Tools for working with pandas in your Django projects

• Christopher Clarke
• Bertrand Bordage
• Guillaume Thomas
• Parbhat Puri
• Fredrik Burman (coachHIPPO)
• Safe Hammad
• Jeff Sternber
• @MiddleFork
• Daniel Andrlik
• Kevin Abbot
• Yousuf Jawwad
• @henhuy
• Hélio Meira Lins
• @utpyngo
• Anthony Monthe
• Vincent Toupet
• Anton Ian Sipos
• Thomas Grainger
• Ryan Smith

This is release facilitates running of test with Python 3.10 and automates the publishing of the package to PYPI as per PR #146 (again much thanks @graingert). As usual we have attempted support legacy versions of Python/Django/Pandas and this sometimes results in deperation errors being displayed in when test are run. To avoid use python -Werror runtests.py

django-pandas supports Django (>=1.4.5) or later and requires django-model-utils (>= 1.4.0) and Pandas (>= 0.12.0). Note because of problems with the requires directive of setuptools you probably need to install numpy in your virtualenv before you install this package or if you want to run the test suite :

pip install numpy
pip install -e .[test]
python runtests.py

Some pandas functionality requires parts of the Scipy stack. You may wish to consult http://www.scipy.org/install.html for more information on installing the Scipy stack.

You need to install your preferred version of Django. as that Django 2 does not support Python 2.

Please file bugs and send pull requests to the GitHub repository and issue tracker.

Start by creating a new virtualenv for your project :

mkvirtualenv myproject

Next install numpy and pandas and optionally scipy :

pip install numpy
pip install pandas

You may want to consult the scipy documentation for more information on installing the Scipy stack.

Finally, install django-pandas using pip:

pip install django-pandas

or install the development version from github :

pip install https://github.com/chrisdev/django-pandas/tarball/master

The django-pandas.io module provides some convenience methods to facilitate the creation of DataFrames from Django QuerySets.

Parameters

• qs: A Django QuerySet.

• fieldnames: A list of model field names to use in creating the DataFrame.

  You can span a relationship in the usual Django way by using double underscores to specify a related field in another model

• index_col: Use specify the field name to use for the DataFrame index.

  If the index field is not in the field list it will be appended

• coerce_float : Boolean, defaults to True

  Attempt to convert values to non-string, non-numeric objects (like decimal.Decimal) to floating point.

• verbose: If this is True then populate the DataFrame with the

  human readable versions of any foreign key or choice fields else use the actual values set in the model.

• column_names: If not None, use to override the column names in the

  DateFrame

Assume that this is your model:

class MyModel(models.Model):

    full_name = models.CharField(max_length=25)
    age = models.IntegerField()
    department = models.CharField(max_length=3)
    wage = models.FloatField()

First create a query set:

from django_pandas.io import read_frame
qs = MyModel.objects.all()

To create a dataframe using all the fields in the underlying model :

df = read_frame(qs)

The df will contain human readable column values for foreign key and choice fields. The DataFrame will include all the fields in the underlying model including the primary key. To create a DataFrame using specified field names:

df = read_frame(qs, fieldnames=['age', 'wage', 'full_name'])

To set full_name as the DataFrame index :

qs.to_dataframe(['age', 'wage'], index_col='full_name'])

You can use filters and excludes :

qs.filter(age__gt=20, department='IT').to_dataframe(index_col='full_name')

django-pandas provides a custom manager to use with models that you want to render as Pandas Dataframes. The DataFrameManager manager provides the to_dataframe method that returns your models queryset as a Pandas DataFrame. To use the DataFrameManager, first override the default manager (objects) in your model's definition as shown in the example below :

#models.py

from django_pandas.managers import DataFrameManager

class MyModel(models.Model):

    full_name = models.CharField(max_length=25)
    age = models.IntegerField()
    department = models.CharField(max_length=3)
    wage = models.FloatField()

    objects = DataFrameManager()

This will give you access to the following QuerySet methods:

• to_dataframe
• to_timeseries
• to_pivot_table

Returns a DataFrame from the QuerySet

Parameters

• fieldnames: The model field names to utilise in creating the frame.

  to span a relationship, use the field name of related fields across models, separated by double underscores,

• index: specify the field to use for the index. If the index

  field is not in the field list it will be appended

• coerce_float: Attempt to convert the numeric non-string data

  like object, decimal etc. to float if possible

• verbose: If this is True then populate the DataFrame with the

  human readable versions of any foreign key or choice fields else use the actual value set in the model.

Create a dataframe using all the fields in your model as follows :

qs = MyModel.objects.all()

df = qs.to_dataframe()

This will include your primary key. To create a DataFrame using specified field names:

df = qs.to_dataframe(fieldnames=['age', 'department', 'wage'])

To set full_name as the index :

qs.to_dataframe(['age', 'department', 'wage'], index='full_name'])

You can use filters and excludes :

qs.filter(age__gt=20, department='IT').to_dataframe(index='full_name')

A convenience method for creating a time series i.e the DataFrame index is instance of a DateTime or PeriodIndex

Parameters

• fieldnames: The model field names to utilise in creating the frame.

  to span a relationship, just use the field name of related fields across models, separated by double underscores,

• index: specify the field to use for the index. If the index

  field is not in the field list it will be appended. This is mandatory.

• storage: Specify if the queryset uses the wide or long format

  for data.

• pivot_columns: Required once the you specify long format

  storage. This could either be a list or string identifying the field name or combination of field. If the pivot_column is a single column then the unique values in this column become a new columns in the DataFrame If the pivot column is a list the values in these columns are concatenated (using the '-' as a separator) and these values are used for the new timeseries columns

• values: Also required if you utilize the long storage the

  values column name is use for populating new frame values

• freq: the offset string or object representing a target conversion
• rs_kwargs: Arguments based on pandas.DataFrame.resample

• verbose: If this is True then populate the DataFrame with the

  human readable versions of any foreign key or choice fields else use the actual value set in the model.

Using a long storage format :

#models.py

class LongTimeSeries(models.Model):
    date_ix = models.DateTimeField()
    series_name = models.CharField(max_length=100)
    value = models.FloatField()

    objects = DataFrameManager()

Some sample data::

========   =====       =====
date_ix    series_name value
========   =====       ======
2010-01-01  gdp        204699

2010-01-01  inflation  2.0

2010-01-01  wages      100.7

2010-02-01  gdp        204704

2010-02-01  inflation  2.4

2010-03-01  wages      100.4

2010-02-01  gdp        205966

2010-02-01  inflation  2.5

2010-03-01  wages      100.5
==========  ========== ======

Create a QuerySet :

qs = LongTimeSeries.objects.filter(date_ix__year__gte=2010)

Create a timeseries dataframe :

df = qs.to_timeseries(index='date_ix',
                      pivot_columns='series_name',
                      values='value',
                      storage='long')
df.head()

date_ix      gdp     inflation     wages

2010-01-01   204966     2.0       100.7

2010-02-01   204704      2.4       100.4

2010-03-01   205966      2.5       100.5

Using a wide storage format :

class WideTimeSeries(models.Model):
    date_ix = models.DateTimeField()
    col1 = models.FloatField()
    col2 = models.FloatField()
    col3 = models.FloatField()
    col4 = models.FloatField()

    objects = DataFrameManager()

qs = WideTimeSeries.objects.all()

rs_kwargs = {'how': 'sum', 'kind': 'period'}
df = qs.to_timeseries(index='date_ix', pivot_columns='series_name',
                      values='value', storage='long',
                      freq='M', rs_kwargs=rs_kwargs)

A convenience method for creating a pivot table from a QuerySet

Parameters

• fieldnames: The model field names to utilise in creating the frame.

  to span a relationship, just use the field name of related fields across models, separated by double underscores,

• values : column to aggregate, optional

• rows : list of column names or arrays to group on

  Keys to group on the x-axis of the pivot table

• cols : list of column names or arrays to group on

  Keys to group on the y-axis of the pivot table

• aggfunc : function, default numpy.mean, or list of functions

  If list of functions passed, the resulting pivot table will have hierarchical columns whose top level are the function names (inferred from the function objects themselves)

• fill_value : scalar, default None

  Value to replace missing values with

• margins : boolean, default False

  Add all row / columns (e.g. for subtotal / grand totals)

• dropna : boolean, default True

Example :

# models.py
class PivotData(models.Model):
    row_col_a = models.CharField(max_length=15)
    row_col_b = models.CharField(max_length=15)
    row_col_c = models.CharField(max_length=15)
    value_col_d = models.FloatField()
    value_col_e = models.FloatField()
    value_col_f = models.FloatField()

    objects = DataFrameManager()

Usage :

rows = ['row_col_a', 'row_col_b']
cols = ['row_col_c']

pt = qs.to_pivot_table(values='value_col_d', rows=rows, cols=cols)