Re: typo injected an undeclared function into the name space ?

[<pl...@pi...>]

On 15/01/16 18:38, Ethan A Merritt wrote:
> On Friday, 15 January, 2016 17:54:37 pl...@pi... wrote:
>
>  > On 15/01/16 17:40, Ethan A Merritt wrote:
>
>  > > On Friday, 15 January, 2016 11:38:01 pl...@pi... wrote:
>
>  > >
>
>  > > > Hi ,
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > I accidentally forgot a * operator when plotting a function. This
> seems
>
>  > >
>
>  > > > to have injected an udefined fn into the name space
>
>  > >
>
>  > > >
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > twopi=2*pi
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > gnuplot> plot datafile" u ($1/365.25+1970):2 w l, cos
> (twopi(x)/p3) +0.05
>
>  > >
>
>  > > > undefined function: twopi
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > User-Defined Functions:
>
>  > >
>
>  > > > cos1(x)=a1*cos(twopi*(x-phi1)/p1)
>
>  > >
>
>  > > > cos2(x)=a2*cos(twopi*(x-phi2)/p2)
>
>  > >
>
>  > > > cos3(x)=a3*cos(twopi*(x-phi3)/p3)
>
>  > >
>
>  > > > triple(x)=c+cos3(x)
>
>  > >
>
>  > > > twopi is undefined
>
>  > >
>
>  > > >
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > gnuplot> print twopi
>
>  > >
>
>  > > > 6.28318530717959
>
>  > >
>
>  > > >
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > Now it won't stop crapping on about "twopi is undefined", even
> though I
>
>  > >
>
>  > > > have plotted something else and make no reference to it.
>
>  > >
>
>  > > >
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > All I can find to clear it is to quit gnuplot.
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > Has this , as it seems, injected an undeclared function into the name
>
>  > >
>
>  > > > space ?
>
>  > >
>
>  > > >
>
>  > >
>
>  > > > How to clear it more elegantly, is this a bug?
>
>  > >
>
>  > > There is currently no way to clear or delete any user-provided
> function.
>
>  > >
>
>  > > That is true for your well-defined functions as well as your undefined
>
>  > >
>
>  > > function. The "undefine FOO" command only looks for FOO as a variable
>
>  > >
>
>  > > name not as a function name.
>
>  > >
>
>  > > If you want to reset the entire namespace you can use "reset session".
>
>  > >
>
>  > > Of course that will clear everything, not just the undefined function.
>
>  > >
>
>  > > Ethan
>
>  > >
>
>  >
>
>  > Thanks Ethan,
>
>  >
>
>  > the main point was, is this a bug?
>
>  >
>
>  > how did I manage to register an undefined function simply by trying to (
>
>  > accidentally ) use it in a plot command?
>
>  >
>
>  > How does that become a function definition, right or wrong ?
>
> gnuplot allows you to refer to a function before you define it.
>
> Consider:
>
> g(x) = f1(x) + f2(x)
>
> In order to store the evaluation template for g(x) it has to reserve
>
> slots for f1() and f2() even if they have not been defined yet.
>
> If you say "show fun" at this point you will see
>
> User-Defined Functions:
>
> g(x) = f1(x) + f2(x)
>
> f1 is undefined
>
> f2 is undefined
>
> Issuing a plot command triggers the same thing implicitly.
>
> The program constructs a template for how to evaluate the values
>
> being plotted, reserving a slot for any functions that are invoked.
>
> If it turns out that one of those functions is not yet defined then
>
> the plot will fail, but just as in the above example it leaves
>
> an entry in the table of functions that can be filled in later.
>
> You can see this at work in the following admittedly bizarre
>
> plot command that plots f(x) twice but redefines it in between:
>
> plot f(x)=sin(x) f(x), f(x)=cos(x) f(x)
>
> That works because both curves in the plot are evaluated using
>
> the template "f(x)" but the definition of "f()" has changed
>
> between the two evaluations of the same template.
>
> So - not a bug.
>
>
> Ethan
>

Thanks for giving a good understanding of what is happening.

Not a bug?  Well it must be feature then.

I can see the point of inserting a place holder when defining functions.
I can see no useful purpose in the same thing happening during a failed 
plot command. When that pollutes the namespace afterwards, it is not a 
very useful feature.

The solution of resetting the session is essentially just like quitting 
and restarting gnuplot. This can be quite a time waster if the user has 
defined several function , constants, string constants for axis labels, 
arrows and graph labels and has to start again.

It looks more  like an unintended side effect than a designed feature. 
Maybe the real problem is that there is no way of removing such zombies 
from the name space as can be done with variables.


I have a numeric constant  twopi that I can remove with undef , maybe 
what is needed is an equivalent fro function definitions.

regards, Peter.