confusion about tiles not filling extent

[AndrewAnnex]

I am attempting to build a set of TMSs using transverse mercator projection for all of the zones, to start though I wanted to test things out with morecantile's UTM31WGS84Quad TMS,

But after some experimentation, I am seeing something unexpected when trying to enumerate all the tiles for the area of use for UTM31N, which is 0 to 6 longitude and 0 to 84 north latitude when using tms.tiles() where the tiles don't extend to the lower left corner of the expected extent in 4326. I am also comparing the results from supermorecado burn as I'll need the performance at higher zoom levels and I get a more truncated response in comparison to morecantile.shapes, even though I am setting truncate to false

here's a plot of this in 4326, at zoom level 9 and the red boxes corresponding to supermorecado, and the green from morecantiles' tiles function

And the same in EPSG:32631

What's interesting, is it could be related to the false easting, which is dead center of the projections only the 1 column of tiles to the left is shown.

I am a little confused about all of this, it should be the case that all of the area of use for the crs should be usable in the tms

[AndrewAnnex]

@vincentsarago I think I might have worked it out at least for supermorecado.

inside https://github.com/developmentseed/supermorecado/blob/7b2a68fce1c25921db81dd317ee4eb5ca574171e/supermorecado/burntiles.py#L135-L148

    def tile_extrema(
        self, bounds: Tuple[float, float, float, float], zoom: int
    ) -> Dict:
        """Tiles min/max at the given zoom for bounds."""
        ulTile = self.tms.tile(bounds[0], bounds[3], zoom)
        lrTile = self.tms.tile(bounds[2], bounds[1], zoom)

        minx, maxx = (min(ulTile.x, lrTile.x), max(ulTile.x, lrTile.x))
        miny, maxy = (min(ulTile.y, lrTile.y), max(ulTile.y, lrTile.y))

        return {
            "x": {"min": minx, "max": maxx + 1},
            "y": {"min": miny, "max": maxy + 1},
        }

you grab the min x tile coordinate from the upper left coordinate. I was able to correct things by grabbing the min x tile from the lower left tile.

For a bounds of 0-6 degrees longitude, 0-84 degrees latitude at zoom level 9, I get the following tiles for each corner:

print(tms.tile(0, 0, 9), tms.tile(6, 0, 9), tms.tile(0, 84, 9), tms.tile(6, 84, 9))

you get:

(Tile(x=123, y=255, z=9),
 Tile(x=132, y=255, z=9),
 Tile(x=127, y=136, z=9),
 Tile(x=128, y=136, z=9))

but tile_extrema returns:

{'x': {'min': 127, 'max': 133}, 'y': {'min': 136, 'max': 256}}

the xmin being 127 is the issue, as can be seen the actual x for tms.tile(0, 0, 9) is 123.

unless I am missing something here entirely this seems to be an issue.

I created a monkey patch for this and got the following

and also made a gist with the geojson files:

https://gist.github.com/AndrewAnnex/da4f4c353a6929dec22e8834c54fd881

As a side note, given my use case I may want to actually buffer the tile range to ensure no areas within the area of use polygon are not covered, like a adding a buffer to the tile extrema rather than buffering the geometry or maybe changing the behavior of truncate a bit.

another side point using xy_bounds resulted in better looking geometries near the pole, and I just reprojected after the fact to 4326 for the plots.

I hope that I am not just super confused about something with the intended usage/expected input coordinate systems for burn/tiles.