[
https://issues.apache.org/jira/browse/NUTCH-1078?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13113567#comment-13113567
]
Lewis John McGibbney commented on NUTCH-1078:
---------------------------------------------
OK so this kinda opened up a couple of avenues. I am interested to see how we
review and act on this one. Having discussed our problem with some slf4j users,
the following seems to sum up the options.
----------------------------------
The problem has to do with how LogUtil is implemented.
Here are two possible solutions.
1) Get rid of the LogUtil class. For example, in HttpBase, instead of
{code}
} catch (Throwable e) {
e.printStackTrace(LogUtil.getErrorStream(logger));
return new ProtocolOutput(null, new ProtocolStatus(e));
}
{code}
write instead
{code}
} catch (Throwable e) {
logger.error("Failed to get protocol output", e);
return new ProtocolOutput(null, new ProtocolStatus(e));
}
{code}
This latter form is the standard way of logging exceptions and this is the
format some other Nutch code uses. I noticed this when I was creating the
various patches.
We would need to repeat the above operation in all places where LogUtil is
used. There are about 60 such places located in about a dozen classes.
2) In LogUtil, replace
{code}
static {
try {
TRACE=Logger.class.getMethod("trace", new Class[] { Object.class });
DEBUG=Logger.class.getMethod("debug", new Class[] { Object.class });
INFO=Logger.class.getMethod("info", new Class[] { Object.class });
} catch ...
}
}
{code}
with
{code}
static {
try {
TRACE=Logger.class.getMethod("trace", new Class[] {String.class});
DEBUG=Logger.class.getMethod("debug", new Class[] {String.class});
INFO=Logger.class.getMethod("info", new Class[] {String.class});
} catch ...
}
}
{code}
On line 104 change
{code}
method.invoke(logger, new Object[] {toString().trim() });
{code}
to
{code}
method.invoke(logger, new String[] {toString().trim() });
{code}
Further to this, a suggestion was raised that unless there is a good reason not
to, it might be an idea to get rid of LogUtil in from the Nutch code base
altogether as it seems like a very convoluted way for logging exceptions. I
(and the kind folks from slf4j who I spoke with) by no means understand every
aspect of why we currently use LogUtil, therefore I really need some opinions
from you guys to sort this one out. Thanks
> Upgrade all instances of commons logging to slf4j (with log4j backend)
> ----------------------------------------------------------------------
>
> Key: NUTCH-1078
> URL: https://issues.apache.org/jira/browse/NUTCH-1078
> Project: Nutch
> Issue Type: Improvement
> Affects Versions: 1.4
> Reporter: Lewis John McGibbney
> Assignee: Lewis John McGibbney
> Priority: Minor
> Fix For: 1.4
>
> Attachments: NUTCH-1078-branch-1.4-20110816.patch,
> NUTCH-1078-branch-1.4-20110824-v2.patch,
> NUTCH-1078-branch-1.4-20110911-v3.patch,
> NUTCH-1078-branch-1.4-20110916-v4.patch
>
>
> Whilst working on another issue, I noticed that some classes still import and
> use commons logging for example HttpBase.java
> {code}
> import java.util.*;
> // Commons Logging imports
> import org.apache.commons.logging.Log;
> import org.apache.commons.logging.LogFactory;
> // Nutch imports
> import org.apache.nutch.crawl.CrawlDatum;
> {code}
> At this stage I am unsure how many (if any others) still import and reply
> upon commons logging, however they should be upgraded to slf4j for branch-1.4.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
