Auto-compute option could be controlled by output type:

• autocompute_scalar: auto-compute scalars (similar to what we had originally)
• autocompute_vector: auto-compute vector results
• autocompute_matrix: auto-compute matrix results

More possibilities:

• infix_scalar_args: infix operations with scalar arguments. v + 1, A > 0, etc are all very clear and intuitive. Well, mostly. Scalars can probably return Scalars, whereas Vector and Matrix objects will return expressions.
  • What about in-place? Would A += 1 turn A into an expression, or do we compute for in-place? Or, do we have another config for this?
• inplace: enable common accumulations to be spelled with e.g. A += expr.
  • What does A += B mean? If inplace means accumulation, then this implies outer join / ewise_mult behavior. Hmph.

In general, I don't like the idea of adding lots of configs to enable variations of spelling. I am in favor of exploring syntax that we may want to add. And, we may find that some syntax can be handy at times, so I'm open to the idea of adding configs. Fwiw, I think auto-compute is worth having a config for.