Django CSV Import
Ed Crewe - January 2023
Overview
django-csvimport is a generic importer tool to allow the upload of CSV files for populating data. The egg installs an admin csvimport model that has a file upload field. Add a new csvimport and upload a comma separated values file or MS Excel file.
The upload triggers the import mechanism which matches the header line of the files field names to the fields in the selected model. Importing any rows that include all required fields. Optionally required fields can be specified as part of the upload. By default duplicate value rows are not inserted.
The import can also be run as a custom command, ie manage.py importcsv filename for possible use via cronjob etc.
For CSV files import where their schema is unknown, and there is no existing model to import to, there is another command, inspectcsv, to generate the model code from the CSV file, guessing data types from the data using https://messytables.readthedocs.org - to use this command please pip install messytables.
The core import code was based on http://djangosnippets.org/snippets/633/ by Jonathan Holst. It adds character encoding handling, model field and column autodetection, admin interface, custom command etc.
Version 2 - Sept 2014
- New management command csvinspect to generate models from CSV files
- General code refactor
- Management command renamed from csvimport to importcsv
- More features to cope with bad encoding and date types
Version Compatibility
version 2.16 tested with Django 3.2.16 on Python 3.9.6 version 2.14 tested with Django 3.0.5 on Python 3.7.6, 3.8.2 version 2.13 was tested with Django 2.2.5 on Python 3.7.3 version 2.6 was tested with Django 1.7, 1.8, 1.9, 1.10, 1.11 on Python 2.7.13, Python 3.5.6
(NOTE To use the inspectcsv model generation command, requires messytables. Latest messytables needs older python magic...
- pip install python-magic-bin==0.4.14
- pip install messytables)
Please use version 2.1, eg. pip install django-csvimport==2.1 for Django versions prior to 1.7
This Django >= 1.7 requirement is because django-csvimport uses the newly added AppConfig for versions > 2.1 (NB: To fix this issue you could install django-appconf to django 1.6 or earlier and tweak csvimport to use it in csvimport.app)
For really old Django versions < 1.4 you may have to dial back the versions until it works!
Note that only versions > 2.2 are compatible with Python 3.4
Installation instructions
Add the following to the INSTALLED_APPS in the settings.py of your project:
>>> pip install django-csvimport
...
... INSTALLED_APPS = (
... ...
... 'csvimport.app.CSVImportConf', # use AppConfig for django >=1.7 csvimport >=2.2
... )
...
... python manage.py migrate (or syncdb if django < 1.9)Note that migrate has the core tables in 0001_initial migration and test tables in 0002 so rm migrations/0002_test_models.py if you do not want these cluttering your database
Custom commands
INSPECTCSV
(pip install messytables to use this command)
manage.py inspectcsv importfile.csv > models.py
This returns the code for a new models file with a guesstimated model for the CSV file. Add it to your app then run
>>> makemigrations your_app
>>> migrateYou can then run the import to that model for importfile.csv
NB: As it says its a guesstimate, you may have to manually tweak the generated models.py to get the import to work better.
If there are no headings in the CSV file, then it just uses automated ones col_1, col_2 ... etc.
IMPORTCSV
(Please note this command used to be csvimport but that caused name clash issues with the module)
manage.py importcsv --mappings='' --model='app_label.model_name' --delimiter='t' importfile.csv
For mappings enter a list of fields in order only if you dont have a header row with matching field names - or you want to override it, eg.
--mappings = '1=shared_code,2=org(otherapp.Organisation|name),3=date'
where (model|foreign key field) is used to specify relations if again, you want to override what would be looked up from your models.
If you have no real field names in your csv file, then you can use --mappings='none' and it will assume the fields are named col_1, col_2 ... etc.
Note that if you have a header row and specify mappings then it will treat the header as a data row, so delete it first.
Admin interface import
Just add a csvimport item, fill in the form and submit. Failed import rows are added to the log field.
Demonstration installation instructions
To see how it works, you can install a demo easily enough eg. via virtual environment, then use the tests settings to have some sample models for importing data, and the fixtures are sample csv files.
- Run the following in your shell:
>>> virtualenv mysite
... cd mysite
... pip install django
... pip install django-csvimport
...
... cat > bin/django-admin.py << EOF
... #!/usr/bin/env python
... from django.core import management
... import os
... os.environ["DJANGO_SETTINGS_MODULE"] = "csvimport.settings"
... if __name__ == "__main__":
... management.execute_from_command_line()
... EOF
...
... django-admin.py migrate
... django-admin.py runserver- Go to http://127.0.0.1:8000/admin/ in your browser - pay attention to the trailing / !
- Click on add Csvimport
- Pick the django-csvimport/csvimport/tests/fixtures/countries.csv [1] and upload it
- Check to see if the Country model is now populated.
[1] also available from https://raw.github.com/edcrewe/django-importcsv/master/importcsv/tests/fixtures/countries.csv
Alternatively you can use the command line to upload
django-admin.py importcsv --model='csvimport.Country' django-csvimport/csvimport/tests/fixtures/countries.csv --settings='csvimport.settings'
Foreign Keys
It is not viable for csvimport to determine complex table relations. However if it finds something marked as an ForeignKey with a lookup field in its model mappings, then it checks if the data exists already for the related model and pulls back an id for the field or creates a new entry if possible in the fkey model and pulls back a new id.
For this to be useful then you need a related table that has a unique and more meaningful field that is being used in your data than a numeric primary key.
eg. for an organisation column, org, that holds the unique name of the organisation from a separate table, you can add
column2=org(Organisation|name)
to the mappings, so it knows that the org field relates to a separate Organisation table with a unique name field to be used for it to lookup and replace with org_id FKey
More complex relations
For any more sophisticated relations, eg. multiple keys, many to many fields etc. The recommended approach is to create a temporary or intermediate import table that holds the data from your CSV file with the lookup data in as columns, you can use
inspectcsv importfile.csv > models.py
to automatically generate the import model from your CSV file.
Whenever you do an import to that table you would use a bulk insert database query to take the data in it and populate complex relations of the final model tables appropriately. If imports are happening repeatedly, eg. once a day, you retain your import CSV format table, and can add a database trigger for the table to automatically run your stored data conversion synchronisation query into the target tables.
DateTime data
Note that the importer uses the standard date input format list in settings to try to convert any datetime types you have in your CSV file. So add this to your settings with a list of any formats that you want to be handled.
DATE_INPUT_FORMATS = ['%Y-%m-%d %H:%M:%S']
In order for dates to be imported outside of the timezone range of 1970-2037 for certain database backends such as sqlite there is a patch of django.utils.timezone using tzinfo monkeypatch
Acknowledgements
This package was created as part of a django dash at the House of Omni, Bristol UK, organised by Dan Fairs and my local django users group, #DBBUG. It was a core component for an application for aid agency supply chain sharing, prompted by Fraser Stephens of the HELIOS foundation and developed by Ed Crewe and Tom Dunham.