Add support to TiledDataset for missing, irregular or overlapping tiles #487
Conversation
CodSpeed Performance ReportMerging #487 will not alter performanceComparing Summary
Benchmarks breakdown
|
| def __init__(self, dataset_array, inventory=None): | ||
| self._data = np.array(dataset_array, dtype=object) | ||
| def __init__(self, dataset_array, inventory=None, mask=False): | ||
| self._data = np.ma.masked_array(dataset_array, dtype=object, mask=mask) |
There was a problem hiding this comment.
Can we auto-generate the mask based on the elements of dataset_array?
There was a problem hiding this comment.
I expect so. We can readily assume anything that isn't a Dataset should be masked out, but we should probably also have some way of flagging invalid Datasets for masking as well.
There was a problem hiding this comment.
Shouldn't the default value of mask be None?
There was a problem hiding this comment.
Either works fine. False is specifying to set the mask to be False (ie: not masked) everywhere. Which is the default if you pass None into masked_array anyway, but explicit is better than implicit.
Co-authored-by: Stuart Mumford <[email protected]>
…o tiledds-improvements
|
There are a few places where this can still be improved - if we're happy to make issues for those and punt them then this is good to go pending review. Otherwise I'll come back to it on Monday. |
Changes the array of datasets underlying TiledDataset into a masked array, defaulting to False everywhere (all tiles are valid).