Skip to content

geopozo/logistro

Repository files navigation

logistro (lo-hī-stro)

logistro is an extremely light addition to logging, providing sensible defaults.

It also includes getPipeLogger() which can be passed to Popen() so that its stderr is piped to the already thread-safe logging library.

Which logger level should I use?

debug2 I use if I'm in a situation where it's okay to dump large amounts of information. I'm planning on scrolling.

info I will maybe use at the beginning and end of functions called.

debug Might print stuff from inside functions or it might print shortened versioning of debug2.

Warning, Error, Critical, Exception, these are all more obvious.

Quickstart

import logistro

logger = logistro.getLogger(__name__)

logger.debug2(...) # new!
logger.debug(...) # or debug1()
logger.info(...)
logger.warning(...)
logger.error(...)
logger.critical(...)
logger.exception(...) # always inside except:

# For subprocesses:

pipe, logger = logistro.getPipeLogger(__name__+"-subprocess")
subprocess.Popen(cli_command, stderr=pipe)
os.close(pipe) # eventually

CLI Flags

  • --logistro-level DEBUG|DEBUG2|INFO|WARNING|ERROR|CRITICAL
  • --logistro-human (default)
  • --logistro-structured which outputs JSON

The help for CLI commands can be included in your program:

parser = argparse.ArgumentParser(parents=[logistro.parser])

Functions

  • logistro.set_structured()
  • logistro.set_human()

Generally, they must be called before any other logging call (See note below).

Additionally

logistro.betterConfig(...) applies our formats and levels. It accepts the same arguments as logging.basicConfig(...) except format=, which it ignores. It is better to call this early in a multithread program.

logistro.getLogger(...) will ensure betterConfig().

You can use our two formatters manually instead:

  • human_formatter
  • structured_formatter

Changing Logger Formatter Mid-Execution

With a typical setup, calling set_structured() or set_human() and then logistro.coerce_logger(logistro.getLogger()) will change the format.

See the tech note for an intro into the complexities of logging.