Batch operations support for AshScylla using ScyllaDB's BATCH statements.
ScyllaDB/Cassandra supports batch operations for executing multiple CQL statements in a single request.
Synchronous Batches
For small batches or when ordering matters, use batch_insert/3,
batch_update/3, or batch_delete/3.
Async Partition-Aware Batches
For large bulk operations, use batch_insert_async/4. This function:
- Groups records by partition key (safe for ScyllaDB)
- Executes sub-batches in parallel using
Task.async_stream - Respects ScyllaDB's recommendation to avoid cross-partition BATCH statements
Examples
# Synchronous batch
statements = [
{"INSERT INTO users (id, name) VALUES (?, ?)", [id1, "Alice"]},
{"INSERT INTO users (id, name) VALUES (?, ?)", [id2, "Bob"]}
]
DataLayer.Batch.batch_insert(repo, statements)
# Async partition-aware batch
DataLayer.Batch.batch_insert_async(repo, statements, max_concurrency: 8)Summary
Functions
Executes a batch of DELETE statements.
Executes a batch of INSERT statements.
Executes batch inserts asynchronously, grouped by partition key.
Executes a batch of UPDATE statements.
Extracts the partition key values from a record for a given resource.
Functions
Executes a batch of DELETE statements.
Executes a batch of INSERT statements.
@spec batch_insert_async(module(), [{String.t(), list()}], keyword()) :: {:ok, [term()]} | {:error, term()}
Executes batch inserts asynchronously, grouped by partition key.
This is the recommended approach for large bulk inserts in ScyllaDB. Records are grouped by their partition key values, and each group is executed as a separate batch in parallel. This avoids the cross-partition BATCH anti-pattern.
Options
:max_concurrency- Maximum number of concurrent batch executions (defaults toSystem.schedulers_online())
Examples
statements = [
{"INSERT INTO users (id, name) VALUES (?, ?)", [id1, "Alice"]},
{"INSERT INTO users (id, name) VALUES (?, ?)", [id2, "Bob"]},
# ... hundreds more
]
DataLayer.Batch.batch_insert_async(repo, statements, resource: MyApp.User)Executes a batch of UPDATE statements.
Extracts the partition key values from a record for a given resource.
Returns a map of partition key column names to their values.