Unsafe regions

In order to avoid inconsistency issues, TensorNetwork checks the index sizes are correct whenever a Tensor is push!ed and it already contains some of the its indices. There are cases in which you may want to temporarily avoid index size checks (for performance or for ergonomy) on push! to a TensorNetwork. But mutating a TensorNetwork without checks is dangerous, as it can leave it in a inconsistent state which would lead to hard to trace errors.

Instead, we developed the @unsafe_region macro. The first argument is the AbstractTensorNetwork you want to disable the checks for, and the second argument is the code where you modify the AbstractTensorNetwork without checks.

@unsafe_region tn begin
    ...
end

When the scope of the @unsafe_region ends, it will automatically run a full check on tn to assert that the final state of the AbstractTensorNetwork is consistent.

Note that this only affects disables the checks for one AbstractTensorNetwork, but multiple @unsafe_regions can be nested.

DocumenterMermaid.MermaidScriptBlock([...])