in reply to Re^2: Reverse Polish Notation Question
in thread RPN Question

Thank you hdb for your comments.

As for the first comment, we have a first degree equation, so, yes, I assumed a straight line on that specific condition, although I was more broadly trying to implement a more general solution.

For your second comment, yes, you are absolutely right, calculating the slope would far easier, especially for a linear equation, and I would not even need to painfully consider whether the various calculated values are positive or negative. And even for a polynomial equation, say second or third degree, calculating the slope would narrow down faster on the solution (something similar to the Newton-Raphson method).

My error was to initially think it could be done in 15 minutes and 20 lines of code, so I started to code something without carefully thinking of the problem, and discovered only when doing it that it was more complicated than I thought. If I have time this weekend, I'll try to come up with a better design.

Je suis Charlie.

Replies are listed 'Best First'.
Re^4: Reverse Polish Notation Question
by hdb (Monsignor) on May 02, 2015 at 14:55 UTC

    For a linear equation it could look like this (and yes, it took me more than 15 minutes...) 

    use strict;
use warnings;

my $equation = "(2*(4*x-8)+3*x)-(10*x+1-9)"; # " = 0" omitted
$equation =~ s/x/\$_[0]/g;
my $eval = sub { eval $equation };
my $val1 = $eval->(0);
my $val2 = $eval->(1);
die "No solution for this equation\n" unless $val1 - $val2;
my $result = -$val1 / ($val2 - $val1);

printf "Result is %.10f\n", $result;

    [download]
[reply]
[d/l]
      I tried to time your version and mine (removing the printouts). As I would expect yours is faster, but mine is converging fast enough to make the difference very small:

      Your solution: 

      $ time perl -e 'use strict;
> use warnings;
>
my $val2 > = $eval->(1);
my $equation = "(2*(4*x-8)+3*x)-(10*x+1-9)"; # " = 0" omitted
> $equation =~ s/x/\$_[0]/g;
> my $eval = sub { eval $equation };
> my $val1 = $eval->(0);
> my $val2 = $eval->(1);
> die "No solution for this equation\n" unless $val1 - $val2;
> my $result = -$val1 / ($val2 - $val1);
>
> printf "Result is %.10f\n", $result;'
Result is 8.0000000000

real    0m0.044s
user    0m0.015s
sys     0m0.031s

      [download]
      and mine: 
      $ time perl solver.pl
Result is: 8.0000000000

real    0m0.050s
user    0m0.031s
sys     0m0.015s

      [download]
      0.44 versus 0.50 sec. I might have been lucky on this one, but, frankly, my narrowing-down algorithm works fairly well.

      Je suis Charlie.
[reply]
[d/l]
[select]
      Hi hdb, that's a pretty neat solution.

      It took me twenty seconds or so to figure out why you were replacing x by $_[0]. I must say that this is a nice clever idea.

      Je suis Charlie.
[reply]
[d/l]

In Section Seekers of Perl Wisdom