I am using Eigen3 Arrays to store Nx2 complex numbers to represent 2 complex series of datapoints vs time. I need to perform some operations (similar to norms, for example) on these arrays (which I can do using a DenseBase plugin), but I also need to store some meta-data information about the arrays (eg they are a time-series so I need to store the time step between datapoints).

To do this I create an Eigen Array and store it in a shared pointer. I then pass this as an argument in the constructor of a class that holds the meta-data as well as the shared pointer. In this way I can work on the Eigen Array directly and avoid needing to provide all the Array accessors, and I can work on the meta-data+array object as well, depending on need. But it's not a great solution as it removes a level of data hiding (I avoid major problems with this by ensuring that the majority of the use of the Array/Meta-data class is via read only const, but it's not ideal).

I am wondering if there is an alternative, such as inheriting from Array, but I notice in http://eigen.tuxfamily.org/dox-devel/To ... Eigen.html that this is really not recommended. Can I ask why this is so? Are there any suggestions for how to implement the above in an improved manner?

Thanks.

inheriting Array or Matrix is fine as long as you know what you are doing. For instance, is Foo inherit Array, then adding to Foo objects and evaluating them ( (foo0+foo1).eval() ) wont give you a Foo but an Array. That said, in your case, I'm not sure there is any benefit of inheriting Array, instead of having a custom class storing an Array and your additional members but well go as you prefer.

Thanks. I think it would be valuable to provide a bit more detail in the docs about the implications of inheriting from Matrix. Your description was enough to put me off!