Warcbase is an open-source platform for managing web archives built on Hadoop and HBase. The platform provides a flexible data model for storing and managing raw content as well as metadata and extracted knowledge. Tight integration with Hadoop provides powerful tools for analytics and data processing via Spark.
There are two main ways of using Warcbase:
- The first and most common is to analyze web archives using Spark.
- The second is to take advantage of HBase to provide random access as well as analytics capabilities. Random access allows Warcbase to provide temporal browsing of archived content (i.e., "wayback" functionality).
You can use Warcbase without HBase, and since HBase requires more extensive setup, it is recommended that if you're just starting out, play with the Spark analytics and don't worry about HBase.
Warcbase is built against CDH 5.4.1:
- Hadoop version: 2.6.0-cdh5.4.1
- HBase version: 1.0.0-cdh5.4.1
- Spark version: 1.3.0-cdh5.4.1
The Hadoop ecosystem is evolving rapidly, so there may be incompatibilities with other versions.
Detailed documentation is available in this repository's wiki.
Supporting files can be found in the warcbase-resources repository.
Clone the repo:
$ git clone http://github.com/lintool/warcbase.git
You can then build Warcbase:
$ mvn clean package appassembler:assemble
For the impatient, to skip tests:
$ mvn clean package appassembler:assemble -DskipTests
To create Eclipse project files:
$ mvn eclipse:clean
$ mvn eclipse:eclipse
You can then import the project into Eclipse.
To generate Scaladocs:
$ mvn scala:doc
Generated Scaladocs will be under the target/site directory
For the impatient, let's do a simple analysis with Spark. Within the repo there's already a sample ARC file stored at src/test/resources/arc/example.arc.gz.
If you need to install Spark, we have a walkthrough here for installation on OS X. This page also has instructions on how to get Spark Notebook, an interactive web-based editor, running.
Once you've got Spark installed, you can go ahead and fire up the Spark shell:
$ spark-shell --jars target/warcbase-0.1.0-SNAPSHOT-fatjar.jar
Here's a simple script that extracts and counts the top-level domains (i.e., number of pages for each top-level domain) in the sample ARC data:
import org.warcbase.spark.matchbox._
import org.warcbase.spark.rdd.RecordRDD._
val r = RecordLoader.loadArc("src/test/resources/arc/example.arc.gz", sc)
.keepValidPages()
.map(r => ExtractTopLevelDomain(r.getUrl))
.countItems()
.take(10)
Tip: By default, commands in the Spark shell must be one line. To run multi-line commands, type :paste in Spark shell: you can then copy-paste the script above directly into Spark shell. Use Ctrl-D to finish the command.
What to learn more? Check out analyzing web archives with Spark.
Warcbase was originally conceived with Pig for analytics, but we have transitioned over to Spark as the language of choice for scholarly interactions with web archive data. Spark has several advantages, including a cleaner interface, easier to write user-defined functions (UDFs), as well as integration with different "notebook" frontends.
The result of analyses of using Warcbase can serve as input to visualizations that help scholars interactively explore the data. Examples include:
- Basic crawl statistics from the Canadian Political Parties and Political Interest Groups collection.
- Interactive graph visualization using Gephi.
- Shine interface for faceted full-text search.
- Ingesting content into HBase: loading ARC and WARC data into HBase
- Warcbase/Wayback integration: guide to provide temporal browsing capabilities
- Warcbase Java tools: building the URL mapping, extracting the webgraph
Licensed under the Apache License, Version 2.0.
This work is supported in part by the National Science Foundation and by the Mellon Foundation (via Columbia University). Any opinions, findings, and conclusions or recommendations expressed are those of the researchers and do not necessarily reflect the views of the sponsors.