Re: Declarative partitioning

On Mon, Mar 7, 2016 at 8:39 PM, Amit Langote
<Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp> wrote:
>> Which means your creates would look like (following Robert Haas's implied
>> suggestion that we leave off the string literal quotes):
>>
>> CREATE TABLE foo_ax1x PARTITION OF foo FOR VALUES ( , (b,2) );
>> CREATE TABLE foo_ax1x PARTITION OF foo FOR VALUES [ (b,2), (b,3) );
>> CREATE TABLE foo_ax1x PARTITION OF foo FOR VALUES [ (b,3), (b,4) );
>> CREATE TABLE foo_ax1x PARTITION OF foo FOR VALUES [ (b,4), (c,2) );
>> CREATE TABLE foo_ax1x PARTITION OF foo FOR VALUES [ (c,2), (c,3) );
>> CREATE TABLE foo_ax1x PARTITION OF foo FOR VALUES [ (c,3), (c,4) );
>>
>> That's not terrible looking.
>
> So I tried a grammar that looked like the following:
>
> range_spec: lb_inc bound ',' bound ub_inc {<create-range-struct>}
>
> lb_inc: '[' { $$ = true; } | '(' { $$ = false; }
> ub_inc: ']' { $$ = true; } | ')' { $$ = false; }
>
> bound: a_expr
> {
> if (IsA($1, RowExpr)) /* (key1, key2, ...) */
> $$ = ((RowExpr) $1)->args;
> else /* key */
> $$ = list_make1($1);
> }
> | /* EMPTY */ { $$ = NIL; }

I think we're converging on a good syntax, but I don't think the
choice of nothingness to represent an open range is a good idea, both
because it will probably create grammar conflicts now or later and
also because it actually is sort of confusing and unintuitive to read
given the rest of our syntax. I suggest using UNBOUNDED instead.

Also, I don't think allowing an a_expr as a bound is remotely sensible
- I think you should allow only Sconst, NumericOnly, NULL, and
UNBOUNDED.

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