nutch-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Marko Bauhardt (JIRA)" <j...@apache.org>
Subject [jira] Commented: (NUTCH-747) inject&Index metadatas and inherit these metadatas to all matching suburls
Date Thu, 06 Aug 2009 10:46:14 GMT

    [ https://issues.apache.org/jira/browse/NUTCH-747?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12739982#action_12739982
] 

Marko Bauhardt commented on NUTCH-747:
--------------------------------------

To update the parse_data.metadata is maybe not the best solution, but i don't want to change
the input path's for the indexer.
so maybe a new crawl_metadata or fetch_metadata path is better to store metadatas. but then
we have to change the input path for the indexer. 
and i think metadatas should be stored in a path inside the shard instead in the crawldb.

> inject&Index metadatas and inherit these metadatas to all matching suburls
> --------------------------------------------------------------------------
>
>                 Key: NUTCH-747
>                 URL: https://issues.apache.org/jira/browse/NUTCH-747
>             Project: Nutch
>          Issue Type: Improvement
>          Components: indexer, injector
>            Reporter: Marko Bauhardt
>             Fix For: 1.1
>
>         Attachments: index-metadata.patch, metadata.patch
>
>
> Hi.
> the following two patches supports
> + inject metadatas to url's into a metadatadb
> url.com <TAB> <METAKEY> : <TAB> <METAVALUE> <TAB> <METAVALUE>
<METAKEY> : <METAVALUE> ...
> ...
> + updates the parse_data metadata from a shard and write the metadatas to all fetched
urls that starts with an url from the metadatadb
> + this patch support's metadata to all matching suburls inheritance
> the second patch implements a index-metadata plugin.
> + this plugin extract all metadats from the parse_data of a shard and index it. which
metadats you can configure in the plugin.properties.
> + to index for example the lang you have to configure the plugin.properties: lang=STORE,UNTOKENIZED
> + that means that the index plugin exract metadata values with key "lang". if exists,
all values are indexed stored and untokenized
> Example
> create start url's in "/tmp/urls/start/urls.txt"
> http://lucene.apache.org/nutch/apidocs-1.0/index.html
> http://lucene.apache.org/nutch/apidocs-0.9/index.html
> create metadata url's in "/tmp/urls/metadata/urls.txt"
> http://lucene.apache.org/nutch/apidocs-1.0/     version:        1.0
> http://lucene.apache.org/nutch/apidocs-0.9/     version:        0.9
> Inject Urls
> bin/nutch inject crawldb /tmp/urls/start/
> bin/nutch org.apache.nutch.crawl.metadata.MetadataInjector metadatadb /tmp/urls/metadata/
> Fetch & Parse & Update
> bin/nutch generate crawldb segments
> bin/nutch fetch segments/20090806105717/
> bin/nutch org.apache.nutch.crawl.metadata.ParseDataUpdater metadatadb segments/20090806105717
> bin/nutch updatedb crawldb/ segments/20090806105717/
> Fetch & Parse & Update Again
> ...
> Index
> bin/nutch invertlinks linkdb -dir segments/
> bin/nutch index index crawldb/ linkdb/ segments/20090806105717 segments/20090806110127
> Check your Index
> All urls starting with "http://lucene.apache.org/nutch/apidocs-1.0/ " are indexed with
"version:1.0".
> All urls starting with "http://lucene.apache.org/nutch/apidocs-0.9/ " are indexed with
"version:0.9".
> This issue is some related to http://issues.apache.org/jira/browse/NUTCH-655

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message