| Authors: | Power Go |
|---|---|
| Version: | 0.2.3 |
This library provides a set of functions that provide open and click tracking when sending emails. This is particularly useful if you rely on an Email Service Provider (ESP) which does not provide open and click tracking.
The library only provides building blocks and does not handle the actual sending of email or the serving of tracking pixel and links, but it comes pretty close to this.
Summary
There are two main steps when tracking email opens and link clicks:
To track email opens, the generally accepted strategy is to add a small 1x1 transparent pixel at the end of an email. When a user opens an email, the email client (e.g., gmail, outlook, thunderbird) will load the pixel by making a GET request. The web server serving the request will then record the open and notify the sender of the email.
To track link clicks, the generally accepted strategy is to rewrite links in an email to change the destination to a proxy. Once a user clicks on the link, the proxy redirects the user to the real link and notifies the sender of the email.
pytracking provides a stateless strategy to open and click tracking: all the information you want to track are encoded in the pixel (open) and proxy (click) URLs. For example, if you want to track the customer id and the transaction id associated with a particular email, pytracking will encode this information in the URL. When the user opens the email or clicks on a link, the customer id and transaction id will be decoded and can then be sent to a webhook.
See the Get Open Tracking Link section for a quick example.
Once a user opens an email or clicks on a link, the email client will send a request to the encoded URL. Your web server will receive such request and pass it to pytracking, which will decode the tracking information. You can then use the tracking information directly (e.g., update your tracking database) or you can send the information to a webhook.
In the case of link tracking, the decoded information will contain the original URL that you must redirect the email client to.
See the Get Open Tracking Data from URL section for a quick example.
- Encryption: pytracking uses base 64 to encode your tracking information, which can be decoded by anyone. You can optionaly encrypt your tracking information, which can only be decoded if you have the key. See the Encrypting Data section for more information.
- HTML modification: pytracking can modify an HTML email to replace all links and add a tracking pixel. See the Modifying HTML emails to add tracking links section.
- Django: if you use Django to serve open and click tracking URLs, you can extend pytracking Django views, which already provides the redirect and pixel serving. See the Using pytracking with Django section.
- Webhooks: pytracking offers a shortcut function to make a POST request to a webhook. See the Notifying Webhooks section.
pytracking works with Python 3.6+. It doesn't require any external library, but there are many optional features that have dependencies.
You can install pytracking using pip:
pip install pytracking
You can install specific features with extras:
pip install pytracking[django,crypto]
You can also install all features:
pip install pytracking[all]
You can generate two kinds of tracking links with pytracking: a link to a transparent tracking pixel and a link that redirects to another link.
You can encode metadata in both kinds of links. For example, you can associate a customer id with a click tracking link so when the customer clicks on the link, you'll know exactly which customer clicked on it.
pylinktracking implements a stateless tracking strategy: all necessary information can be encoded in the tracking links. You can optionally keep common settings (e.g., default metadata to associate with all links, webhook URL) in a separate configuration.
The information is encoded using url-safe base64 so anyone intercepting your links, including your customers, could potentially decode the information. You can optionally encrypt the tracking information (see below).
Most functions take as a parameter a pytracking.Configuration
instance that tells how to generate the links. You can also pass the
configuration parameters as **kwargs argument or can mix both: the kwargs
will override the configuration parameters.
Once you get a request from a tracking link, you can use pytracking to decode
the link and get a pytracking.TrackingResult instance, which contains
information such as the link to redirect to (if it's a click tracking link),
the associated metadata, the webhook URL to notify, etc.
import pytracking
open_tracking_url = pytracking.get_open_tracking_url(
{"customer_id": 1}, base_open_tracking_url="https://trackingdomain.com/path/",
webhook_url="http://requestb.in/123", include_webhook_url=True)
# This will produce a URL such as:
# https://trackingdomain.com/path/e30203jhd9239754jh21387293jhf989sda=
import pytracking
configuration = pytracking.Configuration(
base_open_tracking_url="https://trackingdomain.com/path/",
webhook_url="http://requestb.in/123",
include_webhook_url=False)
open_tracking_url = pytracking.get_open_tracking_url(
{"customer_id": 1}, configuration=configuration)
# This will produce a URL such as:
# https://trackingdomain.com/path/e30203jhd9239754jh21387293jhf989sda=
import pytracking
click_tracking_url = pytracking.get_click_tracking_url(
"http://www.example.com/?query=value", {"customer_id": 1},
base_click_tracking_url="https://trackingdomain.com/path/",
webhook_url="http://requestb.in/123", include_webhook_url=True)
# This will produce a URL such as:
# https://trackingdomain.com/path/e30203jhd9239754jh21387293jhf989sda=
import pytracking
full_url = "https://trackingdomain.com/path/e30203jhd9239754jh21387293jhf989sda="
tracking_result = pytracking.get_open_tracking_result(
full_url, base_open_tracking_url="https://trackingdomain.com/path/")
# Metadata is in tracking_result.metadata
# Webhook URL is in tracking_result.webhook_url
import pytracking
full_url = "https://trackingdomain.com/path/e30203jhd9239754jh21387293jhf989sda="
tracking_result = pytracking.get_click_tracking_result(
full_url, base_click_tracking_url="https://trackingdomain.com/path/")
# Metadata is in tracking_result.metadata
# Webhook URL is in tracking_result.webhook_url
# Tracked URL to redirect to is in tracking_result.tracked_url
import pytracking (pixel_byte_string, mime_type) = pytracking.get_open_tracking_pixel()
You can encrypt your encoded data to prevent third parties from accessing the tracking data encoded in your link.
To use the encryption feature, you must install pytracking with
pytracking[crypto], which uses the cryptography Python library.
Encrypting your data slightly increases the length of the generated URL.
import pytracking
from cryptography.fernet import Fernet
key = Fernet.generate_key()
# Encode
click_tracking_url = pytracking.get_click_tracking_url(
"http://www.example.com/?query=value", {"customer_id": 1},
base_click_tracking_url="https://trackingdomain.com/path/",
webhook_url="http://requestb.in/123", include_webhook_url=True,
encryption_bytestring_key=key)
# Decode
tracking_result = pytracking.get_open_tracking_result(
full_url, base_click_tracking_url="https://trackingdomain.com/path/",
encryption_bytestring_key=key)
pytracking comes with View classes that you can extend and that handle open and click tracking link request.
For example, the pytracking.django.OpenTrackingView will return a 1x1
transparent PNG pixel for GET requests. The
pytracking.django.ClickTrackingView will return a 302 redirect response to
the tracked URL.
Both views will return a 404 response if the tracking URL is invalid. Both views will capture the user agent and the user ip of the request. This information will be available in TrackingResult.request_data.
You can extend both views to determine what to do with the tracking result (e.g., call a webhook or submit a task to a celery queue). Finally, you can encode your configuration parameters in your Django settings or you can compute them in your view.
To use the django feature, you must install pytracking with
pytracking[django].
You can provide default configuration parameters in your Django settings by adding this key in your settings file:
PYTRACKING_CONFIGURATION = {
"webhook_url": "http://requestb.in/123",
"base_open_tracking_url": "http://tracking.domain.com/open/",
"base_click_tracking_url": "http://tracking.domain.com/click/",
"default_metadata": {"analytics_key": "123456"},
"append_slash": True
}
from pytracking import Configuration
from pytracking.django import OpenTrackingView, ClickTrackingView
class MyOpenTrackingView(OpenTrackingView):
def notify_tracking_event(self, tracking_result):
# Override this method to do something with the tracking result.
# tracking_result.request_data["user_agent"] and
# tracking_result.request_data["user_ip"] contains the user agent
# and ip of the client.
send_tracking_result_to_queue(tracking_result)
def notify_decoding_error(self, exception, request):
# Called when the tracking link cannot be decoded
# Override this to, for example, log the exception
logger.log(exception)
def get_configuration(self):
# By defaut, fetchs the configuration parameters from the Django
# settings. You can return your own Configuration object here if
# you do not want to use Django settings.
return Configuration()
class MyClickTrackingView(ClickTrackingView):
def notify_tracking_event(self, tracking_result):
# Override this method to do something with the tracking result.
# tracking_result.request_data["user_agent"] and
# tracking_result.request_data["user_ip"] contains the user agent
# and ip of the client.
send_tracking_result_to_queue(tracking_result)
def notify_decoding_error(self, exception, request):
# Called when the tracking link cannot be decoded
# Override this to, for example, log the exception
logger.log(exception)
def get_configuration(self):
# By defaut, fetchs the configuration parameters from the Django
# settings. You can return your own Configuration object here if
# you do not want to use Django settings.
return Configuration()
Add this to your urls.py file:
urlpatterns = [
url(
"^open/(?P<path>[\w=-]+)/$", MyOpenTrackingView.as_view(),
name="open_tracking"),
url(
"^click/(?P<path>[\w=-]+)/$", MyClickTrackingView.as_view(),
name="click_tracking"),
]
You can send a POST request to a webhook with the tracking result. The webhook
feature just packages the tracking result as a json string in the POST body. It
also sets the content encoding to application/json.
To use the webhook feature, you must install pytracking with
pytracking[webhook].
import pytracking
from pytracking.webhook import send_webhook
# Assumes that the webhook url is encoded in the url.
full_url = "https://trackingdomain.com/path/e30203jhd9239754jh21387293jhf989sda="
tracking_result = pytracking.get_open_tracking_result(
full_url, base_click_tracking_url="https://trackingdomain.com/path/")
# Will send a POST request with the following json str body:
# {
# "is_open_tracking": False,
# "is_click_tracking": True,
# "metadata": {...},
# "request_data": None,
# "tracked_url": "http://...",
# "timestamp": 1389177318
# }
send_webhook(tracking_result)
If you have an HTML email, pytracking can update all links with tracking links and it can also add a transparent tracking pixel at the end of your email.
To use the HTML feature, you must install pytracking with pytracking[html],
which uses the lxml library.
import pytracking
from pytracking.html import adapt_html
html_email_text = "..."
new_html_email_text = adapt_html(
html_email_text, extra_metadata={"customer_id": 1},
click_tracking=True, open_tracking=True)
pytracking uses tox and py.test. If you have tox installed, just run
tox and all possible configurations of pytracking will be tested on Python
3.6-3.9.
- Add various checks to ensure that the input data is sane and does not bust any known limits (e.g., URL length).
- Add more examples.
- Allow multiplpe webhooks and webhooks per tracking method.
- Transform Django views into view mixins.
- Add option to encode the webhook timeout in the tracking URL.
- Document caveats of using pytracking.html (example: long emails are often cut off by the email clients and the tracking pixel is thus not loaded).
- Add some form of API documentation (at least Configuration and TrackingResult), maybe as a separate document.
This software is licensed under the New BSD License. See the LICENSE file in the repository for the full license text.