MathGroup Archive 2009

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Bug In LogPlots with WorkingPrecision option?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg97169] Re: [mg97140] Bug In LogPlots with WorkingPrecision option?
  • From: "David Park" <djmpark at comcast.net>
  • Date: Sat, 7 Mar 2009 02:37:41 -0500 (EST)
  • References: <19041450.1236333952924.JavaMail.root@m02>

I don't understand why Mathematica does that. WorkingPrecision seems to
throw the LogPlot logic off.

It is possible to make perfectly good Log plots by hand using
WorkingPrecision. I do it with  the Presentations package because I also
need to introduce custom ticks.

Needs["Presentations`Master`"]

With[{yticks = CustomTicks[Log[10, #] &, {-16, -2, {1}, {2, 5}},
    CTNumberFunction -> (N[#] &)]},
 Draw2D[
  {Draw[Log[10, Exp[-7 x]], {x, 1, 5},
    WorkingPrecision -> 20]},
  AspectRatio -> 1,
  Frame -> True,
  FrameTicks -> {{yticks, yticks // NoTickLabels}, {Automatic, 
     Automatic}},
  ImageSize -> 400]
 ]

The following plots your more complicated function. With the factor of 10^13
in the Sin function the function oscillates so rapidly that Mathematica
can't do a fine enough sample to catch a smooth set of points.
(WorkingPrecision is no help there.) I changed the factor to 1000 and then
we obtain a nice smooth descending sinusoidal curve on the Log scale.

With[{yticks = CustomTicks[Log[10, #] &, {-10, 2, {1}, {2, 5}},
    CTNumberFunction -> (N[#] &)]},
 Draw2D[
  {Draw[Log[10, Exp[-7 x]/Exp[-10 Sin[1000/(13 \[Pi]) x]]], {x, 1, 2},
    WorkingPrecision -> 20,
    MaxRecursion -> 2]},
  AspectRatio -> 1,
  Frame -> True,
  FrameTicks -> {{yticks, yticks // NoTickLabels}, {Automatic, 
     Automatic}},
  ImageSize -> 400]
 ] 


David Park
djmpark at comcast.net
http://home.comcast.net/~djmpark/  


From: M.Roellig [mailto:markus.roellig at googlemail.com] 

Hi Group,

the Documentation states that LogPlot takes the same options as Plot.
Unfortunately, LogPlot seems to have difficulties to properly swallow
them. Compare the follwoing examples:

p1 = Plot[Exp[-7 x], {x, 1, 5}]
p2 = LogPlot[Exp[-7 x], {x, 1, 5}, PlotRange -> {10^-10, 20}]
p3 = LogPlot[Exp[-7 x], {x, 1, 5}, PlotRange -> {10^-10, 20},
  WorkingPrecision -> 20]
p4 = LogPlot[Exp[-7 x], {x, 1, 5}, WorkingPrecision -> 20]

p2 looks good.  But if I use the WorkingPrecision option (even though
the example does not demands a particularly high precision) the plots
are messed up.
p3 is plotting something very strange but at least is able to keep the
axes ticks and labeling. Once I leave out the PlotRange everything is
completely messed up. I hope this is not a feature because I don't
understand it.
This occurs with all kinds of logarithmic plots.
Even more puzzling is the following example:

LogPlot[Exp[-7 x]/Exp[-10 Sin[10^13/(13 \[Pi]) x]], {x, 1, 2},
 PlotRange -> {10^-10, 20}]
LogPlot[Exp[-7 x]/Exp[-10 Sin[10^13/(13 \[Pi]) x]], {x, 1, 2},
 PlotRange -> {10^-10, 20}, WorkingPrecision -> 20]

Bug ?

I am using logarithmic plots a lot and am always surprised how fragile
the present implementation is. Especially  in science Log(Log)Plots
are used all the day and deserve some more love from the developers
(e.g. better automatic tick selection/control  or consistency in
positioning of text/Insets, etc...).

Cheers,

Markus R=F6llig




  • Prev by Date: Re: Style for function calls?
  • Next by Date: Re: list/sequence problem: a 'keep' list to a list of
  • Previous by thread: Bug In LogPlots with WorkingPrecision option?
  • Next by thread: Re: Bug In LogPlots with WorkingPrecision option?