This is already well tracked over a number of issues (#38, #12, #954), but filing as a replication case as per community slack messages @TomNicholas
The problem
As described in 38:
For example if you concatenate two arrays that each have uniform chunks except for the final chunk, you could end up with an array that has variable-length chunks in the middle somewhere
Using virtual-tiff I'm coming across the same issue in many community tiffs, especially those that are released on degree grids e.g. 10x10 degree images. That means virtual-tiff builds a manifest for a GeoTIFF whose size is not an integer multiple of its internal block size. The final row/column form a smaller "dangle". The single array opens correctly, but doesn't allow for concatenation across many tiles, even when they share the same projection and grid. A stack of these tiles gives a variable chunk length.
The current solution
- Opening a single tiff is of course fine
- Building a datatree of each tile, then building a filter + composite after the fact
Reproducing datasets
JRC Tropical Moist Forests (TMF):
Forest Typology 2020:
This is already well tracked over a number of issues (#38, #12, #954), but filing as a replication case as per community slack messages @TomNicholas
The problem
As described in 38:
Using
virtual-tiffI'm coming across the same issue in many community tiffs, especially those that are released on degree grids e.g. 10x10 degree images. That meansvirtual-tiffbuilds a manifest for a GeoTIFF whose size is not an integer multiple of its internal block size. The final row/column form a smaller "dangle". The single array opens correctly, but doesn't allow for concatenation across many tiles, even when they share the same projection and grid. A stack of these tiles gives a variable chunk length.The current solution
Reproducing datasets
JRC Tropical Moist Forests (TMF):
Forest Typology 2020: