[rfc] avoid assert statements?

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Martin Pool wrote:
> The question of when it's ok to use 'assert' has come up in a few
> reviews recently; Robert suggested we should set a standard for it.  I
> wouldn't say it's hugely urgent but thought I'd write this down rather
> than forget it.
> 
> As you probably know, assert is equivalent to
> 
>   if __debug__:
>     if not CONDITION:
>       raise AssertionError(MESSAGE)
> 
> where __debug__ is controlled by the Python -O flag, which is normally absent.

I tend to think of "assert" as a check for developers, which is meant to
be disabled for most people at runtime.

That way when I'm prototyping something new I'll get an assertion error
if I ask the wrong thing.

I would be happy enough to have the default installer set the '-O' flag.
 So that running from source would be normal, but running from an
installed bzr would have assertions removed.

I agree that there are places which are either "this always needs to be
check, so check for real and raise a real error" and there are some
which are "I'm prototyping, so let me check my assumptions" which we
don't need in production code anymore.

I don't really want to see a general "don't use assert" because I think
it is a useful tool.

But then again, maybe it is only really useful for developing/debugging,
and thus shouldn't ever be in bzr.dev....

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHARV1JdeBCYSNAAMRAvd/AKDZEaP/NE02Ch1xrbEpR/dGLrhfSQCfcwD7
RcL4srHSagVWM4/l76ch+BM=
=dzpl
-----END PGP SIGNATURE-----