| Re: Try this [message #12945] |
Wed, 23 September 1998 00:00 |
Karl Krieger
Messages: 8
Registered: April 1995
|
Junior Member |
|
|
On Wed, 23 Sep 1998, Joe wrote:
> But wouldn't they (RSI) have to go out of their way to purposefully
> convert the doubles to float within plot in order for this to occur?
> I would have thought that the double precision x-value would have
> caused an inheritence cascade throughout the plot routine so that
> everything would be double...
I guess that the internal C-code of "plot" uses automatic conversion
of the arguments to float as demonstrated in the "Advanced Development
Guide". To enable plots of double precision data, you needed either to
switch entirely to C double data inside "plot" or duplicate the internal
code to handle both data types appropiately.
Karl
--
Max-Planck-Institute for Plasma Physics
Boltzmannstr.2, 85740 Garching, Germany Email: krieger@ipp.mpg.de
|
|
|
|
| Re: Try this [message #12946 is a reply to message #12945] |
Wed, 23 September 1998 00:00  |
Joe[2]
Messages: 11
Registered: May 1998
|
Junior Member |
|
|
>
> In article <MPG.1071fa8a1e63e2a49896c8@news.frii.com>
> davidf@dfanning.com (David Fanning) writes:
>>
>> Ray Sterner (sterner@tesla.jhuapl.edu) writes:
>>
>>> Try this:
>>>
>>> x=dindgen(1000)/999.*20.
>>> for f=0.,10.,.02 do begin plot,/xstyl,x+f*1D6,sin(x) & empty & endfor
>>
>> Totally cool! :-)
>
> Weird yes, but definitively not cool for those who *want* to plot
> those data points!
>
> It took me a while to see what was going on here, but
> after running Ray's lines, do:
>
> plot,(x+1D7)(0:200),/xstyle,/ystyle,psym=3
>
> Now, a DOUBLE should have more dynamic range than to say that
> e.g., (x+1D7)(50) is *equal* to (x+1D7)(51). Luckily, it does:
>
> IDL> print,(x+1d7)([50,51]),form='(g20)'
> 10000001.00100100
> 10000001.02102102
>
> However, it appears that the PLOT command internally just works
> with FLOAT precision:
>
> IDL> print,FLOAT((x+1d7)([50,51])),form='(g20)'
> 1.000000e+07
> 1.000000e+07
>
> This could have been OK if some "zero point" value (typically the
> value of one of the tickmarks) had been subtracted in the process of
> converting to float. As it stands, this will have to be done
> by the user/programmer.
>
> Regards,
>
> Stein Vidar
But wouldn't they (RSI) have to go out of their way to purposefully
convert the doubles to float within plot in order for this to occur?
I would have thought that the double precision x-value would have
caused an inheritence cascade throughout the plot routine so that
everything would be double. Hmmmm... Oh wait, I got it! The
!x and !y structures are not double precision in their .range (and
other related) fields so it won't matter. This has to be done for
speed reasons since to make them double would force all plots
to use double precision calculations along with the attendant
decrease in computational rate. Still it is not obvious why they
need to force the x (and presumably y) variables to be float-type.
Z
|
|
|
|
| Re: Try this [message #12948 is a reply to message #12945] |
Wed, 23 September 1998 00:00 |
steinhh
Messages: 260
Registered: June 1994
|
Senior Member |
|
|
In article <MPG.1071fa8a1e63e2a49896c8@news.frii.com>
davidf@dfanning.com (David Fanning) writes:
>
> Ray Sterner (sterner@tesla.jhuapl.edu) writes:
>
>> Try this:
>>
>> x=dindgen(1000)/999.*20.
>> for f=0.,10.,.02 do begin plot,/xstyl,x+f*1D6,sin(x) & empty & endfor
>
> Totally cool! :-)
Weird yes, but definitively not cool for those who *want* to plot
those data points!
It took me a while to see what was going on here, but
after running Ray's lines, do:
plot,(x+1D7)(0:200),/xstyle,/ystyle,psym=3
Now, a DOUBLE should have more dynamic range than to say that
e.g., (x+1D7)(50) is *equal* to (x+1D7)(51). Luckily, it does:
IDL> print,(x+1d7)([50,51]),form='(g20)'
10000001.00100100
10000001.02102102
However, it appears that the PLOT command internally just works
with FLOAT precision:
IDL> print,FLOAT((x+1d7)([50,51])),form='(g20)'
1.000000e+07
1.000000e+07
This could have been OK if some "zero point" value (typically the
value of one of the tickmarks) had been subtracted in the process of
converting to float. As it stands, this will have to be done
by the user/programmer.
Regards,
Stein Vidar
(OK, so the visual effect *was* kind of cool :-)
|
|
|
|
| Re: Try this [message #12952 is a reply to message #12945] |
Tue, 22 September 1998 00:00 |
davidf
Messages: 2866
Registered: September 1996
|
Senior Member |
|
|
Ray Sterner (sterner@tesla.jhuapl.edu) writes:
> Try this:
>
> x=dindgen(1000)/999.*20.
> for f=0.,10.,.02 do begin plot,/xstyl,x+f*1D6,sin(x) & empty & endfor
Totally cool! :-)
David
----------------------------------------------------------
David Fanning, Ph.D.
Fanning Software Consulting
E-Mail: davidf@dfanning.com
Phone: 970-221-0438, Toll-Free Book Orders: 1-888-461-0155
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
