On Thu, 6 Jun 2019 at 18:03, Alvaro Herrera <alvherre@xxxxxxxxxxxxxxx> wrote:
On 2019-Jun-06, Alex V. wrote:
> I think that your position about primary keys in partitional tables is
> not right.
>
> If we see regular table, one-field primary key is cross-table unique.
> In partitional tables for users view we MUST also seen unique
> one-field primary key because this is user requirement and another
> keys can destroy logic between regular and partitional tables and
> functionality of partitioning becomes useless.
> For administrators of table we not-MAY, but MUST see one-field unique
> primary key in cross-table realisation.
> All another realizations are plugging holes in a hurry and non-logical
> at global review of engine.
If you are saying that you think that Postgres should support primary
keys that don't necessarily overlap partition keys, then I agree with
you. Please send a patch to implement that capability.
The point is that unique indexes that overlap partition keys are rather cheap because they can be made from a set of local indexes.
A unique index that doesn't overlap partition keys would be
* unfeasibly huge
* likely to hit the 32TB limit on relations
* unable to cope with dropping partitions
It sounds like you want it, cos its just a line of DDL, but in practice your colleagues wouldn't want those things.
