Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The rules for broadcasting in
2.4.x
seem to be that singleton dimension can be stripped from either side. If both is possible it'll strip from the back, i.e. the dataset with shape(1, 3, 1)
can be read into an array of shape(1, 3)
. This was probably motivated bystd::vector<std::vector<double>>
of shape(1, 3)
is easier to deal with than one that's(3, 1)
.However, the stripping from both sides only works for one-dimensional arrays. Stripping multi-dimensional arrays from either side is unstable and should be avoided anyway. Hence, we only allow stripping any
1
s if the target shape is one-dimensional.This commit achieves the above by:
introducing
details::squeezeDimensions
which strips the required amount of singleton dimesions.improving
details::checkDimensions
by removing bugs related to(3, 1, 1)
not being a valid 2D array.Adds (back) tests for checking the broadcasting rules. As well as
details::checkDimensions
anddetails::squeezeDimensions
.