Am Montag, den 12.12.2011, 23:07 +0100 schrieb Mojca Miklavec:
> On Mon, Dec 12, 2011 at 22:39, Paul Menzel wrote:

> > I have two coefficients saved in a text file
> >
> >        -0.157737135681261
> >        -0.211443105668896
> >
> > and I want to use them to plot a line with gnuplot.
> >
> > Having several of those files I want to use Lua to create the formula
> > for the line for me.
> >
> > I have not yet figured out how to read values from a file in Lua and
> > just tried to print a constant to the gnuplot command.
> >
> > I tried the following.
> >
> >        \usemodule[gnuplot]
> >
> >        \startGNUPLOTscript[lua]
> >           plot \ctxlua{context(1.2 * x + 3)}
> >        \stopGNUPLOTscript
> 
> This won't work. The reason is that contents between
> \startGNUPLOTscript and \stopGNUPLOTscript are not parsed by ConTeXt
> on purpose.
> 
> It was very very very painful if not impossible to get it right with
> parsing. I often do things like
>     \def\E#1{$10^{#1}$}
> 
>     \startGNUPLOTscript[name]
>     set logscale y
>     set format y "\\E{%T}"
>     plot sin(x) t '$\sin(x) [\E{3}]$' # usually something else with
> logscale, but just to show the point
>     \stopGNUPLOTscript
> 
> and if ConTeXt starts parsing this it becomes a pure nightmare.
> Realize that it can easily happen that ConTeXt then writes out
>     set logscale y
>     set format y "<some junk leftovers from \\>
>     E<ungrouped { since % starts comment>
>     plot sin(x) t '\math{\mathrm{sin}(x) [{10^{3}}]}' <arbitrary junk
> from expansion ...>
> 
> 
> 
> However, \startGNUPLOTinclusions is parsed and you could in theory
> misuse that, but it would be very bad coding practice.

Thank you very much for your explanations.

> It would be a lot better to come up with some proposal how to change
> the interface to enable what you want to do. (To be honest, it is
> still on my todo list to migrate to Aditya's filter module in the
> background, but I don't find the courage to start since I have no idea
> how many bugs will creep in.)

I do not know this very well, so I do not know what would change when
using the filter module. I guess in the end being able to pass a
plt-file to `\startGNUPLOTscript` would be enough.

> But a serious question: what *exactly* do you want to do with lua that
> you are unable to do with some (even though ugly) gnuplot trickery?

I do not know. This was just a product out of my ignorance. Searching
for `gnuplot load parameters` or something like that only turned `fit`
related stuff.

Thinking about this, I should have searched harder or looked at the
manual or examples.

> Gnuplot does have some basic programmable capabilities. I can imagine
> that you should be able to do something like:
> 
> # params.dat
> a=-0.157737135681261
> b=-0.211443105668896
> 
> \startGNUPLOTinclusions
> f(x,a,b)=a*x+b
> \stopGNUPLOTinclusions
> 
> \startGNUPLOTscript[name]
> load 'params.dat'
> plot f(x,a,b)
> \stopGNUPLOTscript
>
> And if needed you can create params.dat on the fly (but I believe that
> the same should be doable even if you start with just pure data values
> and do some dirty tricks in gnuplot).

That pretty much what I need. Thank you!

What I do is to do simulations and then do the linear regression using R
[1]. (I know gnuplot can do that too, but I got some errors with it and
using R seems easier for me.) Then the calculated coefficient are saved
to a separate file because having discovered your nice gnuplot module I
want to use that for plotting. Now I need to read in those parameters. I
guess I have to compose the plot manually but those parameters were
essential.

> Yet another trick could be to create a text file on the fly with lua
> (if you really really really need that) and then your script only says
> something like.
> 
> # bla.plt created on the fly with lua scripts
> plot 1.2 * x + 3
> 
> \startGNUPLOTscript[bla]
> load 'bla.plt'
> \stopGNUPLOTscript
> 
> I leave creating a suitable 'bla.plt' with lua to others ...

That is in my opinion the best solution. If I want to script something
there are enough solutions already. Being able to use the tools of once
choice and to insert the result should fit everyone’s need.

Thank you again very much for your help.


Thanks,

Paul


[1] http://www.r-project.org/