a random error in AIA function

[imendez]

A very curious error.

When compiling this code:

Set Outliers = SetOfRation(PulseOut);
TimeSet FechOutput = WD(6)+WD(7);
Set Include("M:/personas/Chakib/serie_aia.bdt");
Polyn MA = 0.59627;
Polyn ARI = (1-B^{7);
Ration PiW = ARI/MA;
Set AIAInputs.Pre = AIA(res, PiW, Outliers);
Real NumCiclos = 100000;
Set CicloAia = For(1, NumCiclos, Set(Real n)
{}

If(EQ(n%100, 0), WriteLn(". "+ FormatReal(n, "%.0lf") +"/"+ FormatReal

(NumCiclos, "%.0lf") +"\n"),

If(EQ(n%10, 0), Write(".")));

Set seriesOutlier01 = For (1, Card(AIAInputs.Pre), Serie(Real i)
{

Text nombre = Identify(AIAInputs.Pre[i]);
Real inicio = TextFind(nombre, "RationExpand");
Real final = TextLength(nombre)-1;
Text nombresub = Sub(nombre, inicio, final);
Text nombrefinal = Replace(nombresub, ",", ", ");
Serie Eval(nombrefinal)

});
seriesOutlier01

});

Under Tol version from september 20th, Tol returns an aleatory error proximately
one of each 800 times (personal an frequentist estimate, not bayesian neither
critical). Curiously, under Tol version from september 29th, the frequency of the
error is reduced to only 2 or 3 per thousand.

I don't know why the error appears and I don't know the difference between these
two Tol versions.

[danirus]

Hello Iván,
Please, could you attach here the BDT file?
Thanks in advance.

[imendez]

serie de residuos

[imendez]

Is it working out, Dani?

I change the priority and severity of this bug becuase one of each 100 estimates
fails due to this bug.

[manuelb]

After a few updates and some other bugs resolved, the beast seems more stable,
and this bug just got out of here...

Tested on 1.1.2, no errors.

[manuelb]

Iván just gave us this code that re-opens this problem:

/
Text PutDumpFile("prueba.log");
Real NumCiclos = 100;
WriteLn("Inicio: "+ Time);
Real Cont = Copy(0);
Real While(LT(Cont, NumCiclos),
{

Real (Cont := Cont+1);
If(EQ(Cont%100, 0),

WriteLn(". "+ FormatReal(Cont, "%.0lf") +"/"+

FormatReal(NumCiclos, "%.0lf") +"\n"),

If(EQ(Cont%10, 0), Write(".")));

Text nombrefinal = "RationExpand(y2000m11d25, Diario, PulseOut)";
Serie sr = Eval(nombrefinal);
Copy(TRUE)

});
WriteLn("Fin: "+ Time);
/

Ivan tells us on his mail that the code seems to fail only when NOT giving the ";" at
the end of the Eval sentence nombrefinal.

numciclos counts the iteration number, for 100 seems ok, but 10,000 starts to give
us a headache!

I am on it!

[danirus]

This bug has just been resolved.
It was a problem related with BFilter module of Tol Parser.
A bad analisis of a pointer to char brang Bfilter module to turn a little bit crazy.

Solution will be available with next release of Tol.
In other way, you can always get a copy of Tol from our CVS.

[danirus]

This bug has just been resolved.
It was a problem related with BFilter module of Tol Parser.
A bad analisis of a pointer to char brang Bfilter module to turn a little bit crazy.

Solution will be available with next release of Tol.
In other way, you can always get a copy of Tol from our CVS.