Salesforce integration for Django's ORM using the SF REST API.
Clone or download
Pull request Compare This branch is 166 commits behind django-salesforce:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
salesforce
tests
.coveragerc
.editorconfig
.gitignore
.travis.yml
AUTHORS.md
LICENSE.md
MANIFEST.in
README.rst
coverage_run.sh
manage.py
pylintrc
requirements.txt
setup.py
tox.ini

README.rst

django-salesforce

https://travis-ci.org/django-salesforce/django-salesforce.svg?branch=master

This library allows you to load and edit the objects in any Salesforce instance using Django models. The integration is fairly complete, and generally seamless for most uses. It works by integrating with the Django ORM, allowing access to the objects in your SFDC instance (Salesforce .com) as if they were in a traditional database.

Python 2.6, 2.7, 3.3, 3.4 or pypy; Django 1.4.2 - 1.7, partly Django 1.8. The best supported version is currently Django 1.7, including relative complicated subqueries. Django 1.8 is only very rudimentally supported, without raw queries and without values_list() and values() methods. The usual support can be expected in the next django-salesforce version. Note that Django 1.4.x is not compatible with Python 3.

Quick Start

  1. Install django-salesforce: pip install django-salesforce

  2. Add the salesforce app to your INSTALLED_APPS setting:

    INSTALLED_APPS = {
        "django.contrib.auth",
        "django.contrib.contenttypes",
        ...
        ...
        "salesforce"
    }
    
  3. Add a salesforce connection to your DATABASES setting:

    'salesforce': {
        'ENGINE': 'salesforce.backend',
        'CONSUMER_KEY': '',
        'CONSUMER_SECRET': '',
        'USER': '',
        'PASSWORD': '',
        'HOST': 'https://test.salesforce.com',
    }
    

    In the example above, all fields should be populated as follows:

    • CONSUMER_KEY and CONSUMER_SECRET values are for the app used to connect to your Salesforce account. Instructions for how get these are in the Salesforce REST API Documentation. Key and secret can be created on web by:
      • Salesforce web > Setup > App Setup > Create > Apps > Connected apps > New.
      • Click "Enable OAuth Settings" in API, then select "Access and manage your data (api)" from available OAuth Scopes.
      • Other red marked fields must be filled, but are not relevant for Django.
    • USER is the username used to connect.
    • PASSWORD is a concatenation of the user's password and security token. Security token can be omitted if the local IP address has been whitelisted in Security Controls / Network Access.
    • HOST is https://test.salesforce.com to access the sandbox, or https://login.salesforce.com to access production.

    If an error message is received while connecting, review the error received. Everything in the error message between {...} is exactly copied from the Salesforce error message to assist debugging.

    See also: Information on settings up Salesforce connected apps.

    Note about permissions: Everything for a project can work under restricted Salesforce user account if it has access to objects in your models. Introspection (inspectdb) doesn't require any permissions. Running tests for django_salesforce requires many permissions or Administrator account for sandbox.

  4. (optional) To override the default timeout of 15 seconds, define SALESFORCE_QUERY_TIMEOUT in your settings file:

    SALESFORCE_QUERY_TIMEOUT = 15
    
  5. (optional) If you want to use another name for your Salesforce DB connection, define SALESFORCE_DB_ALIAS in your settings file:

    SALESFORCE_DB_ALIAS = 'salesforce'
    
  6. Add salesforce.router.ModelRouter to your DATABASE_ROUTERS setting:

    DATABASE_ROUTERS = [
        "salesforce.router.ModelRouter"
    ]
    
  7. Define a model that extends salesforce.models.Model or export the complete SF schema by python manage.py inspectdb --database=salesforce and simplify it to what you need.

  8. You're all done! Just use your model like a normal Django model.

  9. (Optional) Create a normal Django admin.py module for your Salesforce model.

Primary Key

Salesforce doesn't allow you to define custom primary keys, so django-salesforce will add them automatically in all cases. You can override capitalization and use primary key id by configuring SF_PK='id' in your project settings. The previous capitalization of Id is only for old projects, but it will stay as the default variant until django-salesforce>=0.5.

Foreign Key Support

Foreign key filters are currently possible only from child to parent with some restrictions:

They are fully supported with Django 1.7+ without other restrictions. New

With Django 1.6 and older, ForeignKey filters are possible only for the first level of relationship and only for fields whose name equals the name of object. Foreign keys of an object can be normally accessed by dot notation without any restriction Example:

contacts = Contact.objects.filter(Account__Name='FOO Company')
print(contacts[0].Account.Owner.LastName)

But the relationship Owner__Name is not currently possible because the type of Owner is a different name (User).

Along similar lines, it's not currently possible to filter by ForeignKey relationships based on a custom field. This is because related objects (Lookup field or Master-Detail Relationship) use two different names in SOQL. If the relation is by ID the columns are named FieldName__c, whereas if the relation is stored by object the column is named FieldName__r. More details about this can be found in the discussion about #43.

In case of a ForeignKey you can specify the field name suffixed with _id, as it is automatically allowed by Django. For example: account_id instead of account.id, or AccountId instead of Account.Id. It is faster, if you need not to access to the related Account object.

Querysets can be easily inspected whether they are correctly compiled to SOQL. You can compare the meaning with the same compiled to SQL:

my_qs = Contact.objects.filter(my__little_more__complicated='queryset')
print my_qs.query.get_compiler('salesforce').as_sql()    # SOQL
print my_qs.query.get_compiler('default').as_sql()       # SQL

Generic foreign keys are frequently used in SF for fields that relate to objects of different types, e.g. the Parent of Note or Attachment can be almost any type of ususal SF objects. Filters by Parent.Type and retrieving this type is supported:

note = Note.objects.filter(parent_type='Contact')[0]
parent_model = getattr(example.models, note.parent_type)
parent_object = parent_model.objects.get(pk=note.parent_id)
assert note.parent_type == 'Contact'

Example of Note model is in salesforce.testrunner.example.models.Note.

Advanced usage

  • Multiple Inheritance from Abstract Models - Many Salesforce models use the same sets of fields, but using a single inheritance tree would be too complicated and fragile. Proxy models and mixins are also supported.

  • Testing - By default, tests will be run against the SFDC connection specified in settings.py, which will substantially increase testing time.

    One way to speed this up is to change the SALESFORCE_DB_ALIAS to point to another DB connection (preferably SQLite) during testing using the TEST_* settings variables. Django unit tests without SalesforceModel are fast everytimes. Special read only fields that are updated only by SFDC e.g. last_modified_date need more parameters to be possible to save them into an alternate database, e.g. by auto_now=True.

  • Multiple SFDC connections - In most cases, a single connection is all that most apps require, so the default DB connection to use for Salesforce is defined by the SALESFORCE_DB_ALIAS settings variable. This behavior can be also configured by DATABASE_ROUTERS, replacing the use of salesforce.backend.router.ModelRouter.

  • Non SF databases - If SALESFORCE_DB_ALIAS is set to a conventional database, the tables defined by the SF models will be created by syncdb. This behavior can be disabled by adding a Meta class with managed=False.

  • Custom Managers - When creating a custom manager for a model, the manager must be a descendant of salesforce.manager.SalesforceManager.

    In most cases, switching DB connections with .using(alias). will be sufficient, but if you need to call a method on your custom manager, you should instead use .db_manager(alias) to select a DB while returning the correct manager, e.g. Contact.objects.db_manager(alias).my_manager(params...)

  • Automatic Field Naming - Most of database columns names can be automatically deduced from Django field name, if no db_column is specified:

    last_name = models.CharField(max_length=80)     # db_column='LastName'
    FirstName = models.CharField(max_length=80)    # db_column='FirstName'
    custom_bool = models.BooleanField(custom=True)  # db_column='CustomBool__c'
    

    Fields named with an upper case character are never modified, except for the addition of the namespace prefix or the '__c' suffix for custom fields.

  • Custom SF Objects and Fields - Custom SF class objects are indicated by adding a Meta class with parameter 'custom=True'. All child fields are assumed to be custom as well, unless marked otherwise with a field parameter marked "custom=False".

    Similarly, custom fields on standard objects can be indicated by "custom=True", or they can be defined in an standard parent model (the custom Meta parameter is not inherited).

    Also namespace prefixes of managed packages (prefixed with "PackageName__" can be automatically applied to custom fields without db_column.

  • Meta class options - If an inner Meta class is used, it must be a descendant of SalesforceModel.Meta or must have managed=False.

  • Query deleted objects - Deleted objects that are in trash bin are not selected by a normal queryset, but if a special method query_all is used then also deleted objects are searched. If a trash bin is supported by the model then a boolean field IsDeleted can be in the model and it is possible to select only deleted objects

    deleted_list = list(Lead.objects.filter(IsDeleted=True).query_all())
    
  • Migrations - Migrations can be used for an alternate test database with SalesforceModel. Then all tables must have Meta managed = True and attributes db_table and db_column are required. (Migrations in SFDC will be probably never supported, though it was experimantally tested creation of a new simple table in sandbox if a development patch is applied and permissions increased. If anything would be implemented after all, a new attribute will be added to SalesforceModel for safe forward compatibility. Consequently, the setting managed = True can be considered safe as it is related only to the alternate non SFDC database configured by SF_ALIAS.)

Introspection and special attributes of fields

Some Salesforce fields can not be fully used without special attributes. You can see in the output of inspectdb in the most complete form.

  • sf_read_only - Some fields require this special attributes to make the model writable. Some fields are completely read only (READ_ONLY) or insertable only but can not be later updated (NOT_UPDATEABLE) or updateable only but can not be specified on insert (NOT_CREATEABLE). Examples of such fields are automatically updated fields "last_modified_by" and "last_modified_date" or fields defined by a formula like "name" of contact, given by "first_name" and "last_name". Example:

    last_modified_date = models.DateTimeField(sf_read_only=models.READ_ONLY)
    
  • Defaulted on create - Some fields have a dynamic default value unknown by Django and assigned by Salesforce if the field is omitted when a new object is inserted. This rule will not be used if the value is None. Sometimes is None even not accepted by Salesforce, while the missing value is ok. Django-salesforce supports it by a special default value model.BooleanField(default=models.DEFAULTED_ON_CREATE). That means "let it to Salesforce". This is useful for all fields marked by attribute defaultedOnCreate in Salesforce. For example the current user of Salesforce is assigned to owner field if no concrete user is assigned, but None would be rejected. All boolean fields have different default values according to current Checked/Unchecked preferences.

  • Comments # Reference to tables [...] Some builtin foreign keys are references to more tables. The class of first table is used in the exported ForeignKey and all tables are listed in the comment. Any of them can be used instead.:: models.ForeignKey(User) # Reference to tables [SelfServiceUser, User] cl object [SelfServiceUser, User]

  • Partial Database Introspection with inspectdb Tables that are exported into a Python model can be restricted by regular expression:

    python manage.py inspectdb --table-filter="Contact$|Account" --database=salesforce
    

    In this example, inspectdb will only export models for tables with exact name Contact and all tables that are prefixed with Account. This filter works with all supported database types.

  • Verbosity - This package can set correct column names for Salesforce without explicit attribute db_column for many objects automatically. These attributes are not exported if a default verbosity is used. This is intended for use only with SFDC. If an alternate non SFDC test database is also expected and migrations of any SalesforceModel will

  • Accessing the Salesforce SOAP API - There are some Salesforce actions that cannot or can hardly be implemented using the generic relational database abstraction and the REST API. For some of these actions there is an available endpoint in the old Salesforce API (SOAP) that can be accessed using our utility module. In order to use that module, you will need to install an additional dependency

    pip install beatbox    # depends on Python 2.7
    

    Here is an example of usage with Lead conversion

    from salesforce.utils import convert_lead
    
    lead = Lead.objects.all()[0]
    response = convert_lead(lead)
    

    All usual additional parameters are supported in the original letter case. It allows e.g. merging a Lead with an existing Account or Contact and to control much more.

    For the particular case of Lead conversion, beware that having some custom and required fields in either Contact, Account or Opportunity can be more complicated. A mapping from custom fields in your Lead to these fields must be defined in the system and these Lead fields should not be empty at the time of conversion. Follow the instructions for more details.

SSL/TLS settings

The package requests doesn't provide an easy way to set the minimum required SSL/TLS version while ensuring use of the highest version that is available on both sides. (requests issue 2118) The required version can be set in settings.py to one of reasonable values

import ssl
SF_SSL = {'ssl_version': ssl.PROTOCOL_SSLv23}
  • ssl.PROTOCOL_SSLv23 - use the highest available protocol, including TLS. The security depends on the lowest protocol supported by your the installed versions of Python, requests, pyOpenSSL, and installed versions of OpenSSL/libssl.
  • ssl.PROTOCOL_TLSv1 - This will pin the communication protocol to TLS 1.0. This must be changed to PROTOCOL_SSLv23 once SFDC disables TLS 1.0.

The default for django-salesforce is currently PROTOCOL_TLSv1 in hopes of reducing compatibility issues. If you have Python 2.7.9 and newer or Python 3.4.0 and newer, the old insecure protocols including SSL v3 are disabled unless you've installed PyOpenSSL. As long as you have not installed PyOpenSSL, it's recommended you update your settings to use PROTOCOL_SSLv23.

The test of readiness for TLS better than 1.0 and a test of disabled SSL 3 are run by all tests. These tests give also some suggestions for the tested machine. More tests for SSL/TLS client security by popular SSL evaluation sites can be run by the command

python manage.py test salesforce.tests.test_ssl.SslTest

Additional tests are skipped without the word SslTest on the command line, because some vulnerabilities are hopefully not (so?) important for connections to SFDC.

If you have an old Python, you can improve security a little (SNI, validation of certificates, fixed InsecurePlatformWarning) by additional packages:

pip install pyopenssl ndg-httpsclient pyasn1

These have dependencies on the libffi development libararies. Install libffi-dev on Debian/Ubuntu or libffi-devel on RedHat derivatives.

However, once you're using Python 2.7.9 and newer or Python 3.4.0 and newer, installing pyOpenSSL can enable SSLv3 again. If you must install PyOpenSSL on these Python versions, it is more secure to use ssl.PROTOCOL_TLSv1 than other protocols.

Ultimately this will become moot for users of django-salesforce, as SFDC will soon require the updated setting.

Caveats

This package is in continuous development, and the ultimate goal is to support all reasonable features of the Salesforce platform, but for now here are the potential pitfalls and unimplemented operations:

  • Large Objects — Since the entire result set needs to be transferred over HTTP, and since it's common to have extremely high column counts on full object queries, it's assumed that users will create models that are specific to their individual applications' needs. Models that have been included with this library are for example and documentation purposes.
  • Inheritence — When using the default router, all models for object types on Salesforce must extend salesforce.models.SalesforceModel. The model router checks for this to determine which models to handle through the Salesforce connection.
  • Multiple Updates — Multiple update support is not yet implemented.
  • Multiple Deletes — Multiple delete support is not yet implemented.
  • Database Syncsyncdb will only create new databases in non-SF databases (useful for unit tests); SFDC classes are assumed to already exist with the appropriate permissions.

Experimental Features

  • If you use multiple Salesforce databases or multiple instances of AdminSite, you'll probably want to extend salesforce.admin.RoutedModelAdmin" in your admin.py

  • Dynamic authorization - The original use-case for django-salesforce assumed use of a single set of credentials with read-write access to all necessary objects. It's now possible to write applications that use OAuth to interact with a Salesforce instance's data on your end user's behalf. You simply need to know or request the Access Token <https://www.salesforce.com/us/developer/docs/api_rest/Content/quickstart_oauth.htm> for the user in question. In this situation, it's not necessary to save any credentials for SFDC in Django settings. The manner in which you request or transmit this token (e.g., in the Authorization: header) is left up to the developer at this time.

    Configure your DATABASES setting as follows:

    'salesforce': {
        'ENGINE': 'salesforce.backend',
        'HOST': 'https://your-site.salesforce.com',
        'CONSUMER_KEY': '.',
        'CONSUMER_SECRET': '.',
        'USER': 'dynamic auth',
        'PASSWORD': '.',
    }
    

    A static SFDC connection can be specified with the data server URL in "HOST" Note that in this case we're not using the URL of the login server — the data server URL can be also used for login.

    Items with '.' value are ignored when using dynamic auth, but cannot be left empty.

    The last step is to enable the feature in your project in some way, probably by creating a Django middleware component. Then at the beginning of each request:

    from django.db import connections
    # After you get the access token for the user in some way
    # authenticate to SFDC with
    connections['salesforce'].sf_session.auth.dynamic_start(access_token)
    
    # or to override the `instance_url` on a per-request basis
    connections['salesforce'].sf_session.auth.dynamic_start(access_token, instance_url)
    

    Make sure to purge the access token at end of request:

    connections['salesforce'].sf_session.auth.dynamic_end()
    

    You can continue to supply static credentials in your project settings, but they will only be used before calling dynamic_start() and/or after calling dynamic_end().

Backwards-incompatible changes

  • v0.6.1: This is the last code that supports old Django 1.4, 1.5, 1.6 and it will be removed immediately.
  • v0.5: The name of primary key is currently id. The backward compatible behaviour for code created before v0.5 can be reached by settings SF_PK='Id'.

News since version 0.5

  • All child to parent filters are still correctly supported for Django 1.7 in many levels, including foreign keys between custom models or mixed builtin and custom models, also filters where the same model is used multiple times, e.g. filter Account objects by a field of their parent Account.