Skip to main content
Warning This document has not been updated for a while now. It may be out of date.
Last updated: 16 Jan 2018

manuals-publisher: Current state


The purpose of the Manuals Publisher application is to allow content editors to update Manuals - a type of long-form document.

The domain model for the application is quite simple. A Manual is composed of many Sections. Both the Manual and the Section have a version history which is known in the code base as an Edition (ManualRecord::Edition and SectionEdition concretely). Looking at editions of manuals and sections allows citizens to see when a document has changed in a significant way.

Sections can have file Attachments. Publishing a manual makes a call to the Publishing API (which updates the search index via Rummager). For historical reasons the application maintains its own state in MongoDB. The local state should reflect the state of the Publishing API's database at any particular time, although the application itself doesn't enforce this. A user of the application sees data read from the local MongoDB database, not the Publishing API.


The following diagram illustrates the associations between the models in the application

Class relationship diagram

This application models the "domain" and the "persistence" for Manuals and Sections separately. Unfortunately this separation is not rigidly and consistently held throughout the application

  • A "plain Ruby" class Manual is responsible for persisting its data into ManualRecord and ManualRecord::Edition.
  • A "plain Ruby" class Section is responsible for persisting its data into SectionEdition.
  • There is no section equivalent of ManualRecord (e.g. SectionRecord); the concept of a section is only persisted by one or more SectionEditions having a specific section_uuid.
  • A SectionEdition's attachments are modelled using Mongoid's embeds_many association (in the underlying database, they are stored as a nested array of objects under SectionEdition).
  • Some associations between objects are not modelled using Mongoid association methods (has_many, embeds_many, etc.) but rather rely on the presence of foreign keys, or arrays of foreign keys. For example the association between a ManualRecord::Edition and Section is via an array of section_uuids.
  • It is not possible to explicitly determine which edition of a section belongs to which edition of a manual, although this can be determined implicitly for the first and last edition of a manual.


The application has three controllers, one for each of the main domain objects in the application - ManualsController, SectionsController and SectionAttachmentsController. These controllers are largely unsurprising. Most of the logic for each controller action is encapsulated in Service objects.


Service classes in the app/services directories (not to be confused with the Services module in lib/services.rb) contain the logic for creating and updating Mongoid models, logging events and interacting with the Publishing API.

If you need to perform an action from outside of the context of a controller, for example in a one-off Rake task, ideally you should call a method on a Service instance as this ensures the correct sequence of side-effects.


The Adapter classes encapsulate the API calls to 2 external APIs: The (version 2) Publishing API and the Organisations API. These Adapters should be used in preference to making API calls directly from scripts and services.

The classes serve as application-specific wrappers around the GDS API adapters. The idea is that methods on adapters should accept and return instances of domain models and not Hashes/JSON. The translation between domain model objects and Hashes/JSON (and vice versa) should only happen inside the adapters.


The link-checker-api has been integrated on the show page for both Manuals and Sections. It extracts all links from within any GovSpeak fields and sends them to the /batch endpoint of the API. In this request it also sends across a webhook_callback which then populates any information about broken/warning links. More reading about the endpoints can be found here

Rake tasks

A small number of Rake tasks exist. These perform actions that are not exposed via the user interface, or are required to be infrequently run (for example report generation). Some commonly used Rake tasks are documented separately.