Extracting Fetched Web Pages from Nutch in a Map R

问题:

After a Nutch crawl in distributed (deploy) mode as follows:

bin/nutch crawl s3n://..... -depth 10 -topN 50000 -dir /crawl -threads 20 

I need to extract each URL fetched along with it's content in a map reduce friendly format. By using the readseg command below, the contents are fetched but the output format doesn't lend itself to being map reduced.

bin/nutch readseg -dump /crawl/segments/*  /output  -nogenerate -noparse -noparsedata -noparsetext

Ideally the output should be in this format:

http://abc.com/1     content of http://abc.com/1
http://abc.com/2     content of http://abc.com/2

Any suggestions on how to achieve this?

回答1:

The bin/nutch readseg command produces output in a human readable format and not map reduce format. The data is stored in segments in map-reduce format. I dont think that you can directly pull out that info from segements in map-reduce format.

Few options for your concern:

1. Segments are themselves map-reduce format files. Can you re-use those ?
2. The output of readseg command can be converted to map-reduce form by writing a small map-reduce code.

回答2:

The answer lies in tweaking the source code of nutch. This turned out to be quite simple. Navigate to the SegmentReader.java file at apache-nutch-1.4-bin/src/java/org/apache/nutch/segment

Inside the SegmentReader class is a method reduce which is responsible for generating the human readable output the bin/nutch readseg command generates. Alter the StringBuffer dump variable as you see fit - this holds the entire output for a given url which is represented by the key variable.

Make sure you to run ant to create a new binary and further calls to bin/nutch readseg shall generate the output in your custom format.

These references were extremely useful in navigating the code:
[1] http://nutch.apache.org/apidocs-1.4/overview-summary.html
[2] http://nutch.apache.org/apidocs-1.3/index-all.html