Struct QueryConfig 

pub struct QueryConfig {
    pub max_execution_time: Duration,
    pub forced_read_path: Option<ReadPathMode>,
    pub max_result_rows: u64,
    pub max_result_bytes: u64,
    pub plan_cache_size: usize,
    pub enable_optimization: bool,
    pub parallel: ParallelQueryConfig,
    pub query_cache_size: Option<usize>,
    pub query_parallelism: Option<usize>,
    pub analyze_iterations: Option<usize>,
}

Query engine configuration

Fields

max_execution_time: Duration

Maximum query execution time

forced_read_path: Option<ReadPathMode>

Force the SELECT access-path decision (issue #1918).

None (the default) leaves routing to the per-query classifier and the CQLITE_READ_PATH env knob; Some(mode) forces that mode and takes precedence over the env var. A test/debug control — see ReadPathMode. #[serde(default)] keeps configs serialized before this field existed deserializing successfully (absent = None).

max_result_rows: u64

Maximum number of rows to return in a result set.

A secondary safety valve, retained for defense-in-depth (issue #1582). The primary guard on a materialized result is now max_result_bytes: a row count is the wrong unit because 1M skinny rows can fit comfortably while 100k wide rows blow the <128MB memory target. Still load-bearing: the materializing SELECT path enforces this row-count ceiling alongside the byte budget (lowering it makes a wide-row-count result trip even under the byte budget), so it is a real knob, not decoration.

max_result_bytes: u64

Byte ceiling on a MATERIALIZED result set (issue #1582 / D6).

While the SELECT executor collects a materialized Vec<QueryRow>, it tracks a running estimate of the result’s logical size (via the shared crate::memory::estimate_value_size estimator) and fails with crate::Error::ResultTooLarge once this ceiling is crossed — telling the caller to add a LIMIT or use the streaming API. This is the correct-unit primary guard; max_result_rows remains as a secondary valve. Streaming queries are bounded by their channel buffer, so this budget does not apply to them.

Default: DEFAULT_MAX_RESULT_BYTES (64 MiB). Chosen well below the project’s <128MB process memory target: the estimator measures logical content bytes and does not count per-row container overhead (HashMap<Arc<str>, Value> slots, String/Vec capacity slack, row metadata), which in practice roughly doubles real heap use — so a 64 MiB logical ceiling keeps a fully-materialized result comfortably inside the process budget while leaving headroom for readers, caches, and decode buffers.

plan_cache_size: usize

Query plan cache size

enable_optimization: bool

Enable query optimization

parallel: ParallelQueryConfig

Parallel query execution configuration

query_cache_size: Option<usize>

Query cache size (for plan caching)

query_parallelism: Option<usize>

Query parallelism thread count

analyze_iterations: Option<usize>

Number of iterations for query analysis

Trait Implementations

impl Clone for QueryConfig

fn clone(&self) -> QueryConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for QueryConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for QueryConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for QueryConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for QueryConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.