Scoogle is a Search Engine written in Scala and using Akka Actors. Its components are a web crawler, a HBase data store, a server serving a backend API and a frontend written in React. The HBase master, ZooKeeper and servers need to be started first, either locally or remotely, either on the local filesystem or on HDFS. The address and port number of the ZooKeeper Quorum server are given to the Scoogle web crawler and server.

The web crawler takes as input the seed in the shape of one or more XML files. Find below an example of such a file. The different fields designate the following:

• link The link to a source. This can be a file on the local filesystem or a http(s) website. The web crawler will take all the links as a seed.
• depth The depth up to which each source will be crawled.

<sources>
    <source>
        <link>https://www.wikipedia.com</link>
        <depth>3</depth>
    </source>
    <source>
        <link>file:///path/to/file.html</link>
        <depth>1</depth>
    </source>
</sources>

Ideally the program should be terminated with a SIGTERM signal rather than SIGINT, SIGABRT or SIGKILL so that the HBase connection can close gracefully and flush buffered data.

The web crawler exposes Kamon metrics in a Prometheus format on a scraping endpoint on http://localhost:9095. A Prometheus server can be started in order to view the various Akka metrics.

The web crawler is able to crawl and store approximately 35 pages per second on a e2-standard-4 GCE instance (4 vCPUs, 16 GB memory).

After running mvn clean package, run java -jar ./target/web-crawler.jar [options] [source ...]

The options are

• --zooKeeperAddress The address of the ZooKeeper Quorum server. Defaults to localhost.
• --zooKeeperPort The port of the ZooKeeper Quorum server. Defaults to 2181.
• --maxConcurrentSockets The maximum number of sockets that the program will open simultaneously. Defaults to 30. Sometimes if this number is too large, the requests will time out.
• -h, --help Show help message.

followed by a list of source xml files in the format described above.

The server will try to find the webpages that maximize the number of searchbar keywords occurring in it. The keywords are reduced to their stem form. For example, eating or eaten will be mapped to the stem eat. After that, the keywords are mapped to a possible synonym. For example magenta is mapped to red. This was achieved using the Apache Lucene core library. All the keywords typed into the searchbar must occur at least once in the webpage for it to appear in the result list.

A special feature of the searchbar is to surround keywords with quotes (e.g. "apple"). Both quoted and unquoted keywords must appear at least once in the webpage, but only quoted keywords will be highlighted in the result list. If there are no quoted keywords, all the unquoted keywords are highlighted.

The server is extremely fast in retrieving the list of webpages from a query (usually displays the results in less than a second). The HBase database maintains an inverted index table indexing keywords to the webpage they occur in. This table is indexed on the keywords, rendering a keyword lookup extremely fast. The latest query is always cached which means that upon clicking on the pagination buttons, the query does not need to be processed again, and the webpages do not need to be fetched again from the database.

An Akka HTTP server runs and serves the following routes:

• / The frontend consisting of the static files inside the /build folder
• /api Given a query HTTP parameter representing a list of space-separated keywords and a pageNumber parameter, this route will respond with a list of at most ten links that match against those keywords for the given pageNumber.

After running mvn clean package, run java -jar ./target/server.jar [options]

The options are

• --serverInterface The interface on which the server listens. Defaults to localhost.
• --serverPort The port number on which the server listens. Defaults to 8080.
• --zooKeeperAddress The address of the ZooKeeper Quorum server. Defaults to localhost.
• --zooKeeperPort The port of the ZooKeeper Quorum server. Defaults to 2181.
• -h, --help Show help message.

Inside the frontend directory, run npm start in order to serve the frontend in developer mode. Run npm run build to build the production files. When the build has completed, the static files will be located inside the target directory of the maven project and will be served by the Server class.

The frontend consists of a home page with a searchbar. Upon hitting a Search button, the frontend displays a list of websites matching the keywords put in the searchbar. The pagination groups the links by 10.