Re: Boolean partitions syntax

On Tue, Dec 12, 2017 at 3:13 PM, Amit Langote
<Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp> wrote:
> On 2017/12/12 18:12, Ashutosh Bapat wrote:
>> On Tue, Dec 12, 2017 at 7:19 AM, Amit Langote wrote:
>>> Horiguchi-san pointed out [1] on a nearby thread that the partitioning
>>> syntax (the FOR VALUES clause) doesn't accept true and false as valid
>>> partition bound datums, which seems to me like an oversight. Attached a
>>> patch to fix that.
>>
>> May be you should use opt_boolean_or_string instead of TRUE_P and
>> FALSE_P. It also supports ON and OFF, which will be bonus.
>
> Thanks for the suggestion. I tried that but NonReservedWord_or_Sconst
> conflicts with Sconst that partbound_datum itself has a rule for,
> resulting in the following error:
>
> gram.y: conflicts: 6 reduce/reduce
> gram.y: expected 0 reduce/reduce conflicts
> gram.y:2769.25-81: warning: rule useless in parser due to conflicts:
> partbound_datum: Sconst

Probably that would get fixed if you remove Sconst from the
partition_datum and leave NonReservedWord_or_Sconst.

>
> Moreover, it seems like on/off are not being accepted as valid Boolean
> values like true/false are.
>
> insert into rp values (true);
> INSERT 0 1
> insert into rp values (on);
> ERROR: syntax error at or near "on"
> LINE 1: insert into rp values (on);
> ^
> What's going on with that? Maybe on/off values work only with SET
> statements?

But this is more important observation. Looks like on/off work with
SET and EXPLAIN only, not in normal SQL. So probably my suggestion was
wrongheaded.

--
Best Wishes,
Ashutosh Bapat
EnterpriseDB Corporation
The Postgres Database Company