No div impl between numbers and polynomials

[sritchie]

The following code, in a namespace with the proper imports like sicmutils.simplify-test:

(sicmutils.simplify/hermetic-simplify-fixture
 (fn []
   (g/simplify
    (g/mul 0.5
           (g/div 1 (g/mul 2 'x))))))

Throws:

Execution error (Error) at (<cljs repl>:1).
No method in multimethod 'sicmutils.generic/div' for dispatch value: [#object[Number] :sicmutils.polynomial/polynomial]

Everything works if we use (/ 1 2) to generate a ratio. Do we want to handle a floating point number here?

Well, of course we do. I guess I would expect here that maybe the simplifier would do nothing, and return (/ 0.5 (* 2 x)).

[littleredcomputer]

That would be my expectation too. Does 0.5 register as exact now?

…

On Tue, Aug 25, 2020 at 1:56 PM Sam Ritchie ***@***.***> wrote: The following code, in a namespace with the proper imports like sicmutils.simplify-test: (sicmutils.simplify/hermetic-simplify-fixture (fn [] (g/simplify (g/mul 0.5 (g/div 1 (g/mul 2 'x)))))) Throws: Execution error (Error) at (<cljs repl>:1). No method in multimethod 'sicmutils.generic/div' for dispatch value: [#object[Number] :sicmutils.polynomial/polynomial] Everything works if we use (/ 1 2) to generate a ratio. Do we want to handle a floating point number here? Well, of course we do. I guess I would expect here that maybe the simplifier would do nothing, and return (/ 0.5 (* 2 x)). — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#93>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAB3Y7O4I4B7KGC52ZZS4UDSCQQQRANCNFSM4QLCQ2IA> .

[sritchie]

looks like it doesn't. I can try and dig in to where this multiplication is happening, too. possibly notable is that it times out, and only succeeds (with that error) when I wrap it with hermetic-simplify-fixture. More demons to chase!

[littleredcomputer]

If memory serves the 0.5 should have been converted to a symbol by the analyzer, so that it would form a rational function when divided by the polynomial in the denominator. Then the backsubstition pass would replace the symbol with 0.5 after we're done grinding through the simplification, so we definitely don't need an implementation of the div.

…

On Tue, Aug 25, 2020 at 3:04 PM Sam Ritchie ***@***.***> wrote: looks like it doesn't. I can try and dig in to where this multiplication is happening, too. possibly notable is that it times out, and only succeeds (with that error) when I wrap it with hermetic-simplify-fixture. More demons to chase! — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#93 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAB3Y7LHUWZBWYYN6FCZDJLSCQYNXANCNFSM4QLCQ2IA> .

[sritchie]

Okay, I fixed this by adding an explicit implementation of invert for polynomials in the rational-function namespace:

(defmethod g/invert [::p/polynomial] [^Polynomial p]
  (make (p/make-constant (.-arity p) 1) p))

[sritchie]

Adding this to #169 .