W dniu 05.03.2016 o 01:03, David G. Johnston pisze:
[----------------]
>
> I've made my point and am not fluent enough to discuss the issues that
> would need to be addressed to implement a FK-to-a-partial-unique-index
> feature.
>
> I recall having hit this limitation myself previously so I too wouldn't
> mind seeing it implemented - but doing so without degrading the
> performance of all FK-related queries is important too - and the
> increase in risk of having bugs in the short term is quite high given
> that the fundamental operation of a key module needs to be changed. Its
> not something whose cost-benefit analysis reports favorably.
>
I see your point and mostly (not fuly, though) I agree. The questions I
raised aimed at weeding out "mathematical" constraints from
"implementation" one. I just wanted to know if its "impossible", or just
"hard/dangerous to do".
I think I have a broader picture now. Thenx for all the explanations,
-R