[
https://issues.apache.org/jira/browse/NUTCH-1481?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13484927#comment-13484927
]
Arni Sumarlidason commented on NUTCH-1481:
------------------------------------------
I was going to post a work around, but haven't decided where. I actually
modified
gora-sql-0.1.1-incubating.jar::org.apache.gora.sql.store.SqlStore.flush() to be
"less sensitive" to failures. I logged all the URLS that were skipped, around
0.0003% of 150,000 so far, some what insignificant. However, I think if users
are going to use less than 4b/character for their 'id' field I don't think the
job should fail.
I haven't been able to spend a ton of time familiarizing myself with the nutch
source, but it may be possible to add a parameter to webpage that is a hash of
the key and setting it to unique in the database. Then we can remove the
primary key constraint from ID and use a utf8mb4 character set with a good
length.
I am operating under the assumption that the primary key's sole purpose is to
reject duplicates in the database.
please advise.
> When using MySQL as storage unicode characters within URLS cause nutch to fail
> ------------------------------------------------------------------------------
>
> Key: NUTCH-1481
> URL: https://issues.apache.org/jira/browse/NUTCH-1481
> Project: Nutch
> Issue Type: Bug
> Components: crawldb
> Affects Versions: 2.1
> Environment: mysql 5.5.28 on centos
> Reporter: Arni Sumarlidason
> Labels: database, sql, unicode, utf8
>
> MySQL's (innodb) primary key / unique key is restricted to 767 bytes..
> currently the url of a web page is used as a primary key in nutch storage.
> when using latin1 character set on the 'id' column @ length 767
> bytes/characters; unicode characters in urls cause jdbc to throw an exception,
> java.io.IOException: java.sql.BatchUpdateException: Incorrect string value:
> '\xE2\x80\x8' for column 'id' at row 1
> when using utf8mb4 character set on the 'id' column @ length 190 characters /
> 760 bytes to fully support unicode characters; the field length becomes
> insufficient
> It may be better to use a hash of the url as the primary key instead of the
> url itself. This would allow urls of any length and full utf8 support.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
