-6. Can we make the minimal and characteristic polynomial tests work
- for trivial algebras, too? Then we wouldn't need the "nontrivial"
- argument to random_eja(). This would also allow us to potentially
- pass a "size" of zero to the other constructors in random_eja() --
- a code path that is basically untested right now.
+5. Profile the construction of "large" matrix algebras (like the
+ 15-dimensional QuaternionHermitianAlgebra(3)) to find out why
+ they're so slow.
+
+6. Instead of storing a basis multiplication matrix, just make
+ product_on_basis() a cached method and manually cache its
+ entries. The cython cached method lookup should be faster than a
+ python-based matrix lookup anyway.