To better serve Wise business and customer needs, the PipelineWise codebase needs to shrink. We have made the difficult decision that, going forward many components of PipelineWise will be removed or incorporated in the main repo. The last version before this decision is v0.64.1

We thank all in the open-source community, that over the past 6 years, have helped to make PipelineWise a robust product for heterogeneous replication of many many Terabytes, daily

This is a Singer tap that reads data from Google Analytics and produces JSON-formatted data following the Singer spec.

This is a PipelineWise compatible target connector.

The recommended method of running this tap is to use it from PipelineWise. When running it from PipelineWise you don't need to configure this tap with JSON files and most of things are automated. Please check the related documentation at Google Analytics

If you want to run this Singer Tap independently please read further.

This tap is a fork of Meltano's Google Analytics tap that:

• Pulls raw data from the Google Analytics Reporting API.
• Supports generating as many reports as the user of the tap wants (following Google's limit of 7 dimensions and 10 metrics per report).
• Generates a valid Catalog that follows the Singer spec.

As the Google Analytics Reports are defined dynamically and there are practically infinite combinations of dimensions and metrics a user can ask for, the entities and their schema (i.e. the Catalog for this tap) are not static. So, this tap behaves more or less similarly to a tap extracting data from a Data Source (e.g. a Postgres Database).

The difference of tap-google-analytics to a database tap is that the Catalog (available entities/streams and their schema) is dynamic but not available to be discovered at run time by connecting to the Data Source. It must be dynamically generated based on the reports the user wants to generate by connecting to the Google Analytics Reporting API.

To that end, this tap uses an additional JSON file for the definition of the reports that the user wants to be generated. You can check, as an example, the JSON file used as a default in tap-google-analytics/defaults/default_report_definition.json. Those report definitions could be part of the config.json, but we prefer to keep config.json small and clean and provide the definitions by using an additional file.

Based on the report(s) definition, it generates a valid Catalog that follows the Singer spec.

It then behaves as any Singer compatible tap and uses that Catalog (or any Catalog generated by a tap-google-analytics) to generate the requested reports. The additional JSON file for defining the reports is only required for generating an initial Catalog.

When no report definitions are provided by the user, tap-google-analytics generates a default Catalog with some common reports provided:

• website_overview: Most common metrics (users, new users, sessions, avg session duration, page views, bounce rate, etc) per day
• traffic_sources: Most common metrics per day, source, medium and social network
• pages: Most common page metrics (page views, unique page views, avg time on page, etc) per day, host name and page path
• locations: Most common metrics per day and various location dimensions (continent, country, region, city, etc)
• monthly_active_users: Monthly active users (past 30 days) per day
• weekly_active_users: Weekly active users (past 7 days) per day
• daily_active_users: Daily active users (past 1 day) per day
• devices: Most common metrics per day, device category, operating system and browser

First, make sure Python 3 is installed on your system or follow these installation instructions for Mac or Ubuntu.

It's recommended to use a virtualenv:

  python3 -m venv venv
  pip install pipelinewise-tap-google-analytics

or

  python3 -m venv venv
  . venv/bin/activate
  pip install --upgrade pip
  pip install .

tap-google-analytics supports two different ways of authorization:

• Service account based authorization, where an administrator manually creates a service account with the appropriate permissions to view the account, property, and view you wish to fetch data from
• OAuth access_token based authorization, where this tap gets called with a valid access_token and refresh_token produced by an OAuth flow conducted in a different system.

If you're setting up tap-google-analytics for your own organization and only plan to extract from a handful of different views in the same limited set of properties, Service Account based authorization is the simplest. When you create a service account Google gives you a json file with that service account's credentials called the client_secrets.json, and that's all you need to pass to this tap, and you only have to do it once, so this is the recommended way of configuring tap-google-analytics.

If you're building something where a wide variety of users need to be able to give access to their Google Analytics, tap-google-analytics can use an access_token granted by those users to authorize it's requests to Google. This access_token is produced by a normal Google OAuth flow, but this flow is outside the scope of tap-google-analytics. This is useful if you're integrating tap-google-analytics with another system, like Stitch Data might do to allow users to configure their extracts themselves without manual config setup. This tap expects an access_token, refresh_token, client_id and client_secret to be passed to it in order to authenticate as the user who granted the token and then access their data.

In order for tap-google-analytics to access your Google Analytics Account, it needs the Analytics Reporting API and the Analytics API (which are two different things) enabled. If using a service account to authorize, these need to be enabled for a project inside the same organization as your Google Analytics account (see below), or if using an OAuth credential set, they need to be enabled for the project the OAuth client ID and secret come from.

If using the OAuth authorization method, the OAuth flow conducted elsewhere must request at minimum the analytics.readonly OAuth scope to get an access_token authorized to hit these APIs

If you have already have a valid client_secrets.json for a service account, or if you are using OAuth based authorization, you can skip the rest of this section.

As a first step, you need to create or use an existing project in the Google Developers Console:

1. Sign in to the Google Account you are using for managing Google Analytics (you must have Manage Users permission at the account, property, or view level).

2. Open the Service accounts page. If prompted, select a project or create a new one to use for accessing Google Analytics.

3. Click Create service account.

   In the Create service account window, type a name for the service account, and select Furnish a new private key. Then click Save and store it locally as client_secrets.json.

   If you already have a service account, you can generate a key by selecting 'Edit' for the account and then selecting the option to generate a key.

Your new public/private key pair is generated and downloaded to your machine; it serves as the only copy of this key. You are responsible for storing it securely.

The newly created service account will have an email address that looks similar to:

[email protected]

Use this email address to add a user to the Google analytics view you want to access via the API. For using tap-google-analytics only Read & Analyze permissions are needed.

1. Visit the Google Analytics Reporting API dashboard and make sure that the project you used in the Create credentials step is selected.

From this dashboard, you can enable/disable the API for your account, set Quotas and check usage stats for the service account you are using with tap-google-analytics.

2. Visit the Google Analytics API dashboard, make sure that the project you used in the Create credentials step is selected and enable the API for your account.

A sample config for tap-google-analytics might look like this:

sample_config.json

{
  "key_file_location": "client_secrets.json",  // can also use `oauth_credentials`, see below
  "view_id": "123456789",
  "reports": "reports.json",
  "start_date": "2019-05-01T00:00:00Z",
  "end_date": "2019-06-01T00:00:00Z"
}

• view_id: The ID for the Google Analytics View you want to fetch data from.

  You can easily find it by using Google Analytics Account Explorer.

• start_date: The earliest date you want to fetch data for

• end_date: The last date for your report (not included). (Default: Today)

  We choose to follow an open [start_date, end_date) definition to easily chunk tap executions by month or week (using airflow or other tools).

  By using that approach, fetching data for May 2019 can be achieved by setting:

  "start_date": "2019-05-01T00:00:00Z",
  "end_date": "2019-06-01T00:00:00Z"
  

• key_file_location: If using Service Account based authorization, path to the client_secrets.json file you generated and downloaded during the "Create credentials" step.

• oauth_credentials: If using OAuth based authorization, a nested JSON object with the whole config looking like this:

{
  "oauth_credentials": {
      "access_token": "<ya29.GlxtB_access_token_gobbledegook>",
      "refresh_token": "<ya29.GlxtB_refresh_tokeN_gobbledegook>",
      "client_id": "<something.apps.googleusercontent.com>",
      "client_secret": "<some client secret string>"
  },
  "view_id": ...
}

• quota_user: an arbitrary string to use as an identifier for the user this library is being invoked on behalf of. Since the Google Analytics APIs have restrictive call quotas, they have a parameter for the API that allows server side applications like tap-google-analytics to make invocations under different users that each have a quota. See https://developers.google.com/analytics/devguides/reporting/core/v4/limits-quotas and https://developers.google.com/analytics/devguides/reporting/core/v4/parameters for more information. This parameter is only necessary if you are running into 429 Too Many Request errors from the GA API when running this tap.

• reports: path to a file with the definition of the reports to be generated.

If not provided and the tap runs without a --catalog also provided, use tap-google-analytics/defaults/default_report_definition.json as the default definition.

The reports.json file structure expected by the reports config key is really simple:

reports.json

[
  { "name" : "name of stream to be used",
    "dimensions" :
    [
      "Google Analytics Dimension",
      "Another Google Analytics Dimension",
      ... up to 7 dimensions per stream ...
    ],
    "metrics" :
    [
      "Google Analytics Metric",
      "Another Google Analytics Metric",
      ... up to 10 metrics per stream ...
    ]
  },
  {
  	... another stream definition ...
  },
  ... as many streams / reports as the user wants ...
]

For example, if you want to extract user stats per day in a users_per_day stream and session stats per day and country in a sessions_per_country_day stream:

[
  { "name" : "users_per_day",
    "dimensions" :
    [
      "ga:date"
    ],
    "metrics" :
    [
      "ga:users",
      "ga:newUsers"
    ]
  },
  { "name" : "sessions_per_country_day",
    "dimensions" :
    [
      "ga:date",
      "ga:country"
    ],
    "metrics" :
    [
      "ga:sessions",
      "ga:sessionsPerUser",
      "ga:avgSessionDuration"
    ]
  }
]

You can check tap-google-analytics/defaults/default_report_definition.json for a more lengthy, detailed example.

tap-google-analytics -c config.json

This tap makes some explicit decisions:

• All Google Analytics Metrics and Dimensions used are preserved with their name changed to ga_XXX from ga:XXX

  e.g. ga:date --> ga_date

• All dimensions are part of the stream's key definition (table-key-properties).

• The {start_date, end_date} parameters for the report query are also added to the schema as {report_start_date, report_end_date}.

  This is important for defining the date range the records are for, especially when 'ga:date' is not part of the requested Dimensions.

• If 'ga:date' has not been added as one of the Dimensions, then the {report_start_date, report_end_date} attributes are also added as keys.

  For example, if a user requests to see user stats by device or by source, the {start_date, end_date} can be used as part of the key uniquelly identifying the generated stats.

  That way we can properly identify rows with no date dimension included and also update those rows over overlapping runs of the tap.

• All streams and attributes are set to "inclusion": "automatic"

• There is a custom metadata keyword defined for all schema attributes:

  ga_type: dimension | metric

  This keyword is required for processing the catalog afterwards and generating the query to be send to GA API, as dimensions and metrics are not treated equally (they are sent as separate lists of attributes)

• The tap dynamically fetches the available dimensions and metrics and their data type. We have to use the Google Analytics API V3 for fetching those lists, as they are not provided in the Analytics Reporting API V4.

  We use those lists to check for invalid dimension or metric names requested by the user and inform her about those errors before the tap runs and starts making requests to the Google Analytics API.

  We also use the dynamically fetched dimensions and metrics lists to set the data type for those attributes and cast the values accordingly (in case of integer or numeric values)

Tap shortcomings (contributions are more than welcome):

• This tap does not currently use any STATE information for incrementally extracting data. This is currently mitigated by allowing for chunked runs using [start_date, end_date), but we should definitely add support for using STATE messages.

  The difficulty on that front is on dynamically deciding which attributes to use for capturing state for ad-hoc reports that do not include the ga:date dimension or other combinations of Time Dimensions.

• We may be checking for valid Google Analytics Metrics and Dimensions and report back to the user with errors, but we are not checking for valid combinations. Not all dimensions and metrics can be queried together and we should add a black list of metrics and dimensions that can not be used with specific dimensions (e.g. ga:impressions with ga:browser) or with other metrics (e.g. ga:XXdayUsers with other ga:XXdayUsers or ga:users).

  The Available Metrics/Dimensions and combos page provides a way to discover those restrictions.