To add some context, the repository for my plugin can be found here: https://github.com/j19sch/pytest-instrument. And the relevant code here: https://github.com/j19sch/pytest-instrument/blob/add-logger-fixture/pytest_instrument/plugin.py

What I'm trying to achieve:

• exposing a structlog logger to be used by tests
• this structlog logger wraps the stdlib logger, so pytest captures the log records and can sent them to stdout
• these log records are also being written to file as json objects
• additional json objects about pytest's setup/test call/teardown are written to the same file, but they need to bypass stdlib logging, so they don't get captured by pytest

As mentioned in the issue description, I got this to work by using structlog.configure(), but my concern is that this may affect any structlog configuration in the code under test, which I want to avoid.

wrap_logger() wraps a logger, but you're passing None as a logger.

If you pass in a logger, it works:

In [2]: import logging
   ...: import structlog
   ...:
   ...: my_logger = structlog.wrap_logger(logging.getLogger(), logger_factory=structlog.stdlib.LoggerFactory(),
   ...:                                   wrapper_class=structlog.stdlib.BoundLogger)
   ...:
   ...: my_logger.setLevel("INFO")
   ...: my_logger.addHandler(logging.StreamHandler())
   ...:
   ...: my_logger.info("this log record is displayed")
   ...: my_logger.debug("this log record isn't")
2019-12-03 15:46.00 this log record is displayed   logger_factory=<structlog.stdlib.LoggerFactory object at 0x7f97415645e0>

If you go the configure() route and call structlog.get_logger(), structlog will call logging.getLogger() for you.

Thank you! That does indeed work for setting the level, etc.

Unfortunately, it also results in the log records containing filename _base.py, funcName _proxy_to_logger, and lineno 192. If I take the configure() route, this is not the case and the values match the file, function and line number of where I create the log record.