Hi @nperez
This fix is for issue #121 (closed). As the clarification of the problem in the comment in the issue, the actual problem is not about transform + chunk for two datasets, but simply for one dataset in some cases (that the previous unit tests happened not to test out.)
The current method of sri chunking may not be 100% correct still. The difficulty is to know which grid points are the one to pick after transformation. One thing can help is to limit the crop option to only the four numbers of region borders, not TRUE or FALSE. By this means, we can predict the values of sri more.
Anyways, this branch fixes the current bugs identified. We can merge it to master and do the crop development somewhere else. Please let me know if you agree to merge it, thanks.
Cheers,
An-Chi