[postgis-devel] Regular raster blocking (was: Real extent feature lost after metadata as views)

[Sandro Santilli]

On Thu, Dec 06, 2012 at 12:14:10PM +0000, Mateusz Loskot wrote:
> On 6 December 2012 11:45, Sandro Santilli <strk at keybit.net> wrote:

> > Why the distinction between internal and rightmost/bottom blocks ?
> 
> Your suggestion allows partially missing tiles in the middle of tiled coverage.
> List of some possible situations that need to be considered:
> Spatial extent for the tiled coverage is calculated, and it represents
> a real extent.
> Spatial extent of partial tile in the middle of coverage is smaller
> than extent of coverage grid cell,
> leaves gap in extent. Gap in extent is different to gap in data.

Extent never has gaps, it's always the minimum rectangle containing
all the tiles in the table.

I do can see that omitting or reducing tile sizes near the boundary
of the whole extent might effectively change that extent, but that's
the case also when only allowing it on the right or on the bottom.

If you want the extent to remain immutable there should be an explicit
extent constraint (only computed when missing).

> If partial tiles are allowed, one may ask why larger tiles shouldn't be and it
> raises question about relevance of the overlapping constraint.

Larger tiles shouldn't be allowed because they would cross block boundary.

  - Each tile belongs to exactly _one_ block
    (so no tile crosses block boundary)
  - Every block contains at most _one_ tile

--strk;