Bug in Python 1.5.2 exception handling?

[Tim Peters]

[Dave Cole, suprised by locals surviving exit-by-exception]
> ...
> The reason that I was worried about this was that I am doing some
> database stuff using my Sybase module:
>
>         plug:  http://www.itga.com.au/~djc/sybase.html
>
> I wanted to automatically rollback a transaction when an exception
> was raised during processing.  I made a class like this:
>
> class Transaction:
>     def __init__(self, db, name):
>         self.db = db
>         self.name = name
>         self.db.execute('begin transaction %s' % (self.name,))
>         self.commit = 0
>
>     def __del__(self):
>         if self.commit:
>             self.db.execute('commit transaction %s' % (self.name,))
>         else:
>             self.db.execute('rollback transaction %s' % (self.name,))
>
> Then I thought that all I would have to do was:
>
>     def something_or_other(self):
>         tran = Transaction(self.db, 'update_id')
>         # lots o' database stuff
>         tran.commit = 1
>
> I suppose it will still work for the rollback case, but just not
> when I expect it to.

I doubt you want to chance that.

     def something_or_other(self):
         tran = Transaction(self.db, 'update_id')
         try:
             # lots o' database stuff
             tran.commit = 1
         finally:
             tran.dispose()  # contains what __del__ does now

is reliable under CPython or JPython.  I wish there were an even simpler way
to write these kinds of acquire/release pairs, but for now this is as good
as it gets.  It's not like it's *un*readable!  Quite clear, actually.  I
think it's just the "extra" indentation that irks me.

whitespace-is-a-precious-commodity-ly y'rs  - tim