TimeSeries

timeseries is a set django application tools designed to facilitate the collation and maintenance of timeseries data.

Requirements

Django versions 1.8+ are supported for projects running on PostgreSQL.

Installation

pip install django-timeseries

Usage

from datetime import timedelta
from django.db import models
from timeseries.utils import TimeSeriesModel, TimeSeriesManager


class Ad(models.Model):

    objects = TimeSeriesManager()


class RawAdData(TimeSeriesModel):

    TIMESERIES_INTERVAL = timedelta(days=1)  # update daily N.B integers in seconds also work

    NOT_AVAILABLE = -1

    ad = models.ForeignKey(Ad, related_name='rawdata')

    views = models.BigIntegerField(default=NOT_AVAILABLE)
    clicks = models.BigIntegerField(default=NOT_AVAILABLE)


class MonthlyAdReport(TimeSeriesModel):

    TIMESERIES_INTERVAL = timedelta(days=28)

    ad = models.ForeignKey(Ad, related_name='monthlyreports')

    avg_view_growth = models.FloatField()
    avg_click_growth = models.FloatField()


def ad_data_collector(queryset):
    """
        should return an iterable that yields dictionaries of data
        needed to successfully create a RawAdData instance
    """
    return ...


def report_data_collector(queryset):
    """
        should return an iterable that yields dictionaries of data
        needed to successfully create a MonthlyAdReport instance
    """
    return ...


# in a shell
>>> Ad.objects.update_timeseries('rawdata', ad_data_collector)
# this return the results of a bulk_create call from the RawAdData manager
# for ads that hadn't been updated in the last day

>>> Ad.objects.update_timeseries('monthlyreports', report_data_collector)
# this return the results of a bulk_create call from the MonthlyAdReport
# manager for ads that hadn't had a report generated in the last 28 days

>>> ad = Ad.objects.prefetch_latest('rawdata', 'monthlyreports').first()
>>> print ad.latest_rawaddata, ad.latest_monthlyreports

TimeSeries QuerySet Methods

timeseries.utils.TimeSeriesQuerySet

Adds 4 main methods to the Django QuerySet API that can be used to update and maintain timeseries data. These methods include:

• prefetch_latest
• filter_outdated
• last_updated
• update_timeseries

update_timeseries

Inputs: related_name, collector, optional force

Returns: list of instatiated related models.

Updates the queryset's related model table (as given by related_name) using a provider "collector" callable.

"collector" must take a queryset of the referenced models as its only argument. It must also return an iterable of dictionaries that can be used to construct and save instances of the related model.

N.B. Only instances that have outdated data will be updated unless explicitly forced using the "force" keyword argument.

filter_outdated

Inputs: *related_names

Returns: queryset

Returns a queryset that will yield the model instances that have "outdated" data associated to reverse related model as given by the specified related_name.

last_updated

Inputs: *related_names

Returns: queryset

Annotates the created timestamp of the latest related instance as given by the reverse relation's related_name.

Usage:

ad = Ad.objects.last_updated('rawdata').first()
# assuming there's data related to ad
print ad.rawdata_last_updated
# this will print the timestamp of when the associated data was
# last updated

prefetch_latest

Inputs: *related_names

Returns: queryset

Exposes the latest associated reverse relation.

Usage:

ad = Ad.objects.prefetch_latest('rawdata', 'monthlyreports').first()
print ad.latest_rawaddata, ad.latest_monthlyreports
# this will print the reprs of the latest associated data
# instances

Other Utilities

LatestQ

timeseries.utils

Inputs: related_name, **kwargs

Returns: django.db.models.Q instance

Constructs a django.db.models.Q instance that allows queries to be executed against the latest associated reverse relation.

N.B. this method is designed to be used in conjunction with timeseries.utils.TimeSeriesQuerySet.last_updated.

Usage:

Ad.objects.last_updated('rawdata').filter(
    LatestQ('rawdata', views__gt=1000)
)