Maestro v0.3.0 Maestro.Store.Postgres View Source
Ecto+Postgres implementation of the storage mechanism.
Events are never replayed outside of the aggregate's context, so the implementation doesn't support retrieval without an aggregate ID.
Link to this section Summary
Functions
If any transactional projections are present, this function is an extension of
commit_events
that within the same transaction applies all projections to
the store as well. Otherwise, this function dispatches to commit_events
.
Events are validated according to the Event.changeset/1
function. If
successful, events are committed transactionally. In the event of a conflict
on sequence number, the storage mechanism should indicate that the command
could be retried by returning {:error, :retry_command}
. The Aggregate
's
command lifecycle will see the conflict and update the aggregate's state
before attempting to evaluate the command again. This allows for making
stricter evaluation rules for commands. If the events could not be committed
for any other reason, the storage mechanism should raise an appropriate
exception.
Snapshots are committed iff the proposed version is newer than the version already stored. This allows disconnected nodes to optimistically write their snapshots and still have a single version stored without conflicts.
Events are retrieved by aggregate_id and with at least a minimum sequence
number, seq
. They should be ordered by sequence number to ensure that
aggregates always process events in the same order.
Snapshots can also be retrieved by aggregate_id and with at least a minimum
sequence number, seq
.
Link to this section Functions
commit_all(events, projections) View Source
If any transactional projections are present, this function is an extension of
commit_events
that within the same transaction applies all projections to
the store as well. Otherwise, this function dispatches to commit_events
.
Callback implementation for Maestro.Store.Adapter.commit_all/2
.
commit_events(events) View Source
Events are validated according to the Event.changeset/1
function. If
successful, events are committed transactionally. In the event of a conflict
on sequence number, the storage mechanism should indicate that the command
could be retried by returning {:error, :retry_command}
. The Aggregate
's
command lifecycle will see the conflict and update the aggregate's state
before attempting to evaluate the command again. This allows for making
stricter evaluation rules for commands. If the events could not be committed
for any other reason, the storage mechanism should raise an appropriate
exception.
Callback implementation for Maestro.Store.Adapter.commit_events/1
.
commit_snapshot(s) View Source
Snapshots are committed iff the proposed version is newer than the version already stored. This allows disconnected nodes to optimistically write their snapshots and still have a single version stored without conflicts.
Callback implementation for Maestro.Store.Adapter.commit_snapshot/1
.
get_events(aggregate_id, min_seq, map) View Source
Events are retrieved by aggregate_id and with at least a minimum sequence
number, seq
. They should be ordered by sequence number to ensure that
aggregates always process events in the same order.
Additional option(s):
:max_sequence
(integer): a hard upper limit on the sequence number. This is useful when attempting to recreate a past state of an aggregate.
Callback implementation for Maestro.Store.Adapter.get_events/3
.
get_snapshot(aggregate_id, min_seq, map) View Source
Snapshots can also be retrieved by aggregate_id and with at least a minimum
sequence number, seq
.
Additional option(s):
:max_sequence
(integer): a hard upper limit on the sequence number. This is useful when attempting to recreate a past state of an aggregate.
Callback implementation for Maestro.Store.Adapter.get_snapshot/3
.