Struct sled_overlay::SledDbOverlay

pub struct SledDbOverlay { /* private fields */ }

An overlay on top of an entire sled::Db which can span multiple trees

Implementations

impl SledDbOverlay

pub fn new(db: &Db) -> Self

Instantiate a new SledDbOverlay on top of a given sled::Db.

pub fn open_tree(&mut self, tree_name: &[u8]) -> Result<(), Error>

Create a new SledTreeOverlay on top of a given tree_name. This function will also open a new tree inside db regardless of if it has existed before, so for convenience, we also provide SledDbOverlay::purge_new_trees in case we decide we don’t want to write the batches, and drop the new trees.

pub fn purge_new_trees(&self) -> Result<(), Error>

Drop newly created trees from the sled database. This is a convenience function that should be used when we decide that we don’t want to apply any cache changes, and we want to revert back to the initial state.

pub fn contains_key(&self, tree_key: &[u8], key: &[u8]) -> Result<bool, Error>

Returns true if the overlay contains a value for a specified key in the specified tree cache.

pub fn get(&self, tree_key: &[u8], key: &[u8]) -> Result<Option<IVec>, Error>

Retrieve a value from the overlay if it exists in the specified tree cache.

pub fn insert( &mut self, tree_key: &[u8], key: &[u8], value: &[u8] ) -> Result<Option<IVec>, Error>

Insert a key to a new value in the specified tree cache, returning the last value if it was set.

pub fn remove( &mut self, tree_key: &[u8], key: &[u8] ) -> Result<Option<IVec>, Error>

Delete a value in the specified tree cache, returning the old value if it existed.

pub fn aggregate(&self) -> Result<(Vec<Tree>, Vec<Batch>), Error>

Aggregate all the current overlay changes into sled::Batch instances and return vectors of sled::Tree and their respective sled::Batch that can be used for further operations. If there are no changes, both vectors will be empty.

pub fn apply(&self) -> Result<(), TransactionError<Error>>

Atomically apply all batches on all trees as a transaction. This function does not perform a db flush. This should be done externally, since then there is a choice to perform either blocking or async IO.