Re: [Fab-user] I bet it would be pretty to do transaction semantics but it sure would be nice

[Christian Vest Hansen]

On Fri, Jul 24, 2009 at 1:43 PM, Jeff Forcier<address@hidden> wrote:
> On Fri, Jul 24, 2009 at 2:56 AM, Christian Vest
> Hansen<address@hidden> wrote:
>> My tasks are written such that the commands I execute on the servers
>> are "effectively idempotent" to the greatest extent possible, so I can
>> rerun them if I need to.
>
> This is what I do as well.
>
>> In some cases I might decide to catch the
>> SystemExit exception that utils.abort raises, even though it's a
>> butt-ugly practice.
>
> Is that butt-ugly because it raises SystemExit or is it something
> else? :)

Well, it's sort of an implementation detail of abort, is it not? Also,
the behavior of _handle_failure is configurable, so I can easily break
my own assumptions if I'm not careful.

> At one point I started using a custom exception class but
> there didn't seem to be much of a point to it in that particular
> situation

I guess that is true. Aborting is pretty much just stopping python,
which is what SystemExit does.

>
>> Jeff, have you heard about the condition system in Common Lisp?
>
> I haven't, no, my only Lisp experience was a semester of Scheme about
> eight years ago and I don't recall hearing about that. I can google
> it, or you can throw me a link

http://gigamonkeys.com/book/beyond-exception-handling-conditions-and-restarts.html
(common lisp)

> or two

http://209.85.229.132/search?q=cache:cAxMwFh6J4AJ:www.pragprog.com/magazines/download/1.pdf&cd=10&hl=da&ct=clnk&gl=dk#20
(clojure.contrib.error-kit is built on the same concepts)

> if you want.
>
> -Jeff
>



-- 
Venlig hilsen / Kind regards,
Christian Vest Hansen.