You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Some files are large but user may not really care how to read them, they simply need to be chunked.
Some instances of RasterSource may have a tilting scheme in the underlying data format.
GeoTiffRasterSource the tile scheme is the segment layout of the GeoTiff segments. GDAL similarly exposes "native" segment layout (reference needed) GeoTrellisRasterSource has the LayoutDefinition in which the underlying GeoTrellis layer is tiled
This should be optional property, as it's mostly an optimization. When chunking does not matter, using underlying chunking will reduce merging and lookup costs, reducing both memory and IO requirements for a read.
Also we should consider how this information will be used to actually read the chunks to make sure that the Option[TileLayout] is the most useful signature for this method. Other option might be LayoutDefinition.
The text was updated successfully, but these errors were encountered:
This is open question. @metasim had API similar to this but seems to have backed out of the idea. I will say that I have not yet keenly felt its absence in current API.
@echeipesh I reinstated something on top of VLM-RasterSource, but isn't truly "native" (you set the tile size). After the experiment with the prioir "native layout" approach, I'm not sure how much exactly aligning with the internal gridding helps, and when you add the need for BufferedTiles, the work seems too much of a hassle. I think providing a means of getting the RasterSource as a gridded set of tiles of arbitrarily regular size (as in the link above) is sufficient.
Some files are large but user may not really care how to read them, they simply need to be chunked.
Some instances of
RasterSource
may have a tilting scheme in the underlying data format.GeoTiffRasterSource
the tile scheme is the segment layout of the GeoTiff segments.GDAL
similarly exposes "native" segment layout (reference needed)GeoTrellisRasterSource
has theLayoutDefinition
in which the underlying GeoTrellis layer is tiledThis should be optional property, as it's mostly an optimization. When chunking does not matter, using underlying chunking will reduce merging and lookup costs, reducing both memory and IO requirements for a read.
Also we should consider how this information will be used to actually read the chunks to make sure that the
Option[TileLayout]
is the most useful signature for this method. Other option might beLayoutDefinition
.The text was updated successfully, but these errors were encountered: