Re: LATERAL, UNNEST and spec compliance

* David Fetter (david(at)fetter(dot)org) wrote:
> As I see it, the current options are:
>
> 1. Do nothing, and insist on non-standard use of the LATERAL keyword.

I'm not a big fan of this. Providing a good error message saying "you
need to use LATERAL for this query to work" makes it slightly better,
but I don't feel like there's really any ambiguity here.

> 2. Add UNNEST to the grammar (or parse analysis) as a special case, making
> it implicitly LATERAL.
>
> (This would make implementing S301 easier, but special cases are ugly.)

This I really don't like.

> 3. Make all cases of SRFs in the FROM-clause implicitly LATERAL.
>
> (As far as I can tell, those cases whose behaviour would be changed by
> this actually produce errors in versions prior to 9.3, so no working
> code should be affected.)

+1 for me on this idea. If you're calling an SRF, passing in a lateral
value, 'LATERAL' seems like it's just a noise word, and apparently the
SQL authors felt the same, as they don't require it for unnest().

> Since LATERAL is new in 9.3, I think the pros and cons of these choices
> should be considered now, rather than being allowed to slide by unexamined.

I agree that we should really hammer this down before 9.3 is out the
door.

Thanks,

Stephen