| From: | "Christopher Kings-Lynne" <chriskl(at)familyhealth(dot)com(dot)au> | 
|---|---|
| To: | "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> | 
| Cc: | "Hackers" <pgsql-hackers(at)postgresql(dot)org> | 
| Subject: | Re: pg_dump in 7.4 | 
| Date: | 2002-11-14 03:33:16 | 
| Message-ID: | GNELIHDDFBOCMGBFGEFOGEIPCEAA.chriskl@familyhealth.com.au | 
| Views: | Whole Thread | Raw Message | Download mbox | Resend email | 
| Thread: | |
| Lists: | pgsql-hackers | 
> The thought that I'd been toying with is to build a list of inter-object
> dependencies (using pg_depend if available, else fall back on pg_dump's
> native wit, ie, the rather limited set of dependencies it already
> understands).  Then do a topological sort, preferring to maintain OID
> order in cases where the object ordering is underspecified.  When the
> sort fails (ie, there's a circular dependency) then modify the set of
> dumpable objects by breaking some object into two parts (a base
> declaration and an ALTER command); this changes the dependencies too.
> Repeat the sort and adjustment steps until the sort succeeds.
>
> If you're not familiar with topological sorts, look at Knuth or your
> favorite algorithms text.  There are one or two instances of the method
> in PG already (deadlock detection, for example).
C'mon - I _do_ have an honours degree in computer science ;)
I was actually trying to think of the correct term when i wrote my original
email.  I had 'PERT charts' in my head, but I couldn't remember the term for
the required order of activities - thanks for reminding me :)
Chris
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Bruce Momjian | 2002-11-14 03:56:41 | ISDN installed | 
| Previous Message | Christopher Kings-Lynne | 2002-11-14 03:30:30 | Re: RC1? |