999999999^9999999^999^99^999^99999 returns an error

[Tpt]

Not sure of what we should do on such big inputs

[marc-chevalier]

Not my fault! Sympy tries to do the computation since it knows it will
succeed but it is killed by timeout of the subprocess. There is not way
to predict if a computation will be long. If you disagree, think about
the halting problem. So, we cannot do anything. I think the best is just
to consider this input as "I cannot handle it", just as when the input
is not a mathematical formula and to return... nothing.

Le 02/02/2015 17:20, Thomas Tanon a écrit :

Not sure of what we should do on such big inputs

—
Reply to this email directly or view it on GitHub
#17.

Marc Chevalier

ENS de Lyon
Site Monod
M1 Informatique Fondamentale

[Tpt]

Yes, I agree that it is difficult. I've opened this issue thinking more "we should track this issue" than "we should solve this issue now".

[marc-chevalier]

I think there is nothing to do. Or add a error message saying "Oups! Too
long computation, sorry. Do it yourself or buy Mathematica" But the
doesn't change really anything.

[eddycaron]

True. You should have at least a timeout to decide when you cancel the computation. Just one way to avoid malicious guys who try to stress platypus for bad reason.

---

Eddy Caron. Mcf HDR ENS Lyon
ENS Lyon - LIP - Projet AVALON
E-Mail : [email protected]

[http://graal.ens-lyon.fr/~ecaron]

Le 2 févr. 2015 à 18:01, Marc Chevalier [email protected] a écrit :

I think there is nothing to do. Or add a error message saying "Oups! Too
long computation, sorry. Do it yourself or buy Mathematica" But the
doesn't change really anything.
—
Reply to this email directly or view it on GitHub.

[marc-chevalier]

We have a timeout for a long time, almost before the first deployement.
https://github.com/ProjetPP/PPP-CAS/blob/master/ppp_cas/evaluator.py#L25

Le 02/02/2015 18:08, eddycaron a écrit :

True. You should have at least a timeout to decide when you cancel the
computation. Just one way to avoid malicious guys who try to stress
platypus for bad reason.

---

Eddy Caron. Mcf HDR ENS Lyon
ENS Lyon - LIP - Projet AVALON
E-Mail : [email protected]

[http://graal.ens-lyon.fr/~ecaron]

Le 2 févr. 2015 à 18:01, Marc Chevalier [email protected] a
écrit :

I think there is nothing to do. Or add a error message saying "Oups!
Too
long computation, sorry. Do it yourself or buy Mathematica" But the
doesn't change really anything.
—
Reply to this email directly or view it on GitHub.

—
Reply to this email directly or view it on GitHub
#17 (comment).

Marc Chevalier

ENS de Lyon
Site Monod
M1 Informatique Fondamentale

[marc-chevalier]

Moreover, I'm pretty sure the timeout works because, since we use
platypus-angel which is less powerful as the previous server, the
request "Integrate[Sin[x*y], {x, 0, 1}, {y, 0, x}]" (one of my longer
test) caused an error until we increase the timeout. Since, that works
again correctly.