Trait mls_rs::GroupStateStorage

source · 
pub trait GroupStateStorage: Send + Sync {
    type Error: IntoAnyError;

    // Required methods
    fn state(&self, group_id: &[u8]) -> Result<Option<Vec<u8>>, Self::Error>;
    fn epoch(
        &self,
        group_id: &[u8],
        epoch_id: u64,
    ) -> Result<Option<Vec<u8>>, Self::Error>;
    fn write(
        &mut self,
        state: GroupState,
        epoch_inserts: Vec<EpochRecord>,
        epoch_updates: Vec<EpochRecord>,
    ) -> Result<(), Self::Error>;
    fn max_epoch_id(&self, group_id: &[u8]) -> Result<Option<u64>, Self::Error>;
}
Expand description

Storage that can persist and reload a group state.

A group state is recorded as a combination of the current state (represented by the GroupState trait) and some number of prior group states (represented by the EpochRecord trait). This trait implements reading and writing group data as requested by the protocol implementation.

§Cleaning up records

Group state will not be purged when the local member is removed from the group. It is up to the implementer of this trait to provide a mechanism to delete records that can be used by an application.

Required Associated Types§

source

type Error: IntoAnyError

Required Methods§

source

fn state(&self, group_id: &[u8]) -> Result<Option<Vec<u8>>, Self::Error>

Fetch a group state from storage.

source

fn epoch( &self, group_id: &[u8], epoch_id: u64, ) -> Result<Option<Vec<u8>>, Self::Error>

Lazy load cached epoch data from a particular group.

source

fn write( &mut self, state: GroupState, epoch_inserts: Vec<EpochRecord>, epoch_updates: Vec<EpochRecord>, ) -> Result<(), Self::Error>

Write pending state updates.

The group id that this update belongs to can be retrieved with GroupState::id. Prior epoch id values can be retrieved with EpochRecord::id.

The protocol implementation handles managing the max size of a prior epoch cache and the deleting of prior states based on group activity. The maximum number of prior epochs that will be stored is controlled by the Preferences::max_epoch_retention function in mls_rs. value. Requested deletes are communicated by the delete_epoch_under parameter being set to Some.

§Warning

It is important to consider error recovery when creating an implementation of this trait. Calls to write should optimally be a single atomic transaction in order to avoid partial writes that may corrupt the group state.

source

fn max_epoch_id(&self, group_id: &[u8]) -> Result<Option<u64>, Self::Error>

The EpochRecord::id value that is associated with a stored prior epoch for a particular group.

Implementors§