Chronic - Python Instrumentation
Chronic is halfway between a simple timer and a profiler. Add decorators or wrap code in with statements to get the execution time. Chronic keeps track of the call hierarchy to tell you what timed blocks executed within other timed blocks. Attach an event listener to log your timings however you want.
Installation
pip install chronic
Usage
To time a block of code, you can use either the @time decorator or the Timer context manager.
import chronic
@chronic.time
def long_running_method():
# Do something here
with chronic.Timer('crunch_numbers'):
# Crunch numbers herechronic.timings
At any point, you can get the information about completed timing info from chronic.timings. This is a proxy to state held in a thread local about all timed blocks captured so far.
>>> import chronic
>>> chronic.timings
{
'long_running_method': {
'total_elapsed': 3.068,
'average_elapsed': 3.068,
'count': 1
},
'crunch_numbers': {
'total_elapsed': 1.884,
'average_elapsed': 1.884,
'count': 1
}
}chronic.timings is a dict with keys for each timed block of code. If you use the context manager, you must specify a name. When using the decorator, you may optionally specify the name, or chronic will use theh function name as a default.
Each timing is itself a dict with the following keys:
-
total_elapsed: the elapsed execution time of the code (including all subtimings) for all runs of this block. The unit is seconds by default. If you pass in your own clock function, the unit is whatever the unit of the clock. -
count: the number of times the timed block was run. -
average_elapsed: the average elapsed time for each run of this block. -
timings: a dict of all subtimings that were completed while inside this block.
So if you're using Django and want to log all timings to MongoDB, all you need to do is add middleware that looks something like this:
import chronic
from pymongo import MongoClient
mongo_client = MongoClient()
class ProfilingMiddleware(object):
def process_response(self, request, response):
mongo_client.app.timings.insert(chronic.timings)
return responsechronic.stack
chronic.stack is a tuple of the names of all timed blocks in the call stack above the current context.
>>> import chronic
>>> @chronic.time
>>> def time_two():
>>> with chronic.Timer('block2'):
>>> print chronic.stack
>>> time_two()
('time_two', 'block2')chronic.post_timing
Finally, you can install signal handlers to be called on the completion of all timings. Here's how you might send all timing data to statsd:
>>> import chronic
>>> from statsd import StatsClient
>>> statsd = StatsClient()
>>> def send_to_statsd(elapsed, timings, stack):
>>> statsd.timing('.'.join(stack), elapsed * 1000)
>>> chronic.post_timing.connect(send_to_statsd)Examples
import chronic
from pprint import pprint
@chronic.time
def time_one():
pass
@chronic.time
def time_two():
with chronic.Timer('block2'):
print chronic.stack
# ('time_two', 'block2')
pprint(chronic.timings)
# prints local view of timings
# {'block2': {'average_elapsed': 1.0967254638671875e-05,
# 'count': 1,
# 'total_elapsed': 1.0967254638671875e-05}}
with chronic.Timer('block1'):
time_one()
pprint(chronic.timings)
def print_done(elapsed, timings, stack):
print stack
pprint(timings)
time_one()
chronic.post_timing.connect(print_done)
time_one()
# []
# 9.5367431640625e-07
# {'average_elapsed': 9.5367431640625e-07,
# 'count': 2,
# 'total_elapsed': 1.9073486328125e-06}
chronic.post_timing.disconnect(print_done)
time_two()
# ['time_two', 'block2']
# {'block2': {'average_elapsed': 5.9604644775390625e-06,
# 'count': 1,
# 'total_elapsed': 5.9604644775390625e-06}}This software is licensed under the MIT license. See the details in the file called LICENSE.
