
);

filter

This is a compile-time metaclass method that enables you to declare a custom filter using the following syntax in your model class:

filter filtration => (
    type     => "edge_ngram",
    min_gram => 1,
    max_gram => 24,
);

has_document_namespace

This is a compile-time metaclass method that enables you to declare the namespace in which to search for your document classes in your model class:

has_document_namespace "MyApplication::Elasticsearch::DocumentClasses";

PRIVATE METHODS

_apply_namespace

Takes the index name and the namespace, and constructs the namespaced_name.

Used internally, when adding an index to the model's metaclass. See Elasticsearch::Model::Role::Metaclass.

_add_index_for_name

Takes the name of an index and some arguments, and adds that index to the model's metaclass. See Elasticsearch::Model::Role::Metaclass.