Valkyrie

Valkyrie is a gem for enabling multiple backends for storage of files and metadata in Samvera.

Code:

Docs:

Jump in:

Alexandra Dunn

Trey Pendragon

The Samvera community is here to help. Please see our support guide.

Add this line to your application's Gemfile:

gem 'valkyrie'

And then execute:

$ bundle

Valkyrie is configured in two places: an initializer that registers the persistence options and a YAML configuration file that sets which options are used by default in which environments.

Here is a sample initializer that registers a couple adapters and storage adapters, in each case linking an instance with a short name that can be used to refer to it in your application:

# frozen_string_literal: true
require 'valkyrie'


Rails.application.config.to_prepare do

  # To use the postgres adapter you must add `gem 'pg'` to your Gemfile
  Valkyrie::MetadataAdapter.register(
    Valkyrie::Persistence::Postgres::MetadataAdapter.new,
    :postgres
  )

  # To use the solr adapter you must add gem 'rsolr' to your Gemfile
  Valkyrie::MetadataAdapter.register(
    Valkyrie::Persistence::Solr::MetadataAdapter.new(
      connection: Blacklight.default_index.connection
    ),
    :solr
  )

  # To use the fedora adapter you must add `gem 'ldp'` to your Gemfile
  Valkyrie::MetadataAdapter.register(
    Valkyrie::Persistence::Fedora::MetadataAdapter.new(
      connection: ::Ldp::Client.new("http://localhost:8988/rest"),
      base_path: "test_fed",
      schema: Valkyrie::Persistence::Fedora::PermissiveSchema.new(title: RDF::URI("http://bad.com/title"))
    ),
    :fedora
  )

  Valkyrie::MetadataAdapter.register(
    Valkyrie::Persistence::Memory::MetadataAdapter.new,
    :memory
  )

  Valkyrie::StorageAdapter.register(
    Valkyrie::Storage::Disk.new(base_path: Rails.root.join("tmp", "files")),
    :disk
  )

  Valkyrie::StorageAdapter.register(
    Valkyrie::Storage::Fedora.new(connection: Ldp::Client.new("http://localhost:8988/rest")),
    :fedora
  )


  Valkyrie::StorageAdapter.register(
    Valkyrie::Storage::Memory.new,
    :memory
  )
end

The initializer registers four Valkyrie::MetadataAdapter instances for storing metadata:

• :fedora which stores metadata in a Fedora server.
• :memory which stores metadata in an in-memory cache (this cache is not persistent, so it is only appropriate for testing).
• :postgres which stores metadata in a PostgreSQL database.
• :solr which stores metadata in a Solr Index (Solr Persister issues a warning if it has to generate an ID for a new resource because it is intended to be used as a secondary persister).

Other adapter options include Valkyrie::Persistence::BufferedPersister for buffering in memory before bulk updating another persister, Valkyrie::Persistence::CompositePersister for storing in more than one adapter at once, Valkyrie::Persistence::Solr for storing in Solr, and Valkyrie::Persistence::Fedora for storing in Fedora.

The initializer also registers three Valkyrie::StorageAdapter instances for storing files:

• :disk which stores files on disk
• :fedora which stores files in Fedora
• :memory which stores files in an in-memory cache (again, not persistent, so this is only appropriate for testing)

require 'valkyrie'
Rails.application.config.to_prepare do
  Valkyrie.config.resource_class_resolver = lambda do |resource_klass_name|
    # Do complicated lookup based on the string
  end
end

A sample configuration file that configures your application to use different adapters:

development:
  metadata_adapter: postgres
  storage_adapter: disk

test:
  metadata_adapter: memory
  storage_adapter: memory

production:
  metadata_adapter: postgres
  storage_adapter: fedora

For each environment, you must set two values:

• metadata_adapter is the store where Valkyrie will put the metadata
• storage_adapter is the store where Valkyrie will put the files

The values are the short names used in your initializer.

Further details can be found on the Persistence Wiki page.

Define a custom work class:

# frozen_string_literal: true
class MyModel < Valkyrie::Resource
  include Valkyrie::Resource::AccessControls
  attribute :title, Valkyrie::Types::Set    # Sets deduplicate values
  attribute :date, Valkyrie::Types::Array   # Arrays can contain duplicate values
end

Attributes are unordered by default. Adding ordered: true to an attribute definition will preserve the order of multiple values.

attribute :authors, Valkyrie::Types::Array.meta(ordered: true)

Defining resource attributes is explained in greater detail on the Using Types Wiki page.

# initialize a metadata adapter
adapter = Valkyrie::MetadataAdapter.find(:postgres)

# create an object
object1 = MyModel.new title: 'My Cool Object', authors: ['Jones, Alice', 'Smith, Bob']
object1 = adapter.persister.save(resource: object1)

# load an object from the database
object2 = adapter.query_service.find_by(id: object1.id)

# load all objects
objects = adapter.query_service.find_all

# load all MyModel objects
Valkyrie.config.metadata_adapter.query_service.find_all_of_model(model: MyModel)

The Wiki documents the usage of Queries, Persistence, and ChangeSets and Dirty Tracking.

A Valkyrie repository may have concurrent updates, for example, from a load-balanced Rails application, or from multiple Sidekiq background workers). In order to prevent multiple simultaneous updates applied to the same resource from losing or corrupting data, Valkyrie supports optimistic locking. How to use optimistic locking with Valkyrie is documented on the Optimistic Locking Wiki page.

Valkyrie's public API is defined by the shared specs that are used to test each of its core classes. This include change sets, resources, persisters, adapters, and queries. When creating your own kinds of these kinds of classes, you should use these shared specs to test your classes for conformance to Valkyrie's API.

When breaking changes are introduced, necessitating a major version change, the shared specs will reflect this. When new features are added and a minor version is released there will be no change to the existing shared specs, but there may be new ones. These new shared specs will fail in your application if you have custom adapters, but your application will still work.

Using the shared specs in your own models is described in more detail on the Shared Specs Wiki page.

When configuring your adapter, include the fedora_version parameter in your metadata or storage adapter config. If Fedora requires auth, you can also include that in the URL, e.g.:

Valkyrie::Storage::Fedora.new(
  connection: Ldp::Client.new("http://fedoraAdmin:fedoraAdmin@localhost:8988/rest"),
  fedora_version: 5
)

Fedora 4 and 6.5+ support automatic creation of "pairtree" paths, which means that identifiers are stored in a hashed directory structure to avoid excessive child nodes directly in the root node.

Unlike Fedora 4 where this is the default behavior, it has to be enabled via options in Fedora 6.5. This is done by configuring both the count and length of the pairtree algorithm, so it needs to be configured in the Fedora adapters so that both sides match.

In the following example, Fedora 6.5 is being started with options to create pairtree paths consisting of 4 segments, 2 characters in length:

CATALINA_OPTS=-Dfcrepo.home=/fcrepo-home ...
...
-Dfcrepo.pid.minter.length=2 -Dfcrepo.pid.minter.count=4

For the Fedora metadata/storage adapters to correctly translate identifiers into URI's, they should be initialized like:

Valkyrie::Persistence::Fedora::MetadataAdapter.new(
    connection: ::Ldp::Client.new("http://localhost:8080/fcrepo/rest"),
    fedora_version: 6.5,
    fedora_pairtree_count: 4,
    fedora_pairtree_length: 2
  )

In the configuration above, an ID of AaBbCcDd will be created in Fedora under the root node as /Aa/Bb/Cc/Dd/AaBbCcDd. If count and length correctly match in the Fedora adapters, that same path will be calculated and returned.

Note that the configuration above is not required for pairtree paths to work correctly in Fedora 4, and automatic pairtree creation is not supported under Fedora 5 or Fedora 6 < 6.5.

For ease of development we use Lando to abstract away some complications of using Docker containers for development.

1. Install the latest released > 3.0 version of Lando from here.
2. bundle install(Ruby 2.6+ required)
3. bundle exec rake server:start
4. bundle exec rspec spec

1. bundle exec rake server:clean

1. bundle exec rake server:stop

You can also run lando poweroff from anywhere.

This software has been developed by and is brought to you by the Samvera community. Learn more at the Samvera website.

Bug reports and pull requests are welcome on GitHub at https://github.com/samvera/valkyrie/.

If you're working on PR for this project, create a feature branch off of main.

If you’re developing an application that uses Valkyrie, consider adding it to the list of Valkyrie apps!

This repository follows the Samvera Community Code of Conduct and language recommendations. Please do not create a branch called master for this repository or as part of your pull request; the branch will either need to be removed or renamed before it can be considered for inclusion in the code base and history of this repository.