alpha values variable with interval colormaps

[gagecarto]

I am working to filter a raster between variable range intervals as chosen by the user. To start I've been doing this using colormaps like below where I just make the upper and lower ranges transparent
myColormap=[ [[0,0.21],[0,0,0,0]], [[0.21,0.75],[22,37,194,255]], [[0.75,1],[0,0,0,0]] ]

This works great and tiles are returned as transparent in those areas where values are within the lower and upper bounds. E.g. below

https://titiler.xyz/cog/tiles/WebMercatorQuad/10/200/383@1x?url=myfilelocation&bidx=1&colormap=[[[0,0],[0,0,0,0]],[[0,0.17],[12,208,32,255]],[[0.17,1],[0,0,0,0]]]

For some reason there is variability in the tiles that are returned and sometimes tiles are not transparent where they should be as defined in the color map. The call below returns a tile with no opacity but just black in those areas where the alpha value should be 0.

https://titiler.xyz/cog/tiles/WebMercatorQuad/10/202/377@1x?url=myfilelocation&bidx=1&maxzoom=14&colormap=[[[0,0.29],[0,0,0,0]],[[0.25,0.75],[22,37,194,255]],[[0.75,1],[0,0,0,0]]]

Any ideas why zoom would effect the alpha value on those tile requests?

[gagecarto]

I got this sorted out. I had my tile format set to automatic and titiler would sometimes send jpg. Adding &format=png cleared things up

https://titiler.xyz/cog/tiles/WebMercatorQuad/10/202/377@1x?url=myfilelocation&bidx=1&maxzoom=14&colormap=[[[0,0.29],[0,0,0,0]],[[0.25,0.75],[22,37,194,255]],[[0.75,1],[0,0,0,0]]]&format=png

[vincentsarago]

I'm not sure but this is mostly where it come from https://github.com/cogeotiff/rio-tiler/blob/main/rio_tiler/utils.py#L444-L449

would you be able to share the file? ( you can use https://github.com/cogeotiff/rio-faux to impersonate the file)

[gagecarto]

See the faux file below.
https://www.dropbox.com/s/k0c3r1wbdtm26yx/test.tif?dl=0

[vincentsarago]

@gagecarto well I can't really reproduce this. When using discrete colormap you are creating a new Alpha band which then should be taken into account when creating the PNG.

Sadly I the faux file is only nodata (bug/missing feature in rio-faux 🤦) so I tried with another file:

[gagecarto]

@vincentsarago - It's all good today. I think some of those other tiles were still cached. Forcing to png fixed it

[vincentsarago]

🥳