Topology identifiers datatypes (was: Migrating TIGER from Oracle to POSTGIS)

[Sandro Santilli]

On Mon, Mar 17, 2025 at 11:52:20AM +0000, Ayorinde O Adesugba (CENSUS/DITD FED) wrote:

> Thanks for the response. I do not have the authority to hire or
> recommend a contracting company, but I am definitely open to contributing
> time and code to make the topology extension better for all users.

Great!

> If you can send any links/documentation of the process - I can start
> looking into it. I have made a few alterations to table definitions and
> the topology.topogeometry type, and some functions (creating a variant
> with int8 parameters but was restricted on AWS which is the target
> location for the final migration).

We plan changes via trac, so that's where you could drop a link to
your ongoing work, when you decide where to put it.

The most convenient way to work would be in a git branch associated
to a pull request on https://git.osgeo.org/gitea/postgis, so that
anyone with an OSGeo UserID can easily comment on it. You'll be able
to use the same credentials that you'd use on https://trac.osgeo.org.

Discussion can be done on this mailing list, I'm starting with this
message a new thread specific to that. Did you review the ticket ?
You'll read that upgrade paths need be provided, which is probably
the most sensible part of this work.

I've been thinking for some time now that we could extend the
topology schemas with a "metadata" table, which may help with
such upgrades. Changing size of primitive identifier datatype
in particular could result in crashing the backend if the code
does not consider the case in which the library is upgraded while
the database structure still isn't, this is something you should
keep in mind.

--strk;