Horton

Re: Cache lookup errors with functions manipulation object addresses

Michael Paquier <michael@paquier.xyz>

View thread View on postgr.es
From: Michael Paquier <michael@paquier.xyz>
To: Alvaro Herrera <alvherre@2ndquadrant.com>
Cc: Andrew Dunstan <andrew.dunstan@2ndquadrant.com>, pgsql-hackers@lists.postgresql.org
Date: 2018-12-25T04:34:38Z
Lists: pgsql-hackers

Attachments

Álvaro, all,

On Fri, Dec 14, 2018 at 09:04:36AM +0900, Michael Paquier wrote:
> Thanks again for looking up at what was proposed.  I'll see if I can
> finish the refactoring part for the next CF, and be done with this
> stuff.

Please find attached an updated patch set which reworks the SQL
interface for object addresses so as NULL-ness is handled consistently
for all object types.  This has needed a couple of extensions to some
current object-lookup APIs, which are added as refactoring bits
without breaking existing callers:
- Procedures and operators gain a new _extended routine which can take
a set of flags to enforce a NULL result or force qualification.
- format_type_extended needs an extra flag to enforce NULL.

On top of the refactoring issues, here is the list of changes to make
things work:
- The object type can never be NULL, meaning that for undefined
relations, routines or constraints the caller gets a generic name even
if the object is not defined.  (That's the definition mentioned a
couple of messages upthread).
- Large object existence is checked with LargeObjectExists().
- Table columns show all the fields as NULL if the object identity
cannot be found instead of showing the table and its schema.

With that all the regression tests show a consistent behavior for all
object types when those are undefined.  Thoughts?
--
Michael

Commits

  1. Eliminate cache lookup errors in SQL functions for object addresses

  2. Refactor routines for name lookups of procedures and operators

  3. Add new flag to format_type_extended() to get NULL for undefined type

  4. Introduce new extended routines for FDW and foreign server lookups

  5. Refactor routines for subscription and publication lookups

  6. get_relid_attribute_name is dead, long live get_attname

  7. Tweak parser so that there is a defined representation for datatypes