An object-oriented interface for abstracting away the ugly parts of ad server APIs
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


What is Parselmouth?

Parselmouth serves as a facade/abstraction to adserver interfaces. As of this writing (2015-11-03) Parselmouth only interfaces with Google's Doubleclick for Publishers (DFP).

External API services like Google's DFP can be painful to interface with given the irregularity of the data structures they use to communicate and the frequency with which the API changes. Parselmouth aims to alleviate the pain of using these services by standardizing a simple, pythonic object-oriented interface that abstracts away the frustrating implementation details of an external API.

For background on how this project began, click here

How do I get started?

Installing the library

Install or update the library from PyPI. If you're using pip, this is as easy as:

$ pip install [--upgrade] parselmouth

If you don't want to install directly from PyPI, you can download the library as a tarball and then install it manually. The download can be found here: Navigate to the directory that contains your downloaded unzipped client library and run the "" script to install the "parselmouth" module.

$ python build install

Intializing Connection

Parselmouth can be initialized in one of the following ways:

Manually enumerating credentials:
from parselmouth import Parselmouth
from parselmouth import ParselmouthProviders

client = Parselmouth(
    ParselmouthProviders.google_dfp_premium,  # Or google_dfp_small_business
Loading credentials from a configuration file:

Credentials may optionally be stored in a YAML file like the following:

    client_id: 'ID'
    client_secret: 'SECRET'
    refresh_token: 'REFRESH_TOKEN'
    network_code: 'NETWORK_CODE'
    application_name: 'APPLICATION_NAME'

    client_id: 'ID'
    client_secret: 'SECRET'
    refresh_token: 'REFRESH_TOKEN'
    network_code: 'NETWORK_CODE'
    application_name: 'APPLICATION_NAME'

And then loaded like this:

client = Parselmouth(
Loading credentials from a configuration file:

Parselmouth may additionally be initialized using implementation-specific configuration classes. These classes can be written to retrieve credentials from external sources such as a SQL database or key-value store.

from parselmouth import ParselmouthConfig  # Or your own custom config object

config = ParselmouthConfig(

client = Parselmouth(config)

##Basic Usage ####Campaigns

A primary use-case of Parselmouth is to obtain information about ad campaigns from an ad service provider.

>>> campaign = client.get_campaign('ORDER_ID')
>>> campaign.start
datetime.datetime(2013, 7, 31, 16, 30, tzinfo=<DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD>)
>>> campaign.stats
Stats({'video_completions': 0, 'impressions': 572, 'click_through_rate': 0, 'clicks': 0, 'video_starts': 0})

####Line Items and Creatives

You can also get line item and creative objects

>>> line_items = client.get_campaign_line_items(campaign)
>>> line_item = line_items[0]
>>> creatives = client.get_line_item_creatives(line_item)
>>> creatives[0].id

Click here for more details on working with delivery objects.

####Line Item Targeting

One of the most convenient features of Parselmouth is it's abstraction of targeting information. Native API implementations of targeting configurations are terribly complex, unwieldy, and difficult to manipulate. Parselmouth defines a simple way of applying arbitrary boolean inclusion/exclusion criterion to a LineItem.

from parselmouth.targeting import Geography

# Initialize some geography targets
usa = Geography(name='USA')
canada = Geography(name='Canada')
uk = Geography(name='UK')
scotland = Geography(name='Scotland')

# Initialize some regional TargetingCriterion
na_region = TargetingCriterion([usa, canada], TargetingCriterion.OPERATOR.OR)
uk_region = TargetingCriterion([uk, scotland], TargetingCriterion.OPERATOR.OR)

# Build new TargetingCriterion using boolean operators
target_either = na_region | uk_region
target_na_only = na_region & ~uk_region
target_neither = ~(na_region | uk_region)

Click here for more details on targeting.


DFP also supports ad unit hierarchies governed by complex tree structures. For example a site might have an ad unit (or zone) called Sports with child ad units called Sports/Hockey and Sports/Baseball. Both of these ad units share a parent ad unit called Sports and are sibblings in this tree structure. Parselmouth also supports functions which help in working with these tree structures using objects called NodeTrees.

>>> from parselmouth.constants import ParselmouthTargetTypes
>>> adunit_tree = client.construct_tree(ParselmouthTargetTypes.adunit)
>>> sports_tree = adunit_tree.get_subtree('name', 'Sports')
>>> for sub_unit in sports_tree.flatten():
...     print

Click here for more details on trees.

####Object Serialization

All objects within Parselmouth can also be serialized to a dictionary.

>>> line_item_doc = line_item.to_doc()
>>> print line_item_doc
{'status': 'COMPLETED', 'domain': None, 'targeting': {'video_position': None, 'day_part': None, 'custom': None, u'_metadata': {u'cls': 'TargetingData'}, 'inventory': {u'_metadata': {u'cls': 'TargetingCriterion'}, 'OR': [{'adunits': None, 'external_name': None, 'name': None, u'_metadata': {u'cls': 'Placement'}, 'parent_id': None, 'external_id': None, 'id': '1904883'}]}, 'video_content': None, 'user_domain': None, 'technology': None, 'geography': None}, 'name': 'Flight 1', 'cost_per_unit': {'budget_currency_code': 'USD', 'budget_micro_amount': 0.0, u'_metadata': {u'cls': 'Cost'}}, 'type': 'standard', 'campaign_id': '134419323', 'last_modified_by': 'Goog_DFPUI', 'value_cost_per_unit': {'budget_currency_code': 'USD', 'budget_micro_amount': 0.0, u'_metadata': {u'cls': 'Cost'}}, 'delivery': {'stats': {'video_completions': 0, u'_metadata': {u'cls': 'Stats'}, 'click_through_rate': 0, 'video_starts': 0, 'impressions': 572, 'clicks': 0}, 'pace': 5.72e-06, 'expected_delivery_percent': 100.0, u'_metadata': {u'cls': 'DeliveryMeta'}, 'delivery_rate_type': 'FRONTLOADED', 'actual_delivery_percent': 0.000572}, 'start': datetime.datetime(2013, 7, 31, 16, 30, tzinfo=<DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD>), 'campaign_name': 'Test', 'cost_type': 'CPM', 'creative_placeholder': [{u'expectedCreativeCount': '1', u'creativeSizeType': 'PIXEL', u'size': {u'width': '300', u'isAspectRatio': False, u'height': '600'}}, {u'expectedCreativeCount': '1', u'creativeSizeType': 'PIXEL', u'size': {u'width': '728', u'isAspectRatio': False, u'height': '90'}}, {u'expectedCreativeCount': '1', u'creativeSizeType': 'PIXEL', u'size': {u'width': '300', u'isAspectRatio': False, u'height': '250'}}], u'_metadata': {u'cls': 'LineItem'}, 'last_modified': datetime.datetime(2013, 11, 20, 13, 34, 52, tzinfo=<DstTzInfo 'PST8PDT' PST-1 day, 16:00:00 STD>), 'budget': {'budget_currency_code': 'USD', 'budget_micro_amount': 0.0, u'_metadata': {u'cls': 'Cost'}}, 'primary_goal': {'unit_type': 'IMPRESSIONS', 'units': 100000000, 'goal_type': 'LIFETIME', u'_metadata': {u'cls': 'Goal'}}, 'end': datetime.datetime(2013, 9, 1, 23, 59, tzinfo=<DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD>), 'target_platform': 'WEB', 'id': '74067003'}

And these dictionaries can be re-serialized back into their native type.

>>> from import LineItem
>>> reconstructed_line_item = LineItem.from_doc(line_item_doc)
>>> line_item == reconstructed_line_item



Copyright 2015 Chartbeat, Inc.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
See the License for the specific language governing permissions and
limitations under the License.