[logback-dev] [JIRA] Created: (LBCLASSIC-217) Exception stack trace printing starting from root cause

[Tomasz Nurkiewicz (JIRA)]

Exception stack trace printing starting from root cause
-------------------------------------------------------

                 Key: LBCLASSIC-217
                 URL: http://jira.qos.ch/browse/LBCLASSIC-217
             Project: logback-classic
          Issue Type: New Feature
    Affects Versions: 0.9.22
            Reporter: Tomasz Nurkiewicz
            Assignee: Logback dev list


A typical exception in multi-tier application looks similar to this:

com.acme.BusinessException: Can't process request
	at ...
	at ...
Caused by: com.acme.dao.PersistenceException: Can't access database
	at ...
	at ...
	... 49 common frames omitted
Caused by: org.springframework.jdbc.datasource.lookup.DataSourceLookupFailureException: Unable to locate datasource
	at ...
	at ...
	... 65 common frames omitted
Caused by: java.net.UnknownHostException: Unknown host localhostt
	at ...
	at ...
	... 84 common frames omitted

The deeper exception, the more specific information it gives. Typically the last "Caused by" is the most interesting one. So it seems like reversing the order in which the exceptions are thrown (from most specific to consecutive, less specific wrappers) might be more intuitive:

java.net.UnknownHostException: Unknown host localhostt
	at ...
	at ...
Wrapped by: org.springframework.jdbc.datasource.lookup.DataSourceLookupFailureException: Unable to connect to the database
	at ...
	at ...
Wrapped by: com.acme.dao.PersistenceException: Can't access database
	at ...
	at ...
Wrapped by: com.acme.BusinessException: Can't process request
	at ...
	at ...

It is not only easier to read and follow (business exception is interesting for the end user while the most precise, technical information is for developers and system administrators - so it should be easily accessible in the logs), but also the stack trace isn't mixed. You can read stack frames in exactly the same order as they were executed, no need to jump from one stack trace line to another (see attachment from the real application). This also means that the beginning of the thread is always at the bottom and the line that caused the very first exception - at the top. (look at the attached real exception)

I already implemented this feature by adding RootCauseFirstThrowableProxyConverter extending ch.qos.logback.classic.pattern.ExtendedThrowableProxyConverter. It is turned on by appending "%rEx" to the encoder pattern. If you like this feature, I will push the changes (just few classes have changed) into my GitHub fork.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.qos.ch/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira