[postgis-devel] [PostGIS] #2122: [raster] Real extent feature lost after metadata as views

[Mateusz Loskot]

On 30 November 2012 15:36, Sandro Santilli <strk at keybit.net> wrote:
> On Fri, Nov 30, 2012 at 03:33:50PM +0000, Mateusz Loskot wrote:
>> On 30 November 2012 15:28, Sandro Santilli <strk at keybit.net> wrote:
>
>> > Still sounds like an higher level structure to me, not something
>> > to encode in each value. It's like topology-to-geometry.
>>
>> That's why there was real spatial extent preserved,
>> at the very high level of metadata.
>
> Yep, glad we're on the same tune about levels :)

Indeed :)

We've broke a feature.
We can't say "but we've never liked that, have we, so let's ditch it"
and back it up with argument, it's user's responsibility.

> Now let's see if those metadata can move from raster_table to
> table constraints ...

OK

Imagine, PostGIS geometry is 3D only storage. User loads 2D geometry, the loader
adds extra data, the Z. Then, SQL queries return 3D byte stream
and there is no functionality that enables 2D only queries.
That's close analogy I could think of ATM.
The point is, raster is not (always) 1:1 mapping of raster in file
or 1:1 equivalent to raster definition by GDAL.
The loader may change physical raster data if requested parameters
of storage require such change. That is an important flexibility of PostGIS.

Best regards,
--
Mateusz Loskot, http://mateusz.loskot.net