[
https://issues.apache.org/jira/browse/NUTCH-2184?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15058977#comment-15058977
]
Lewis John McGibbney commented on NUTCH-2184:
---------------------------------------------
To describe what this patch does
* implements much clearer command line logic (using commons-cli) for
IndexingJob
* permits continuation of IndexerMapReduce based upon exclusion of dbDatum ==
null in OR logic meaning that we can still index segment(s) even if they are
not accompanied by a crawldb CrawlDatum
* undertakes all necessary checks to ensure that a dbDatum object with value
null is never accessed
* implements the same error checking in all ScoringFilters which @override the
#indexerScore method ensuring that we never try to access a null value within
the MR job.
That's it folks. I've tried this with a bunch of test segments and I am now
able to index segments without a crawldb or linkdb.
It should be noted that this complies with backwards compatibility ensuring
that existing scripts, etc. will still work when running the index command.
> Enable IndexingJob to function with no crawldb
> ----------------------------------------------
>
> Key: NUTCH-2184
> URL: https://issues.apache.org/jira/browse/NUTCH-2184
> Project: Nutch
> Issue Type: Improvement
> Components: indexer
> Reporter: Lewis John McGibbney
> Assignee: Lewis John McGibbney
> Fix For: 1.12
>
> Attachments: NUTCH-2184.patch
>
>
> Sometimes when working with distributed team(s), we have found that we can
> 'loose' data structures which are currently considered as critical e.g.
> crawldb, linkdb and/or segments.
> In my current scenario I have a requirement to index segment data with no
> accompanying crawldb or linkdb.
> Absence of the latter is OK as linkdb is optional however currently in
> [IndexerMapReduce|https://github.com/apache/nutch/blob/trunk/src/java/org/apache/nutch/indexer/IndexerMapReduce.java]
> crawldb is mandatory.
> This ticket should enhance the IndexerMapReduce code to support the use case
> where you ONLY have segments and want to force an index for every record
> present.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
