Обсуждение: Clarifying the ImportForeignSchema API

Folks,

I noticed that the API document for IMPORT FOREIGN SCHEMA states in
part:

    It should return a list of C strings, each of which must contain a
    CREATE FOREIGN TABLE command. These strings will be parsed and
    executed by the core server.

A reasonable reading of the above is that it disallows statements
other than CREATE FOREIGN TABLE, which seems overly restrictive for no
reason I can discern.  The list of C strings seems reasonable as a
requirement, but I think it would be better to rephrase this along the
lines of:

    It should return a list of C strings, each of which must contain a
    DDL command, for example CREATE FOREIGN TABLE. These strings will
    be parsed and executed by the core server in order to create the
    objects in the schema.

as a foreign schema might need types (the case I ran across) or other
database objects like CREATE EXTERNAL ROUTINE, when we dust off the
implementation of that, to support it.

I was unable to discern from my draft version of the spec whether
statements other than CREATE FOREIGN TABLE are specifically
disallowed, or whether it is intended to (be able to) contain CREATE
ROUTINE MAPPING statements.

Best,
David.
-- 
David Fetter <david(at)fetter(dot)org> http://fetter.org/
Phone: +1 415 235 3778

Remember to vote!
Consider donating to Postgres: http://www.postgresql.org/about/donate

2020年8月4日(火) 12:08 David Fetter <david@fetter.org>:
>
> Folks,
>
> I noticed that the API document for IMPORT FOREIGN SCHEMA states in
> part:
>
>     It should return a list of C strings, each of which must contain a
>     CREATE FOREIGN TABLE command. These strings will be parsed and
>     executed by the core server.
>
> A reasonable reading of the above is that it disallows statements
> other than CREATE FOREIGN TABLE, which seems overly restrictive for no
> reason I can discern.  The list of C strings seems reasonable as a
> requirement, but I think it would be better to rephrase this along the
> lines of:
>
>     It should return a list of C strings, each of which must contain a
>     DDL command, for example CREATE FOREIGN TABLE. These strings will
>     be parsed and executed by the core server in order to create the
>     objects in the schema.
>
> as a foreign schema might need types (the case I ran across) or other
> database objects like CREATE EXTERNAL ROUTINE, when we dust off the
> implementation of that, to support it.

+1

A while back I was considering using IMPORT FOREIGN SCHEMA to import
object comments (which IMHO can be considered part of the schema) and was
puzzled by the above. I never pursued that further due to lack of
time/priorities;
IIRC technically it wouldn't have been an issue regardless of what the spec
may or may not say (I couldn't find anything at the time).

Regards

Ian Barwick

--
Ian Barwick                   https://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services