All it basically does is replace the dimensions with the requested dimensions in the internal variable header. There wouldn't be much to see even if you could see the source code for that part. Most of the code is probably for the checking, e.g.,
- Are the requested number of elements the same?
- Are the requested dimension values all non-negative integers?
- If there is a missing dimension, calculate this value and make sure it is an integer.
- Can the result be a shared data copy? (e.g., for a full matrix)
- Will the result need to be a deep data copy? (e.g., for a sparse matrix)
- Etc.
