Pixel grid dimensions (40818x1) error fixed.

[dabhicusp]

Fixed: #85.

The current code gives an error(Pixel grid dimensions (40818x1) must be less than or equal to 32768) when creating an lat, lon array for the image_to_array conversion, for certain lengthy images. To address this, I've made updates to the code. Now, when generating an lat, lon array for certain lengthy images, I break it down into smaller array based on Users preferred size or default size. This adjustment prevents the above error.

[dabhicusp]

Hello @alxmrs I tried with just making the requests in a loop. but it takes much more time(~21x) as compare to threadpool for the calculating in the image which user provided.

Code:

    tiles_threadpool = [None for _ in range(total_tile)]
    tiles_forloop = [None for _ in range(total_tile)]

    start_time = time.time()
    with concurrent.futures.ThreadPoolExecutor() as pool:
      for i, arr in pool.map(
          self._get_tile_from_EE, list(zip(data, cycle([coordinate_type])))
      ):
        tiles_threadpool[i] = (
            arr.tolist() if coordinate_type == 'longitude' else arr.tolist()[0]
        )
    print("threadpool_time: ", coordinate_type,time.time() - start_time)

    start_time = time.time()
    for i in range(total_tile):
      _, res = self._get_tile_from_EE((data[i], coordinate_type))
      tiles_forloop[i] = (res.tolist() if coordinate_type == 'longitude' else res.tolist()[0])
    print("for_loop_time : ", coordinate_type, time.time() - start_time)

Image which tested:

ee_class = EarthEngineBackendEntrypoint()
geom = ee.Geometry.Rectangle([-61, -7, -50, -18])
county = ee.Feature(geom)
image = ee.Image('projects/mapbiomas-workspace/TRANSVERSAIS/ZONACOSTEIRA6/mosaic_1985').clip(county)
coll = ee.ImageCollection(image)
ds = ee_class.open_dataset(coll.select(['NDVI']), crs='EPSG:3857',
                         scale=30, geometry=geom,
                         io_chunks={'time':24,'X': 512, 'Y': 512}
                         )

Result:

threadpool_time:  longitude 4.9227306842803955
for_loop_time :  longitude 95.86925983428955
threadpool_time:  latitude 4.389440298080444
for_loop_time :  latitude 74.4054172039032

Also tested on other images too and got the same result as above(taken ~21X more times).