How does the Garbage Collection algorithm work in modern programming languages?

Official Answer

Garbage collection reclaims memory that the runtime determines is no longer reachable or otherwise reclaimable under its rules. It does not know whether the application still needs an object that remains strongly referenced. An unbounded cache can therefore leak usable memory even in a garbage-collected program.

Tracing collectors start from roots and follow references to identify live objects. Mark-and-sweep reclaims unmarked objects; copying or compacting techniques can move live objects. Generational techniques exploit common object-lifetime patterns. These are techniques that collectors combine differently, not one fixed layout shared by all runtimes.

CPython combines reference counting with collection of reference cycles. That is an implementation characteristic, not a guarantee that every Python implementation destroys every object immediately when it becomes unnecessary. JVM behavior likewise depends on the selected collector and version.

Garbage collection is also not a reliable resource-management plan for files, locks or network connections. Use explicit lifetime mechanisms such as context managers or try/finally. When diagnosing memory growth, inspect retained references and allocation patterns, then measure collection pauses, heap use and throughput before changing runtime settings.

Sources

Python data model and CPython gc interface.