Experiment designed to reflect on the organization of the 3 important elements of web scraping/data processing:
Optionally, monitoring might be added to each step.
Scheduling includes the configuration & tracking of when scrapping and processing should take place.
Scraping is organized by target and each "scraper" responds to
#run which returns an array of objects which will each respond to
#to_json. Scrapers shouldn't stop the execution, in other words, they
shouldn't raise exception but instead handle errors internally and
expose the issues via data structures.
The term scraping isn't quite right in the context of this experiment
since a scraper could simply be an API client consuming a web API for
Once the data is scraped, the fetched information can be sent to one or many processors until it reaches its final form. If the data is meant to be persisted, the persistence layer should be implemented as a processor. Examples of processors are data extractors, event triggers, persistence layers etc...
Each unit should be autonomous, easy to test and chainable. Raised exceptions should really be exceptional and always mean that human intervention is needed. Ideally, each unit should also be designed to run concurrently.