Proper use of Solve App

[ebswift]

I'm at a point where I'm horribly confused with the solve app. I can get it to work for different things, but there is one scenario where I'm finding it tough to get right - a simple solve for 2 unknowns. I tried a method from this thread, but get a cannot solve message:

http://www.hpmuseum.org/cgi-sys/cgiwrap/...ead=258059

Export attempt from above:
   

symbolic view:
   

Cannot solve:
   

Now, the REALLY confusing part is that I got it working using X & Y variables. BUT, after playing with the l & w variables for a while I can't even get X & Y to solve anymore. I also mucked about using STO to no avail.

Please help with the solve workflow.

[Tim Wessman]

Are you up to date?

[ebswift]

Hi Tim, I believe so, same version on virtual and real calc:

   

[rgallier]

I have found that clearing the variables will resolve this behavior (sometimes). Also delete the variables and re-export will work (sometimes). I'm very confused with the inconsistent behaviour.

[mandresve]

(08-09-2015 01:27 AM)rgallier Wrote:  I have found that clearing the variables will resolve this behavior (sometimes). Also delete the variables and re-export will work (sometimes). I'm very confused with the inconsistent behaviour.

Same here =/

[Tim Wessman]

Do you see any problems if you enter any value other then "0" for your two guesses?

It seems to me the CAS fsolve command (which is what is being called in the solve app behind the scenes) is doing something strange when you put two guesses of 0 in. You can see this by typing this in the CAS screen.

fsolve([l = (w+10),(l*w) = 1200],[l,w])

Note that if you change that to a, b;

fsolve([a = (b+10),(a*b) = 1200],[a,b])

It works fine, and spits out the two results. If you store a float into the vars though of 0., using a:=0. b:=0. , you will start to see the misbehavior.

Store b:=.1, and suddenly it starts working again.


So I'd say put any number other then 0 in for a guess and it most likely will work. I think this might be something to go on Bernard's "list"...

[ebswift]

Yep, that's it, thanks Tim. Simply entering X: 1 Y: 1 in the guesses solves my example. It also works as expected with L and W.

When I originally solved it, I must have had values in there somehow.

[Tim Wessman]

Looks like Bernard actually fixed this a while back in the cas (after the last public release). It returns the vector of a result instead of the single number like I'd reported prior.

So when/if there is another release it should be resolved I think.