wagnerflo / noblklog Goto Github PK

More complex logging configurations might create instances of the same Handler type multiple times. For example to support different Formatters.

The current implementation will create one queue and asyncio task per instance even if the share they same file descriptor. This is necessary and might even lead to problems: Instance hijacking each others writers.

Improve this by storing per-filedescriptor references to queue and task on the class. Look these up instead of the instance ones and make sure to cleanup in the destructor.

This problem is not trigger on systems with low throughput / small messages, since these will get sent on the fast-path.