Re: Packages: Again

2017-01-13 19:35 GMT+01:00 Serge Rielau <serge(at)rielau(dot)com>:

>
> > On Jan 13, 2017, at 10:23 AM, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>
> wrote:
> >
> > I have not clean feeling from this - I am pretty sure so I am afraid
> schizophrenic between MODULES, SCHEMAS. Nested schemas increase complexity
> of searching complexity and breaks a logic database.schema.object
> Yes my proposal to nest schemata is “radical” and this community is not
> falling into that camp.
> But there is nothing holy about database.schema.object.attribute
>

sure not - but lot of logic in SQL parser and PLpgSQL parser is based on it.

> .
> >
> > Currently almost all in PostgreSQL PL design is primitive, but that
> means pretty simple too.
> We are having > 30,000 functions with a total of millions of lines of code.
>

I understand so your working life is pretty hard :).

>
> You are describing a self fulfilling prophecy here.
> As long as the community codebase only caters to its existing users you
> will not see a change in the usage pattern of the base.
>
> > It is hard to see a advantages of this proposal.
>
> At least I tried :-)
>

I would be careful to translate a Oracle concept to PostgreSQL - The
packages is Ada language concept - and there has clean role. In PL/SQL
Oracle used packages like we used schema because Oracle has different
concept of terms "database", of term "schema". The packages in Postgres is
more redundant than in Oracle. More the packages and related features are
probably most difficult PL/SQL feature. Lot of people don't understand well
- and it is not surprise for me, because PL/SQL is really hard mix of two
very different worlds.

I agree so there is some gap - there is nothing like package variables,
package constants. Can be nice to fill it.

With Postgres we should to think much more about other PL - there is not
only PL/pgSQL. So any what we create should be available for any PL. Our
PLpgSQL is based on total different technology design - so some benefits of
sharing compiled code across databases has not too value in Postgres.

Maybe I am starting be old :) - I don't believe so stronger tools helps do
things better - like Java - some applications are pretty good, some are the
big heap of shit - and due strong language this heap is sometimes pretty
big :)

If you need nested schemas, maybe you do some scary things, that should be
better solved in application server.

So I am not 100% against, but really I am sceptic if it is good idea. We
don't design Postgres as platform for migration legacy Oracle code - on
second hand we should not to create artificial breaks against this
migrations.

Regards

Pavel

> Serge
>
>
>
>