- Bugfix: Class method instrumentation now properly handles inherited methods when parent classes are instrumented first.
- Breaking: Changed instrument_estimator to instrument_instance
- Breaking: Instrument API now includes the class/instance on which instrumentation is applied
- Bugfix: Class instrumentation now properly handles inherited methods, properties,and delegates.
- Bugfix: _get_estimator_classes method now excludes inherited classes.
- Fix a bug with full uninstrumentation
- Reverting from 0.8.0: Change the order of instrumentation so that an estimator is instrumented after its attributes
- Skip instrumentation recursion of estimator attributes which share a name with a method intended to be instrumented
- Change the order of instrumentation so that an estimator is instrumented after its attributes
- Removed numpy as dependency since it is already a dependency of sklearn
- Modified instrumentation to prevent double instrumentation of methods
- Updated dependencies for some instruments
- Removed an extra type exclusion check in instrumentor.
- Exclude bytes from instrumentation
- Move ignored types from instrumentor into
DEFAULT_EXCLUDE
- Add
predict_log_proba
and_predict_log_proba
toDEFAULT_METHODS
- Exclude enums from instrumentation
- Avoid instrumentor recursion against instrumentor-created attributes
- Add instrument for cProfile
- Add instruments for prometheus, datadog, opentelemetry, statsd, memory-profiler, pyinstrument
- Add class instrumentation
- Add instrumentation of properties on classes (but not on instances)
- Refactor instruments into classes with base class (breaking)
- Rename module: instrumentation -> instruments (breaking)
- Change args decorator/decorator_kwargs -> instrument/instrument_kwargs (breaking)
- Use module loggers instead of root logger
- First release