On Tue, Sep 24, 2019 at 12:38:30PM -0300, Alvaro Herrera wrote:
> On 2019-Sep-24, Nikolay Shaplov wrote:
>> В письме от вторник, 24 сентября 2019 г. 9:25:54 MSK пользователь Alvaro
>> Herrera написал:
>>> 0003 looks useful, thanks for completing it. I think it would be a good
>>> idea to test invalid values for each type of reloption too (passing
>>> floating point to integers, strings to floating point, and so on).
>>
>> We already do it in reloption regression tests.
>>
>> My idea was to test here only the things that can't be tested in regression
>> tests, on in real indexes like bloom.
>
> I suppose that makes sense. But of course when I push enum reloptions
> I will have to add such a test, since bloom does not have one.
Good point. We rely now on the GUC parsing for reloptions, so having
cross-checks about what patterns are allowed or not is a good idea for
all reloption types. I have added all that, and committed the
module. The amount of noise generated by the string validator routine
was a bit annoying, so I have silenced them where they don't really
matter (basically everything except the initial creation).
--
Michael