Re: Adding support for Default partition in partitioning

On Thu, Sep 7, 2017 at 3:15 PM, Rajkumar Raghuwanshi <
rajkumar(dot)raghuwanshi(at)enterprisedb(dot)com> wrote:

> On Wed, Sep 6, 2017 at 5:25 PM, Jeevan Ladhe <
> jeevan(dot)ladhe(at)enterprisedb(dot)com> wrote:
>
>> Hi,
>>
>> Attached is the rebased set of patches.
>> Robert has committed[1] patch 0001 in V26 series, hence the patch
>> numbering
>> in V27 is now decreased by 1 for each patch as compared to V26.
>>
>
> Hi,
>
> I have applied v27 patches and while testing got below observation.
>
> Observation: in below partition table, d1 constraints not allowing NULL to
> be inserted in b column
> but I am able to insert it.
>
> steps to reproduce:
> create table d0 (a int, b int) partition by range(a,b);
> create table d1 partition of d0 for values from (0,0) to
> (maxvalue,maxvalue);
>
> postgres=# insert into d0 values (0,null);
> INSERT 0 1
> postgres=# \d+ d1
> Table "public.d1"
> Column | Type | Collation | Nullable | Default | Storage | Stats
> target | Description
> --------+---------+-----------+----------+---------+--------
> -+--------------+-------------
> a | integer | | | | plain
> | |
> b | integer | | | | plain
> | |
> Partition of: d0 FOR VALUES FROM (0, 0) TO (MAXVALUE, MAXVALUE)
> Partition constraint: ((a IS NOT NULL) AND *(b IS NOT NULL) *AND ((a > 0)
> OR ((a = 0) AND (b >= 0))))
>
> postgres=# select tableoid::regclass,* from d0;
> tableoid | a | b
> ----------+---+---
> *d1 | 0 | *
> (1 row)
>

Good catch. Thanks Rajkumar.
This seems to be happening because of the following changes made in
get_partition_for_tuple() for default range partition support as part of
patch 0002.

@@ -1971,27 +2204,10 @@ get_partition_for_tuple(PartitionDispatch *pd,
ecxt->ecxt_scantuple = slot;
FormPartitionKeyDatum(parent, slot, estate, values, isnull);

- if (key->strategy == PARTITION_STRATEGY_RANGE)
- {
- /*
- * Since we cannot route tuples with NULL partition keys through a
- * range-partitioned table, simply return that no partition exists
- */
- for (i = 0; i < key->partnatts; i++)
- {
- if (isnull[i])
- {
- *failed_at = parent;
- *failed_slot = slot;
- result = -1;
- goto error_exit;
- }
- }
- }

Instead of getting rid of this. If there isn't a default partition then
we still do not have any range partition to route a null partition
key and the routing should fail.

I will work on a fix and send a patch shortly.

Regards,
Jeevan Ladhe