Zarządzanie pamięcią i ograniczanie obciążenia GC

The JVM automatically reclaims unused memory through garbage collection. But GC pauses can hurt latency. Reducing allocations means less GC work and smoother performance.

The JVM heap is split into generations. Most objects die young, so the young generation is collected often and cheaply; survivors move to the old generation.

Clojure's immutable data structures share structure efficiently, but transformations still allocate. In tight loops these allocations add up.

Transients allow temporary, mutable-but-safe building of collections, avoiding intermediate allocations. Convert back with persistent!.

Primitive numbers get boxed into objects when stored in generic collections. Type hints and primitive math avoid this overhead.

For I/O and number crunching, reusing arrays instead of allocating per call reduces GC pressure significantly.

Lazy sequences allocate seq nodes and can hold the head in memory. For pure transformations, prefer reduce or transducers, which avoid intermediate collections.

Modern JVMs offer collectors tuned for different goals:

• G1: balanced default
• ZGC / Shenandoah: very low pause times
• Parallel: max throughput, longer pauses

Enable GC logging to see pause frequency and duration before tuning. Never guess; measure.

Set min and max heap equal to avoid resize pauses, and size based on real working set. Too small causes frequent GC; too large wastes memory and lengthens old-gen collections.

Practical loop: profile to find hot allocations, apply transients/primitives/transducers, measure GC logs, then tune the collector and heap only if needed.

Test your memory tuning knowledge.

You learned to reduce GC pressure in Clojure.

• Use transients, primitives, and transducers to cut allocations
• Reuse buffers in hot paths
• Measure with GC logs, then tune collector and heap size