View Source AshPostgres.Repo behaviour (ash_postgres v2.4.12)
Resources that use AshPostgres.DataLayer
use a Repo
to access the database.
This repo is a thin wrapper around an Ecto.Repo
.
You can use Ecto.Repo
's init/2
to configure your repo like normal, but
instead of returning {:ok, config}
, use super(config)
to pass the
configuration to the AshPostgres.Repo
implementation.
Installed Extensions
To configure your list of installed extensions, define installed_extensions/0
Extensions can be a string, representing a standard postgres extension, or a module that implements AshPostgres.CustomExtension
.
That custom extension will be called to generate migrations that serve a specific purpose.
Extensions that are relevant to ash_postgres:
- "ash-functions" - This isn't really an extension, but it expresses that certain functions
should be added when generating migrations, to support the
||
and&&
operators in expressions. "uuid-ossp"
- Sets UUID primary keys defaults in the migration generator"pg_trgm"
- Makes theAshPostgres.Functions.TrigramSimilarity
function available- "citext" - Allows case insensitive fields to be used
"vector"
- Makes theAshPostgres.Functions.VectorCosineDistance
function available. SeeAshPostgres.Extensions.Vector
for more setup instructions.
def installed_extensions() do
["pg_trgm", "uuid-ossp", "vector", YourCustomExtension]
end
Transaction Hooks
You can define on_transaction_begin/1
, which will be invoked whenever a transaction is started for Ash.
This will be invoked with a map containing a type
key and metadata.
%{type: :create, %{resource: YourApp.YourResource, action: :action}}
Summary
Callbacks
Return a list of all schema names (only relevant for a multitenant implementation)
Should the repo should be created by mix ash_postgres.create
?
The default prefix(postgres schema) to use when building queries
Should the repo should be dropped by mix ash_postgres.drop
?
Use this to inform the data layer about what extensions are installed
The path where your migrations are stored
Configure the version of postgres that is being used.
Use this to inform the data layer about the oldest potential postgres version it will be run on.
Allows overriding a given migration type for all fields, for example if you wanted to always use :timestamptz for :utc_datetime fields
Whether or not to explicitly start and close a transaction for each action, even if there are no transaction hooks. Defaults to true
.
Whether or not to explicitly start and close a transaction for each atomic update action, even if there are no transaction hooks. Defaults to false
.
The path where your tenant migrations are stored (only relevant for a multitenant implementation)
Callbacks
@callback all_tenants() :: [String.t()]
Return a list of all schema names (only relevant for a multitenant implementation)
@callback create?() :: boolean()
Should the repo should be created by mix ash_postgres.create
?
@callback default_prefix() :: String.t()
The default prefix(postgres schema) to use when building queries
@callback drop?() :: boolean()
Should the repo should be dropped by mix ash_postgres.drop
?
Use this to inform the data layer about what extensions are installed
@callback migrations_path() :: String.t() | nil
The path where your migrations are stored
@callback min_pg_version() :: Version.t()
Configure the version of postgres that is being used.
@callback on_transaction_begin(reason :: Ash.DataLayer.transaction_reason()) :: term()
Use this to inform the data layer about the oldest potential postgres version it will be run on.
Must be an integer greater than or equal to 13.
Combining with other tools
For things like Fly.Repo
, where you might need to have more fine grained control over the repo module,
you can use the define_ecto_repo?: false
option to use AshPostgres.Repo
.
Allows overriding a given migration type for all fields, for example if you wanted to always use :timestamptz for :utc_datetime fields
@callback prefer_transaction?() :: boolean()
Whether or not to explicitly start and close a transaction for each action, even if there are no transaction hooks. Defaults to true
.
@callback prefer_transaction_for_atomic_updates?() :: boolean()
Whether or not to explicitly start and close a transaction for each atomic update action, even if there are no transaction hooks. Defaults to false
.
@callback tenant_migrations_path() :: String.t() | nil
The path where your tenant migrations are stored (only relevant for a multitenant implementation)