Re: hstore parser incorrectly handles malformed input

On Thu, Apr 26, 2012 at 9:12 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> I wrote:
>> Ryan Kelly <rpkelly22(at)gmail(dot)com> writes:
>>> In my mind, all of these should have been rejected as erroneous input.
>>> To that end, I have attached a patch which causes all of these inputs
>>> to be rejected as invalid.
>
>> Hm, I would have expected all three of these to result in "a" having
>> an empty-string value.  I see nothing in the hstore documentation
>> suggesting that I must write a=>"" or some such to get an empty value,
>
> Attached is an alternative patch that fixes it that way.
>
> Does anybody else have an opinion as to which of these solutions is
> more preferable?  And should we regard this as a back-patchable bug
> fix, or a definition change suitable only for HEAD?

I vote for not back-patching, regardless of exactly what we decide to do here.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company