For the linear node, if it only has one input, and that input has a coefficient of 1, there is no need to use a linear invariant. This change should reduce the height of the propagation graph and make it easier to debug (which is exactly why I implemented it, I'm chasing down a bug in a different branch).
In the future it is probably a good idea to use the view if there is only one input regardless of the coefficient associated with it. However, that would require a new view as well, which would make the PR a lot bigger.
For the linear node, if it only has one input, and that input has a coefficient of 1, there is no need to use a linear invariant. This change should reduce the height of the propagation graph and make it easier to debug (which is exactly why I implemented it, I'm chasing down a bug in a different branch).
In the future it is probably a good idea to use the view if there is only one input regardless of the coefficient associated with it. However, that would require a new view as well, which would make the PR a lot bigger.