Skip to main content

Crate laminar_core

Crate laminar_core 

Source
Expand description

§LaminarDB Core

The core streaming engine for LaminarDB, implementing the Ring 0 (hot path) components.

This crate provides:

  • Reactor: Single-threaded event loop with zero allocations
  • Operators: Streaming operators (map, filter, window, join)
  • State Store: Lock-free state management with sub-microsecond lookup
  • Time: Event time processing, watermarks, and timers

§Design Principles

  1. Zero allocations on hot path - Uses arena allocators
  2. No locks on hot path - SPSC queues, lock-free structures
  3. Predictable latency - < 1μs event processing
  4. CPU cache friendly - Data structures optimized for cache locality

§Example

use laminar_core::{Reactor, Config};

let config = Config::default();
let mut reactor = Reactor::new(config)?;

// Run the event loop
reactor.run()?;

Re-exports§

pub use reactor::Reactor;
pub use reactor::ReactorConfig;

Modules§

aggregation
Cross-partition aggregation.
alloc
Zero-allocation enforcement for Ring 0 hot path.
budget
Task Budget Enforcement
checkpoint
Distributed checkpoint barrier protocol.
compiler
Plan compiler infrastructure for Ring 0 event processing.
dag
DAG Pipeline Topology
detect
Automatic Hardware/Software Feature Detection
error_codes
Structured error code registry (LDB-NNNN) and Ring 0 hot path error type. LaminarDB structured error code registry.
index
Secondary index support using redb. Secondary index support using redb.
io_uring
io_uring Advanced Optimization
lookup
Lookup table types and predicate pushdown.
mv
Cascading Materialized Views
numa
NUMA-Aware Memory Allocation
operator
Operator Module
reactor
Reactor Module
sink
Exactly-Once Sinks
state
State Store Module
streaming
Streaming API
subscription
Reactive Subscription System
time
Time Module
tpc
Thread-Per-Core (TPC) Module
xdp
XDP/eBPF Network Optimization

Macros§

hot_path
Macro to mark a function as hot path.

Enums§

Error
Error types for laminar-core

Type Aliases§

Result
Result type for laminar-core operations