Provide parameter for retrieving additional data around tile edges to minimise edge effects

[rowanwins]

As proposed in #56 this PR aims to reduce the edge effects associated with tiles by introducing a tile_edges_padding parameter. Additional data is retrieved around the tiles edges during the tile generation process which helps to reduce sharp edges that occur due to resampling.

@vincentsarago did suggest

and then overwrite it to 0 if we don't need resampling (if the internal tiles match the requested bounds)

I'm not really sure how we'd know about the internal tiles so happy to take any advice on this one :)

Let me know if you need anything else on this.

[vincentsarago]

thanks @rowanwins this is great. I need to think about how to check if the tile is aligned then also check the performances (which should not vary).

[vincentsarago]

@rowanwins maybe we can do something like

def is_aligned(src_dst, bounds, tilesize):
    """Check if tile is aligned with internal tiles."""
    if src_dst.crs != CRS.from_epsg(3857):
        return False

    col_off, row_off, w, h = windows.from_bounds(
        *bounds, height=tilesize, transform=src_dst.transform, width=tilesize
    ).flatten()

    if round(w) % 64 and round(h) % 64:
        return False

    if (src_dst.width - round(col_off)) % 64:
        return False

    if (src_dst.height - round(row_off)) % 64:
        return False

    return True

The idea is to check:

• if the CRS is EPSG:3857
• if the size of the window is a multiple of 64 px (minimum size of internal tile)
• if the offset is a multiple of 64px

why 64px ?

ideally we want to check if the window is a multiple of the internal tile size (raw data and overview) but we don't have a way to do it with rasterio and I'm not sure this is needed. By checking if we have a multiple of 64 we make sure that even on border tiles we don't need to do resampling.

Note: this might be totally wrong, but did a quick check with a web-optimized file and it work

[rowanwins]

Hi @vincentsarago

I've made those changes.
I think it would be good to add a test around the _requested_tile_aligned_with_internal_tile however I wasn't sure if any of the test fixtures had those internal web-optimised tiles, if you can point me in the right direction there that would be great.

[vincentsarago]

@rowanwins you can use this file https://www.dropbox.com/s/op0szbw18szqupc/web.tif?dl=0

[vincentsarago]

@rowanwins The PR looks excellent (I like the tests you added).
Couples things before we merge this:

• Can you change the base to maint-1.0 branch
• Make sure to follow https://github.com/cogeotiff/rio-tiler#contribution--development and have pre-commit install (this will make sure black ran and validate the format or CI will fail)
• add a note in the changelog

Thanks

[rowanwins]

Hi @vincentsarago

I attempted to change the base but I'm not sure if that worked TBH (I ran git rebase --onto maint-1.0 followed by git push -force origin pad-tile), my git skills are average :)

I think everything else is ticked off other than codecov complaining about a decrease in coverage. It would be good to test that _requested_tile_aligned_with_internal_tile function a little more, I'm just not entirely sure how...

[vincentsarago]

this is great @rowanwins I'll take it from here. thanks for your help

[rowanwins]

Out of interest how did you change the base branch @vincentsarago ?

[vincentsarago]

@rowanwins In fact I had the option in the Github UI 😬 (I didn't know I could do that before)