Skip to content

Latest commit

 

History

History
229 lines (163 loc) · 10.3 KB

Readme.rst

File metadata and controls

229 lines (163 loc) · 10.3 KB

Django Pivot-Tables

This package provides utilities for turning Django Querysets into Pivot-Tables and Histograms by letting your database do all the heavy lifting.

Examples

I am going to shamelessly lift examples from the wikipedia page referenced in the header. Here is part of the table of shirt sales:

Region Gender Style Ship Date Units Price Cost
East Boy Tee 1/31/2005 12 11.04 10.42
East Boy Golf 1/31/2005 12 13 12.6
East Boy Fancy 1/31/2005 12 11.96 11.74
East Girl Tee 1/31/2005 10 11.27 10.56
East Girl Golf 1/31/2005 10 12.12 11.95
East Girl Fancy 1/31/2005 10 13.74 13.33
West Boy Tee 1/31/2005 11 11.44 10.94
West Boy Golf 1/31/2005 11 12.63 11.73
West Boy Fancy 1/31/2005 11 12.06 11.51
West Girl Tee 1/31/2005 15 13.42 13.29
West Girl Golf 1/31/2005 15 11.48 10.67
Etc.            

We might want to know how many Units did we sell in each Region for every Ship Date? And get a result like:

Region 1/31/2005 2/1/2005 2/2/2005 2/3/2005 2/4/2005
East 66 80 102 93 114
North 86 91 95 88 107
South 73 78 84 76 91
West 92 103 111 104 123

It takes 3 quantities to pivot the original table into the summary result, two columns and an aggregate of a third column. In this case the two columns are Region and Ship Date, the third column is Units and the aggregate is Sum

Basic usage

The pivot function

Pivot tables are generated by the pivot function, which takes a Model and 3 attribute names, to make a pivot table like the example above:

>>> pivot_table = pivot(ShirtSales, 'shipped', 'region', 'units')

The result is a list of dictionaries. Each dictionary has a key for the row ('shipped' dates in this case) and a key for every value of the column ('region' in this case).

>>> for record in pivot_table:
...     print(record)
... {u'West': 59, 'shipped': datetime.date(2004, 12, 24), u'East': 71, u'North': 115, u'South': 56}
... {u'West': 55, 'shipped': datetime.date(2005, 1, 31), u'East': 65, u'North': 121, u'South': 66}
... {u'West': 56, 'shipped': datetime.date(2005, 2, 1), u'East': 62, u'North': 124, u'South': 68}
... {u'West': 56, 'shipped': datetime.date(2005, 2, 2), u'East': 59, u'North': 127, u'South': 71}
... {u'West': 66, 'shipped': datetime.date(2005, 3, 1), u'East': 55, u'North': 131, u'South': 65}
... {u'West': 68, 'shipped': datetime.date(2005, 3, 2), u'East': 56, u'North': 130, u'South': 62}
... {u'West': 71, 'shipped': datetime.date(2005, 4, 3), u'East': 56, u'North': 130, u'South': 59}
... {u'West': 65, 'shipped': datetime.date(2005, 5, 6), u'East': 66, u'North': 120, u'South': 55}

The first argument can be a Model, QuerySet, or Manager. This allows you to generate a pivot table filtered by another column. For example, you may want to know how many units were sold in each region for every shipped date, but only for Golf shirts:

>>> pivot_table = pivot(ShirtSales.objects.filter(style='Golf'), 'region', 'shipped', 'units')

The pivot function takes an optional parameter for how to aggregate the data. For example, instead of the total units sold in each region for every ship date, we might be interested in the average number of units per order. Then we can pass the Avg aggregation function

>>> from django.db.models import Avg
>>> pivot_table = pivot(ShirtSales, 'region', 'shipped', 'units', aggregation=Avg)

The pivot function can optionally include a Total column, containing all the data aggregated to a single column:

>>> pivot_table = pivot(ShirtSales.objects.filter(style='Golf'), 'region', 'shipped', 'units', include_total=True)

If your data is stored across multiple tables, use Django's double underscore notation to traverse foreign key relationships. For example, instead of the ShirtSales model having a region attribute, it might have a foreign key to a Store model, which in turn has a foreign key to a Region model, which has an attribute called name. Then our pivot call looks like

>>> pivot_table = pivot(ShirtSales, 'store__region__name', 'shipped', 'units')

It's also possible that the data column we are aggregating over should be a computed column. In our example ShirtSales model we are storing the number of units and the price per unit, but not the total cost of the order. If we want to know the average order size in dollars in each region for every ship date, we can pivot the ShirtSales table:

>>> from django.db.models import F, Avg
>>> pivot_table = pivot(ShirtSales, 'region', 'shipped', F('units') * F('price'), Avg)

If the rows should be grouped on a compound column, for example, you want to know how many Units were sold on each ship date not just split by region, but the combination of region and gender, you can pass a list to the first argument:

>>> pivot_table = pivot(ShirtSales, ['region', 'gender'], 'shipped', 'units')

To change the way the row keys are displayed, a display_transform function can be passed to the pivot function. display_transform is a function that takes an object and returns a string. For example, instead of getting the results with North, East, South, and West for the regions you want them all lower cased, you can do the following

>>> def lowercase(s):
>>>     return s.lower()
>>> pivot_table = pivot(ShirtSales, 'region', 'shipped', 'units', display_transform=lowercase)

The display_transform option is also useful if your column attribute is not a hashable type. Since it will be used as a key in a dictionary, you need to do something to make it hashable, for example converting it to its string representation:

>>> pivot_table = pivot(ShirtSales, 'region', 'shipped', 'units', display_transform=str)

If there are no records in the original data table for a particular cell in the pivot result, SQL will return NULL and this gets translated to None in python. If you want to get zero, or some other default, you can pass that as a parameter to pivot:

>>> pivot_table = pivot(ShirtSales, 'region', 'shipped', 'units', default=0)

The above call ensures that when there are no units sold in a particular region on a particular date, we get zero as the result instead of None. However, the results will only contain shipped dates if at least one region had sales on that date. If it's necessary to get results for all dates in a range including dates where there are no ShirtSales records, we can pass a target row_range:

>>> from datetime import date, timedelta
>>> row_range = [date(2005, 1, 1) + timedelta(days) for days in range(59)]
>>> pivot_table = pivot(ShirtSales, 'region', 'shipped', 'units', default=0, row_range=row_range)

Will output a result with every shipped date from Jan 1st to February 28th whether there are sales on those days or not.

The histogram function

This library also supports creating histograms from a single column of data with the histogram function, which takes a Model, a single attribute name and an iterable of left edges of bins.

>>> hist = histogram(ShirtSales, 'units', bins=[0, 10, 15])

Like pivot, the first argument can be a Model, QuerySet, or Manager. The result is a list of dictionaries:

>>> hist
[{'bin': '0', 'units': 0},
{'bin': '10', 'units': 0},
{'bin': '15', 'units': 0}]

It's also possible to get several histograms from a single query by slicing the data on one of the columns. For example, instead of the histogram above, we might want two histograms, one for boys and one for girls. The gender column of ShirtSales has two values, 'Boy' and 'Girl'. Passing the gender column as a 4th optional parameter to histogram will slice the data on that column.

>>> hist = histogram(ShirtSales, 'units', bins=[0, 10, 15], slice_on='gender')

The result is a ValuesQuerySet where each row corresponds to one bin

>>> for row in hist:
        print(row)
{'bin': u'0', u'Boy': 53, u'Girl': 53}
{'bin': u'10', u'Boy': 40, u'Girl': 41}
{'bin': u'15', u'Boy': 27, u'Girl': 26}

Installation

Just:

pip install django-pivot

and then you:

from django_pivot.pivot import pivot
from django_pivot.histogram import histogram

And off you go.

Tests

The test suite is run by via Github actions for pushes to master and pull requests to master with Django versions 3.2.21, 4.1.11, and 4.2.5 and backends sqlite, MySQL, and Postgres. If you want to run the test suite locally, from the root directory:

python runtests.py

That will use sqlite as the backend and whatever version of Django you have in your current environment.

License

MIT

Copyright 2017 - 2023 Brad Martsberger

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Contributors

rafal-jaworski