Logging

OptunaSharp emits progress and diagnostic messages — new best values, independent-sampling warnings, deprecations — through its own logging module, a port of optuna.logging in OptunaSharp.Logging. You control how much it says and where the output goes.

Setting verbosity

using OptunaSharp;
using OptunaSharp.Logging;

// Quiet everything below warnings (a common choice for scripted runs).
OptunaLogging.SetVerbosity(LogLevel.Warning);

var study = Optuna.CreateStudy();
study.Optimize(Objective, nTrials: 100);   // no per-trial INFO chatter

LogLevel mirrors Python's numeric levels: Debug = 10, Info = 20, Warning = 30, Error = 40, Critical = 50 (with Warn/Fatal aliases and NotSet = 0). Read the current level with OptunaLogging.GetVerbosity().

Handlers

By default, records go to a built-in handler that writes to standard error. Turn it off, back on, or add your own sinks:

OptunaLogging.DisableDefaultHandler();          // stop the built-in stderr output
OptunaLogging.AddHandler(new StreamLogHandler(myWriter));   // route records elsewhere
// ...
OptunaLogging.EnableDefaultHandler();

Implement ILogHandler to bridge into your application's logging framework (Serilog, Microsoft.Extensions.Logging, a file, a test sink). GetLogger(string name) returns the Logger used internally, should you need to emit records under the same configuration.

Threads. Under nJobs several trials log at once. As in Python, where Handler.handle holds the handler's lock around emit, OptunaSharp gives each handler one record at a time under a lock of its own, so a handler needs no lock for what only its Emit touches; different handlers still run at the same time. Emit runs on the thread that logged and other threads that log wait for it, so keep it quick, and do not throw from it, call OptunaSharp from it, or take the Python GIL in it. StreamLogHandler writes each record, its exception included, in one write and flushes it, as Python's StreamHandler does, so records from parallel trials never interleave even over a plain StreamWriter. Adding a handler that is already registered does nothing, as in Python.

Divergence. Optuna's enable_propagation / disable_propagation — which toggle whether records bubble up to Python's root logger — have no analog here: OptunaSharp is not built on the Python logging package, so there is no root logger to propagate to. Route records into another framework by adding your own ILogHandler (or removing the default handler) instead.

Relationship to Optuna

Optuna (Python)	OptunaSharp
optuna.logging.set_verbosity(optuna.logging.WARNING)	OptunaLogging.SetVerbosity(LogLevel.Warning)
optuna.logging.get_verbosity()	OptunaLogging.GetVerbosity()
optuna.logging.disable_default_handler()	OptunaLogging.DisableDefaultHandler()
optuna.logging.enable_default_handler()	OptunaLogging.EnableDefaultHandler()

See also

Getting Started · Study and Trials · the API reference.