charcoal 0.2.0

Installing

$ pip install charcoal

Use it like so:

import charcoal

my_client = charcoal.StatsClient(prefix, host, port)

For development, the client also provides a disabled kwarg, so you do not have change anything in the code when you want to not send stats:

my_client = charcoal.StatsClient(prefix, host, port, disabled=True)

By itself, the client does not provide much use, but it does provide easy ways to get specific sub-clients, such as timers, counters, etc.

Timing

timer = my_client.timer('my-timer-name').start()

this_takes_a_while()
timer.intermediate('first-pass')

this_takes_a_while()
timer.intermediate('second-pass')
timer.stop()

When you have end up with a measurement from somewhere else, perhaps from an external service, you can also send that, using the .send function on the timer class, in fact this is what the higler level functions above use, behind the scenes.

timer = my_client.timer('my-timer-name')
timer.send('db-call', 12.5)

Counting

counter = my_client.counter('my-counter-name')

counter.increment('some-value', 10)
counter.decrement('some-other-value', 10)

The counter can even be fed a dict like object, such as the Counter from the standard library and send the stats as a single message.

pre_counted = {
    'a-name': 5,
    'another-name': 10,
}

counter.from_mapping(pre_counted)

Gauges

For setting the current value.

gauge = my_client.gauge('my-gauge')

gauge.set('a-name', 10)
gauge.update('a-name', 10)

Sets

For counting unique events, such as unique users on a page.

visitors = my_client.set('visitors')
visitors.add('ids', user.id)

Custom

In case the server you are using supports more metric types than this library, you can send custom metrics:

metric_to_send = 'metric.name:{value}|{type_suffix}'.format(
    value=str(value),
    type_suffix=type_suffix,
)

my_client.send(metric_to_send)

The prefix given to the client when creating it, is then prepended to the metric name, encoded, and sent to the server.

Currently it can even accept multiple metrics in one go:

my_client.send(metric_to_send, other_metric_to_send)