The general principal of the publisher framework (PF) is the same as connector framework. There is a generic publisher component that calls in to a repository specific provider to access the repository to which content is being published. In the case of standard “targets” (Solr, Elasticsearch, Sharepoint via Stager etc), the provider is also supported by PF. However, a developer is able to create his own if required to publish to a customer specific target, only needing to consider how to perform actions at the target, rather than needing to consider all the general functionality such as when a new batch should be used. All common functionality (connections, batch handling, commit/clear jobs etc) is handled by the framework and this call methods in the provider as required.
The user selects a publisher jar to load – this is a component (aspire-XXX-publisher). There is a common app bundle that is automatically loaded when required. The app bundle loads the publisher framework jar and the originally requested provider.
The framework has the ability to perform optional groovy or xml transforms and the appropriate parameters are collected by the framework. However, the actual use of the transform is controlled by the developer.
Where possible, connections are pooled.
Similar to the “SourceInfo”, the framework uses a “PublisherInfo”. This holds information used to connect to the target repository (url, username, password etc) but also controls the functionality the framework provides. For example, the framework allows for a transformation, but a connector may not require this functionality, so the developer could disable it. . The developer is able to extend this if required, but the framework allows for the following configuration.
When a connection is required
Pool connections
True/false that connections should be pooled
Use transform
Transform type – none/xml/json + default transform file (to pick out of component)
Supports authentication
http(s)
We implemented a set of properties in the provider that control the DXF in the app bundle and the options in the framework. This could then be used to control the app bundle loaded (this is via the aspire application), and the configuration of the component “publisher info” (for example to control if the publisher supports “clear” and “commit” operations. The same configuration could then we used, via the dxf, to offer the option to (say) “process commits” only if it’s supported
Installation settings are collected when the component is installed. The obvious items required are the location and connection (user/password) details of the target. The desire is that only options that the developer has enabled (in the developer settings above) will be presented to the user. The settings are collected using DXF. A publisher specific DXF is merged with a common piece to present the entire set. The framework collects the following parameters:
Target URL
The url for the search engine etc
Authentication
Yes/no/type
Gather username/password
Clear before full crawls
True/false. If true, the publisher will react to start jobs for full crawl by calling a clear method
Commit after crawls
True/false. If true, the publisher will react to end jobs for crawls by calling a commit method
Transform data before sending
True/false
Transform file name
For cases when transformation is required
On startup, the framework connects to the provider and calls a method “newPublisherInfo”. This returns a class (much like the SourceInfo) holding all the configuration for the publisher (including the common options – perform clear etc). This can be passed to other calls later. If required, connection pools will be initialised here.
When processing a document, the framework first categorises the job in to “control” or “document”. “Control” jobs are commit and clear and the framework calls the provider’s commit or clear methods (if enabled and processing is selected) passing a connection as required. For “document” jobs, the framework will decide whether a new batch is required and calls the provider’s startBatch() method. The framework provides “standard” batch implementations.
Then the provider establishes the specific type of job (add/update, delete or delete by query) and call the appropriate provider method.
Closing the component release all of the connections.
Control of the DXF etc is via a new properties file added to the component. This file allows the developer to add properties. Whilst these properties are generic, the publisher framework looks for certain properties and use these to default the various settings in the PublisherInfo. These properties are passed to the dxf with the rest of the configuration, allowing control of the options shown to the installer. These properties also allows control of the app-bundle loaded when the component is selected.