Page 2 of 2

Re: Feature requests - general topic

Posted: April 24th, 2019, 9:53 am
by MagicPlot
Thank you, confirmed as a bug. The correct rows number is available only if all checkboxes in Data area are unset. We'll fix it.

Re: Feature requests - general topic

Posted: April 25th, 2019, 10:37 am
by Lautibah
I am sorry for beeing too much demanding, but as developers you will think what is better or not for the program. This might not be useful or is just too unoptimized (obviously the best is: do the formula derivative , equal it to 0, after solve. I think this is too expensive and requires expert software, that's one reason is not implemented on magicplot)
I've exported the fit curve to a table with 10^5 points (unchecking the checkboxes) (by the way, the Add to project into > only new folder option is available). After exporting the table, I've calculated in a new column the cmax() for Y (this gives us the Y component of the maximum) and after, on a new column I have added the formula if(col(Y) == cmax(Y), col(X),0), after another column and cmax() (this gives us X component of the maximum). This gave me a pretty good aproximation of the maximum (depending on the number of points the precision is greater).

I was thinking on what you could do. I think I am not that wrong when I think that the maximas and minimas should appear within the fit plot interval. Then, using simple tools as solving the fit formula for multiples X values (number of points), cmax() and iterations, an aproximation might be possible.
I don't know, I thought that "scanning" the morphology of the fit formula with 1000 points, getting a maximum there with cmax (relative maximas won't be detected with cmax), now scanning the formula with 10000 points closeby the previous maximum value and finally, scanning it with 10^n points should give us a really small error on the maxima coordinates). The aproximation could be given in the Report tab because is not an exact value. I might not be that wrong in my thinking... I just don't know.

I know is not that easy, you'd had programmed that feature years ago. It could exist a plot whose maxima width is smaller than the X interval on the scanning:
- One solution for that is do the first scan of the morphology with a number of points > the smallest interval within experimental data. It is not that simple as (Xmax-Xmin)/rows, because the interval between experimental points might not be constant. So through performing a "for loop", the smallest interval between experimental points should be obtained. Then, the number of points for the scan must be greater or much greater than that. "Only by this way" a true maximum might be obtained.
- Other solution for that is cmax() the experimetal points and scanning through those, but it could also be that that experimental point is an error on the measure and casually is the maximum point.

The best is give by hand a relative precise X range in which the maxima is contained, afterwards do "all this".

Re: Feature requests - general topic

Posted: April 25th, 2019, 11:10 am
by MagicPlot
OK, I have added a feature request ticket for finding maximum/minimum of the fit curve, it seems to be a reasonable feature. I cannot say now when it can be implemented.