RLA is set up like a simple pipe and filter system.
This allows you to easily add extra reports, filters and outputs.
-> Aggregator (database) Source -> Filter -> Filter -> Aggregator (summary report) -> Output -> Aggregator (...)
When the pipeline has been constructed, we Start chunk producer (source) and push requests through pipeline.
RequestLogAnalyzer::Source is an Object that pushes requests into the chain. At the moment you can only use the log-parser as a source. It accepts files or stdin and can parse then into request objects using a RequestLogAnalyzer::FileFormat definition. In the future we want to be able to have a generated request database as source as this will make interactive down drilling possible.
The filters are all subclasses of the RequestLogAnalyzer::Filter class. They accept a request object, manipulate or drop it, and then pass the request object on to the next filter in the chain. At the moment there are three types of filters available: Anonymize, Field and Timespan.
The Aggregators all inherit from the RequestLogAnalyzer::Aggregator class. All the requests that come out of the Filterchain are fed into all the aggregators in parallel. These aggregators can do anything what they want with the given request. For example: the Database aggregator will just store all the requests into a SQLite database while the Summarizer will generate a wide range of statistical reports from them.
Running the pipeline
All Aggregators are asked to report what they have done. For example the database will report: I stuffed x requests into SQLite database Y. The Summarizer will output its reports.
The output is pushed to a RequestLogAnalyzer::Output object, which takes care of the output. It can generate either ASCII, UTF8 or even HTML output.