Thread: [Springframework-developer] somewhat confusing behavior w/ LSFB create/update/dropSchema methods.

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

I was punting around this for a good while before taking five minutes
to look at the source for the LocalSessionFactoryBean and having one
of those 'oh-doh' moments.

I now feel that the javadocs for those methods are misleading.  What I
was working on was a setup module for our webapp that would
create/initialize the schema via a web-based UI.. and was using the
LSFB methods to facilate that functionality.

What I wanted to do was trap any exceptions that arose from the schema
methods so that they could be displayed on the UI if there was a
schema creation error.   Looking at the javadocs for those schema
methods, in particular the 'Throws' part in the javadoc where it says
'DataAccessException - in case of script execution errors' it seemed
natural to assume that catching DataAccessException would do the
trick.  But no matter how many sacrifical lambs i put up, I was unable
to trap *any* kind of exception, even though I could see the exception
output in the logs.

I was starting to question my competence as a programmer and my sanity.

So after looking thru the source for the LSFB, I see that those schema
methods call the 'executeSchemaStatement' and 'executeSchemaScript'
methods internally.  And to their credit to some degree, the javadocs
for those methods do mention this:

"Note that the default implementation will log unsuccessful statements
and continue to execute. Override the executeSchemaStatement method to
treat failures differently."

So THAT explains why I wasn't able to trap the exception being thrown,
and was seeing them in the logs.

Simply by looking at the javadocs by itself, a programmer (or at least
a simple one like me) wouldn't immediately make the association that
those schema methods would call the executeSchemaStatements internally
and that any schema errors would be logged and that DAE would not be
thrown at all.  That was only obvious after looking at the source
code.

So it would seem that my only option is to subclass LSFB and modify
the behavior of those to collect the error messages and pass them on
up.  I would think, however that those methods could benefit from
clearer javadoc comments on what I've just covered, and it would be
nice to be able to obtain that output (or to at least know somehow) to
be able to determine if any errors were generated, and have some sort
of output on the reasons why, since there is no way to accomplish that
right now without subclassing the LSFB itself (as far as I can see) -
and I'm somewhat leery of having to maintain a subclass of LSFB for
the entire foreseeable future of the application since it is such a
critical part of the spring/hibernate functionality.  If I'm mistaken
on this, I'd greatly appreciate a point in the right direction.

Also, slightly related, but is there a reason why
'updateDatabaseSchema' throws HibernateException while the other two
'create'/'dropDatabaseSchema' methods throw DataAccessException?

Thanks!

-tim

Thread: [Springframework-developer] somewhat confusing behavior w/ LSFB create/update/dropSchema methods.

springframework-developer