Forum OpenACS Development: Re: package_id for acs_objects

Collapse
5: Re: package_id for acs_objects (response to 4)
Posted by Guan Yang on 02/27/04 09:54 PM
As for package_id, how about this simple rule: If in doubt, objects belong to the package that contains the SQL file where acs_object_type__create_type is called.
Collapse
6: Re: package_id for acs_objects (response to 5)
Posted by Timo Hentschel on 02/29/04 08:16 AM
Well, the thing is, for objects created during install there is no package_id yet, is there? so, should i assign null then?
Collapse
7: Re: package_id for acs_objects (response to 6)
Posted by Dirk Gomez on 02/29/04 10:03 AM
What would be the other options? Claim that they "inherit" from "Main Site" maybe?
Collapse
8: Re: package_id for acs_objects (response to 7)
Posted by Timo Hentschel on 03/01/04 05:37 PM
How should they inherit? and when? as i said, during install there is no package_id available - not even of "Main Site".
Collapse
9: Re: package_id for acs_objects (response to 8)
Posted by Mark Aufflick on 03/02/04 12:05 PM
this is a bogus idea, but here goes:

  • you make a temp table:

    ___________________________________
| fake_pkg_id |    package_key    |
+-------------+-------------------+
|    -1000    |    acs-subsite    |
|    -1001    |   acs-templating  |
|     ...     |        ...        |
  • you use the fake package id's to populate the acs_objects table
  • as soon as package_id's exist, you reverse lookup the real package_id's using the package_key and replace the fake id's in acs_objects.
of course there is a limitation for packages mounted more than once during install, but if that (unlikely) event occurs, you could just use the first one and still be in a sensible position. it may be possible to even correctly deal with that situation, but i'm not too familiar on how package_id's are created during bootstrap.