Tom Lane wrote:
>>> By the way, PL/SQL screams if you want to do an assignment with '='.
>>> But I guess that the current behaviour of PL/pgSQL should not reflect
>>> that to maintain backward compatibility, right?
>>>
>
>
>> I think it should. The current behaviour is undocumented and more than icky.
>>
>
> The lack of documentation is easily fixed ;-). I don't think this is
> icky enough to justify breaking all the existing functions we'd
> undoubtedly break if we changed it.
>
>
We have tightened behaviour in ways much harder to fix in the past, e.g.
actually following UTF8 rules. Fixing breakage in this case would be
pretty trivial, and nobody has any real right to expect the current
behaviour to work.
But I won't be surprised to be in a minority on this ....
cheers
andrew