[FEATURE]: Enable no resampling by raster tiler

Thank you for helping us improve solaris!

Summary of the bug

I ran the rio tiler tutorial example on a Landsat Analysis Ready Data Scene to tile it into 512x512 chips. But new negative values are added that weren’t there before in the original dataset (which only had -9999).

Steps to reproduce the bug

Please either paste sample code used to generate the buggy behavior below, or provide step-by-step instructions to reproduce the problem.

import solaris as sol
import os

raster_tiler = sol.tile.raster_tile.RasterTiler(dest_dir='test_solaris',  # the directory to save images to
                                                src_tile_size=(512, 512),  # the size of the output chips
                                                verbose=True,
                                               nodata=-9999)
raster_tiler.tile('/mnt/cropmaskperm/test-landsat/scene/LT05_L1TP_029031_20050316_20160912_01_T1_sr.tif')
bigarr = skio.imread('/mnt/cropmaskperm/test-landsat/scene/LT05_L1TP_029031_20050316_20160912_01_T1_sr.tif')
arr = skio.imread("test_solaris/LT05_L1TP_029031_20050316_20160912_01_T1_sr_696765_4625505.tif")
np.unique(arr[...,2])
np.unique(bigarr)
8578 in np.unique(bigarr)

array([-9999, -9998, -9997, ...,  8467,  8578, 19423], dtype=int16) #chip arr
array([-9999,  -898,  -897, ...,  8923,  8960, 20000], dtype=int16) # src arr
False # negative and positive values are changed, the src arr value should be in the chip arr

Steps to reproduce the behavior: Run the rio tiler on a Landsat scene and check if the chip array values are in the source scene. If not it means values have been mutated in the tiling step.

Expected behavior

The tiling step shouldn’t mutate the values

Environment information

• OS: ubuntu 16
• solaris version: 0.1.3
• python version: 3.6

[nrweir]

Hi @rbavery,

Great point on the EPSG code issue. Let me chat with the team here regarding defaulting to re-project or leaving it alone. This topic came up a lot as we were writing the initial implementation and we eventually settled on the solution that’s implemented, but given this context might be willing to switch.

[nrweir]

Resolved with #294!