Re: Proof of concept: auto updatable views [Review of Patch]

Thanks for looking at this.
Attached is a rebased patch using new OIDs.

On 11 October 2012 02:39, Peter Eisentraut <peter_e(at)gmx(dot)net> wrote:
> Compiler warning needs to be fixed:
>
> rewriteHandler.c: In function 'RewriteQuery':
> rewriteHandler.c:2153:29: error: 'view_rte' may be used uninitialized in this function [-Werror=maybe-uninitialized]
> rewriteHandler.c:2015:17: note: 'view_rte' was declared here
>

Ah, my version of gcc doesn't give that warning. Looking at the code
afresh though, I think that code block is pretty ugly. The attached
version rewrites that block in a more compact form, which I think is
also much more readable, and should cure the compiler warning.

> Maybe we should distinguish updatable from insertable in error messages
> like this one:
>
> ERROR: cannot insert into view "foov2"
> DETAIL: Views containing DISTINCT are not updatable.
>
> The SQL standard distinguishes the two, so there could be differences.
> I'm not sure what they are right now, though.
>
> This hint could use some refreshing:
>
> HINT: You need an unconditional ON INSERT DO INSTEAD rule or an INSTEAD OF INSERT trigger.
>
> Maybe something along the lines of
>
> HINT: To make the view insertable anyway, supply an unconditional ... etc.
>

I've not updated the error messages - I need to think about that a bit more.

Regards,
Dean