Datomic Database Plugin — com.fulcrologic/fulcro-rad 0.0.1-alpha-12

Datomic Database Plugin

The following namespace aliases are used in the content of this document:

(ns x
  (:require
    [com.fulcrologic.rad.attributes :as attr]
    [com.fulcrologic.rad.database-adapters.datomic :as datomic]))

Schemas

It is common for larger applications to desire some kind of sharding of their data, particularly in cases of high write load. The Datomic plugin has you model attributes on a Datomic schema which can then be applied to any number of runtime databases in your application. Of course, you must have some scheme for selecting the correct runtime database for mutations and read resolvers during operation.

Attribute Facets

Every attribute in the system that is stored in Datomic using this plugin must include a ::datomic/schema <k> entry, where <k> is a keyword representing a schema name. All attributes that share the same schema name will be stored together in a database that has that schema (see Selecting a Database During Operation).

The following common attribute keys are supported automatically:

::attr/unique?: Causes the attribute to be a unique value, unless a Datomic-specific override for uniqueness is supplied or the attribute is marked as an ::attr/identity?.
::attr/identity?: Causes the attribute to be a unique identity. If supplied, then ::attr/unique? is assumed.
::attr/cardinality: :one or :many (or you can specify it with normal datomic keys).

The plugin-specific attribute parameters are:

::datomic/schema keyword: (required) A name that groups together attributes that go together in a schema on a database. A schema can be used on any number of databased (e.g. for sharding).
::datomic/entity-ids #{k k2 …}: Required on non-identity attributes. A set of attribute keys that are :attr/identity? true. This set indicates that the attribute can be placed on an entity that is identified by one of those identity attributes. This allows the Datomic plugin to figure out which properties can be co-located on an entity for storage in form saves and queries in resolvers. It is a set in order to support the fact that Datomic allows an attribute to appear on any entity, but your domain model will put it on a more limited subset of different entities. Failing to list this entry will result in failure to generate resolvers or form save logic for the attribute.

Any of the normal Datomic schema definition attributes can be included as well (e.g. :db/cardinality), and will take precedence over anything above. Be careful when changing these, as the result can cause incompatible change errors.

Automatic Generation

During development you may choose to create a transaction that can be used to create/update the schema in one or more Datomic databases. This feature is will work as long as your team follows a strict policy of only ever making compatible changes to schema (e.g. mostly additions).

This feature is great for early development, but it may become necessary over time to adopt a migration system or other more manual schema management policy. This feature is therefore easy to opt in/out of at any time.

You can pull the current generated schema from the attribute database using (datomic/automatic-schema schema-key). NOTE: You must require all namespaces in your model that define attributes to ensure that they are all included in the generated schema.

Enumerations

The idiomatic way to represent enumerations in Datomic is with a ref and a set of entities known by well-known idents. The following is supported during automatic schema generation:

(new-attribute :address/state :enum
  {:attr/enumerated-values #{:AL :address.state/CA {:db/ident :address.state/OR :db/doc "Oregon"}}})

All three of the above forms are allowed. An unqualified keyword will be auto-qualified (AL and CA above could be represented either way), and a map will be treated like a full-blown entity definition (will be passed through untouched to the schema).

Validation

If you are not using automatic schema generation then it is recommended that you at least enable schema validation. This feature can be used to check the schema of an existing database against the current code of your attributes to detect inconsistencies between the two at startup. This ensures you will not run into runtime errors due to code that does not match the schema of the target database.

TODO: Write the validator, and document it.

Databases

It is up to you to configure, create, migrate, and manage your database infrastructure; however, this plugin comes with various ustilities that can help you set up and manage the runtime environment. You must follow certain conventions for things to work at all, and may choose to opt into various features that make it easy to get started and evolve your system.

Runtime Operation

Mocked Connections during Development

The Datomock library is a particularly useful tool during experimental phases of development where you have yet to stabilize a particular portion of schema (attribute declarations). It allows you to "fork" a real database connection such that any changes (to schema or otherwise) are thrown away on application restarts.

This allows you to play with new schema without worrying about incompatible schema changes.

It is also quite useful for testing, since it can be used to pre-create (and cache) an in memory database that can be used to exercise Datomic code against your schema without the complete overhead of starting an external database with new schema.

Selecting a Database During Operation

When you set up your Pathom parser you can provide plugins that modify the environment that will be passed by Pathom to all resolvers and mutations on the server. The generated resolvers and mutations for the Datomic plugin need to be able to decide which database should be used for a particular schema in the context of the request. Atomic consistency on reads requires that such a database be provided as a value, whereas mutations will need a connection.

The env must therefore be augmented to contain the following well-known things:

::datomic/connections - A map, keyed by schema, of the database connection that should be used in the context of the current request. ::datomic/databases - A map, keyed by schema, of the most recent database value that should be used in the context of the current request (for consistent reads across multiple resolvers).

TODO: Supply helper funtions that can help with this

Testing

Custom mutations and resolvers are easiest to write if you have a simple way of testing them against a database that looks like your real one. This plugin supports some helpful testing tools that leverage Datomock to give you a fast an consistent starting point for your tests.

Seeding Development Data

We recommend using UUID domain IDs for all entities (e.g. :account/id). This not only enables much of the resolver logic, it also allows you to easily and consistently seed development data for things like live coding and tests.

The com.fulcrologic.rad.ids/new-uuid function can be used to generate a new random UUID in CLJC, but it can also be used to generate a constant (well-known) UUID for testing.

A Sample Test

The core function to use is datomic/empty-db-connection, which can work with automatically-generated schema or a manual schema. It returns a Datomic connection which has the supplied schema (and is memoized for fast startup on sequences of tests).

A typical test might look like the following:

(deftest sample-test
  ;; the empty-db-connection can accept a schema txn if needed.
  (let [conn        (datomic/empty-db-connection :production)
        sample-data [{::acct/id   (new-uuid 1)
                      ::acct/name "Joe"}]]
    @(d/transact conn sample-data)

    (let [db (d/db conn)
          a  (d/pull db [::acct/name] [::acct/id (new-uuid 1)])]
      (is (= "Joe" (::acct/name a))))))

The connection is memoized based on the schema key (not any supplied migration data). You can use (datomic/reset-test-schema k) to forget the current memoized version.