4.1. Designing the Search Engine

Dissecting the typical search engine reveals the anatomy of a simple beast in three parts: a crawler, an indexer, and a front end. A crawler goes out looking for fresh content to queue up for the indexer. In larger search engines, the crawler may actually download these pages and scan them for links to still more pages. The indexer in turn takes content, analyzes it, and formulates a searchable index. A front-end component of the search engine then accepts a query, uses it to search the index and presents the results back to the user. The devil, as always, is in the details.

The best algorithms to create an efficient index or rank results by perceived relevancy are closely guarded secrets in the search engine industry, and developing these is where a few lucky programmers have earned millions of dollars.

The full text-searching capabilities of MySQL make it possible to build low-grade search engines for small amounts of content, but this approach has some drawbacks as I will discuss. Instead, this project will have a crawler/indexer and front end.

The functionality of the crawler and indexer components will be combined in the same code file for this project. The document will be indexed to an inverted-index in a MySQL database immediately after it is retrieved. It will be written to run as a stand-alone cron job or scheduled task, the same way the scripts in the mailing list project from the previous chapter work.

The front end of the search engine will take in search terms from the user, query the index and return links to the appropriate documents. I will show you how to offer suggested alternatives for any misspelled words in the query to help make things more user-friendly, although this will require the pspell extension to be loaded in PHP.

The project will also have an administrative back end where an administrator can adjust the configuration of the crawler/indexer. Specifically, the admin will have the ability to define which documents are placed in the index and provide a list of stop words —common words such as a, is, and but that have little relevancy in a search query and should not be included in the index.