Library for efficiently adding analytics to your project.


License
Apache-2.0
Install
pip install analytics==0.5.4

Documentation

py-analytics

https://secure.travis-ci.org/numan/py-analytics.png?branch=master

Py-Analytics is a library designed to make it easy to provide analytics as part of any project.

The project's goal is to make it easy to store and retrieve analytics data. It does not provide any means to visualize this data.

Currently, only Redis is supported for storing data.

Install

You can install the latest official stable version using pypi:

>>> pip install analytics

Or get the latest version directly from github:

>>> pip install -e git+https://github.com/numan/py-analytics.git#egg=analytics

Requirements

Required

Requirements should be handled by setuptools, but if they are not, you will need the following Python packages:

  • nydus
  • redis
  • dateutil

Optional

  • hiredis

analytics.create_analytic_backend

Creates an analytics object that allows to to store and retrieve metrics:

>>> from analytics import create_analytic_backend
>>>
>>> analytics = create_analytic_backend({
>>>     'backend': 'analytics.backends.redis.Redis',
>>>     'settings': {
>>>         'defaults': {
>>>             'host': 'localhost',
>>>             'port': 6379,
>>>             'db': 0,
>>>         },
>>>         'hosts': [{'db': 0}, {'db': 1}, {'host': 'redis.example.org'}]
>>>     },
>>> })

Internally, the Redis analytics backend uses nydus to distribute your metrics data over your cluster of redis instances.

There are two required arguements:

  • backend: full path to the backend class, which should extend analytics.backends.base.BaseAnalyticsBackend
  • settings: settings required to initialize the backend. For the Redis backend, this is a list of hosts in your redis cluster.

Example Usage

from analytics import create_analytic_backend
import datetime

analytics = create_analytic_backend({
    "backend": "analytics.backends.redis.Redis",
    "settings": {
        "hosts": [{"db": 5}]
    },
})

year_ago = datetime.date.today() - datetime.timedelta(days=365)

#create some analytics data
analytics.track_metric("user:1234", "comment", year_ago)
analytics.track_metric("user:1234", "comment", year_ago, inc_amt=3)
#we can even track multiple metrics at the same time for a particular user
analytics.track_metric("user:1234", ["comments", "likes"], year_ago)
#or track the same metric for multiple users (or a combination or both)
analytics.track_metric(["user:1234", "user:4567"], "comment", year_ago)

#retrieve analytics data:
analytics.get_metric_by_day("user:1234", "comment", year_ago, limit=20)
analytics.get_metric_by_week("user:1234", "comment", year_ago, limit=10)
analytics.get_metric_by_month("user:1234", "comment", year_ago, limit=6)

#create a counter
analytics.track_count("user:1245", "login")
analytics.track_count("user:1245", "login", inc_amt=3)

#retrieve multiple metrics at the same time
#group_by is one of ``month``, ``week`` or ``day``
analytics.get_metrics([("user:1234", "login",), ("user:4567", "login",)], year_ago, group_by="day")
>> [....]

#set a metric count for a day
analytics.set_metric_by_day("user:1245", "login", year_ago, 100)

#sync metrics for week and month after setting day
analytics.sync_agg_metric("user:1245", "login", year_ago, datetime.date.today())

#retrieve a count
analytics.get_count("user:1245", "login")

#retrieve a count between 2 dates
analytics.get_count("user:1245", "login", start_date=datetime.date(month=1, day=5, year=2011), end_date=datetime.date(month=5, day=15, year=2011))

#retrieve counts
analytics.get_counts([("user:1245", "login",), ("user:1245", "logout",)])

#clear out everything we created
analytics.clear_all()

BACKWARDS INCOMPATIBLE CHANGES

V0.6.0

  • This version introduces prefixes. Any old analytics data will be unaccessable.

v0.5.2

  • get_metric_by_day, get_metric_by_week and get_metric_by_month return series as a set of strings instead of a list of date/datetime objects

TODO

  • Add more backends possibly...?
  • Add an API so it can be deployed as a stand alone service (http, protocolbuffers, ...)