3
$\begingroup$

I think the best way to state my problem is to illustrate it with an example. Assume I have the following toy-list of rules:

toylist = {a -> 10^12 b + 2 c, d  -> 10^-11 e - 0.5 b, f -> 2 g}

What I am trying to find is a way of "reversing" the rules according to the coefficients present in the righthand side. Namely, if the coefficient of a certain variable in the righthand side is above a certain value, say 10^10, I would like to rewrite the rule so that the variable in question is on the lefthand side. The desired output would be:

toylist = {b -> 10^-12 a - 2*10^-12 c, d -> 10^-11 e - 0.5 b, f -> 2 g}

Note that I do not want anything to happen to the other rules in toylist, they're good as they are because all the coefficient are below the cutoff.

I am sure there's a clever way of doing this, I have tried in a lot of ways but failed miserably. Maybe using associations? I'm not so sure.

Improve this question
$\endgroup$
5
  • $\begingroup$ Is the right hand side of the rules always linear in the parameters ? $\endgroup$
    – b.gates.you.know.what
    Mar 12 '15 at 9:45
  • $\begingroup$ Is it true that only one coefficient is to be tested against the condition? Will that coefficient be specified or does it need to be detected? $\endgroup$
    – m_goldberg
    Mar 12 '15 at 9:57
  • $\begingroup$ @b.gatessucks Yes, it's a linear system $\endgroup$
    – user50473
    Mar 12 '15 at 10:22
  • $\begingroup$ @m_goldberg In principle, you could have something like a->10^12 b + 2 c + 3*10^15 d. Then you solve for the biggest one of the two above cutoff, in this case d. $\endgroup$
    – user50473
    Mar 12 '15 at 10:22
  • $\begingroup$ @Kuba give me a sec I want to understand what you're doing :) $\endgroup$
    – user50473
    Mar 12 '15 at 10:27
5
$\begingroup$

Brute force:

toylist /.  Rule[lhs_, rhs_] :> With[{
   coeff = Coefficient[rhs, Variables[rhs]]
   },
   Solve[lhs == rhs, #][[1, 1]] &[
     Last[Variables[rhs][[Ordering[coeff]]]]
     ] /; Max[coeff] > 10^5
   ]

{b->(a-2 c)/1000000000000,d->-0.5 b+e/100000000000,f->2 g}

You can reverse condition: /; Min[coeff] < 10^-5

{a->1000000000000 b+2 c,e->-1.*10^11 (-0.5 b-d),f->2 g}

That was automatic, if you want less general solution, such as one you've asked for:

rewrite[expr_, var_, limit_] := expr /. Rule[lhs_, rhs_] :> 
  Solve[lhs == rhs, var][[1, 1]] /; Coefficient[rhs, var] > limit

rewrite[toylist, b, 10^5]

{b->(a-2 c)/1000000000000,d->-0.5 b+e/100000000000,f->2 g}
Improve this answer
$\endgroup$
5
  • $\begingroup$ Your answer seems to work in this toy model but not with my list. Here's a sample of it. I suspect the problem has to do with the subscripts, unfortunately is that I can't get rid of them in my code. u = {Subscript[c, 15] -> - 1.08*^16 Subscript[c, 24] - 0.14 Subscript[c, 29], Subscript[c, 23] -> 5.83*^-15 Subscript[c, 24] - 0.84 Subscript[c, 29]} $\endgroup$
    – user50473
    Mar 12 '15 at 10:38
  • $\begingroup$ Actually it was just the minus sign! Just replaced coeff with Abs[coeff] in your code and it works like a charm. Thank you! $\endgroup$
    – user50473
    Mar 12 '15 at 10:43
  • $\begingroup$ @user50473 Yes it is, -10^15 isn't larger than 10^10. So you have to decide what do you want to check, add Abs to dondition if you want. $\endgroup$
    – Kuba ♦
    Mar 12 '15 at 10:44
  • $\begingroup$ @user50473 ok, you've got it, great ;) $\endgroup$
    – Kuba ♦
    Mar 12 '15 at 10:45
  • $\begingroup$ @user50473 Please consider holding on with an accept, better answers may appear, let's do not discourage others :) $\endgroup$
    – Kuba ♦
    Mar 12 '15 at 10:47
2
$\begingroup$ 
rewriteF[e_, v_, cf_] := If[cf[Coefficient[#[[2]], v]], 
                            Solve[Equal @@ #, v][[1, 1]], #] & /@ e

rewriteF[toylist, b, # > 10^5 &]
(* {b -> (a - 2 c)/1000000000000, d -> -0.5` b + e/100000000000, f -> 2 g} *)

rewriteF[toylist, g, Positive]
(* {a -> 1000000000000 b + 2 c, d -> -0.5` b + e/100000000000, g -> f/2} *)

Fold[rewriteF[#, #2, Positive] &, toylist, {g, b}]
(* {b -> (a - 2 c)/1000000000000, d -> -0.5` b + e/100000000000, g -> f/2} *)
Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.