Introduction
An aspect engine builds on the flow element concept to introduce the ability to:
- cache processing results
- handle cases where a property has not been populated by an aspect engine but it could be if certain configuration changes were made.
- lazily load values in the flow data
In the same way a flow element can produce an element data containing results of its processing, an engine can produce an aspect data. This builds on element data to add some of the extra functionality listed above.
Caching
For performance reasons, it may be desirable to add results caching to an engine which has a particularly large overhead for processing.
Logic in an engine makes use of a cache passed at construction time to avoid doing any processing when a result has already been found for a certain set of evidence. The key for this cache is a data key generated using the evidence key filter from the engine. This means that only evidence relevant to the engine is used in the key for the cache, making the cache more space-efficient.
Missing Property Handling
An engine adds handling of cases where a value is requested for a property which is not available for some reason.
Consider the case where an engine is set up with free data, and the user attempts to retrieve a value for an enterprise property. An error would be returned stating that the property was only available in a the enterprise data tier.
Consider another case where an engine has been built with specific properties included rather than the full set, and the user attempts to retrieve a value for a property which was not included. An error would be returned stating that the engine was not configured correctly to return values for that property.
Both these cases are distinct from that where the engine simply does not know what the property is, and cannot return values for it. So it is important to distinguish between these scenarios and be clear to the user about what they need to do to get access to the property they want.
Lazy Loading
The aspect data produced by an engine can be returned before it has been completely populated. This means that the caller does not need to wait for processing to finish before carrying on with other things, and the processing will continue in the background. This is particularly useful in a web server where processing can start the instant a web request is received and the server can continue serving the page until it reaches a point where the result of the processing is needed.
By default lazy loading is not enabled, but can be via the engine's builder.
Note that lazy loading is only available in a subset of languages. This is because it either is not possible or doesn't make sense in some cases. For example, PHP is single-threaded so lazy loading is not possible. The Node.js Pipeline is entirely asynchronous so a lazy loading capability is unnecessary.