gnuplot-beta — Developers and beta testers of new gnuplot code

Re: error passing compound strings to system command

[Ethan A M. <merritt@u.washington.edu>]

On Monday 31 March 2008 04:02, pl...@pi... wrote:
> Hi,
> 
> I appear to have hit a limitation on setting up a command string for  
> system command. I have a "simple" bash command line that works in a  
> terminal:
> 
> bash-3.2#tail -n1 "../htdocs/data/adc.data"| awk 'format= " c3=%g\n  
> c4=%g\n c5=%g\n" {printf format,$3,$4,$5}' >temps.tmp
> bash-3.2#cat temps.tmp
>   c3=1.796
>   c4=1.676
>   c5=2.788
> 
> However, when I try to replicate this from within gnuplot is seems not to  
> get correct send to awk
> 
> 
>    system "tail -n1 '../htdocs/data/adc.data'| awk 'format= \" c3=%g\n  
> c4=%g\n c5=%g\n\"  {printf format,$3,$4,$5}' >temps.tmp  "
> 
> awk throws an error back to gnuplot:
> 
> gnuplot> load "control.gnu"
> awk: format= " c3=%g
> awk:         ^ unterminated string
> 
> 
> For some reason it seems not to see the second \"

No. The error message is exactly correct; the problem is that the
carriage return (\n) immediatedly after the c3=%g is terminating the
awk command.  It is safer to define a complex string in single
quotes rather then double quotes.


> I have tried to rearrange this in different ways but it always seems to  
> hit the same root cause.
> 
> Is this a possible bug in the parser?
> 
> Thx, Peter.
> 
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
> _______________________________________________
> gnuplot-beta mailing list
> gnu...@li...
> https://lists.sourceforge.net/lists/listinfo/gnuplot-beta
> 

-- 
Ethan A Merritt

error passing compound strings to system command

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

Hi,

I appear to have hit a limitation on setting up a command string for  
system command. I have a "simple" bash command line that works in a  
terminal:

bash-3.2#tail -n1 "../htdocs/data/adc.data"| awk 'format= " c3=%g\n  
c4=%g\n c5=%g\n" {printf format,$3,$4,$5}' >temps.tmp
bash-3.2#cat temps.tmp
  c3=1.796
  c4=1.676
  c5=2.788

However, when I try to replicate this from within gnuplot is seems not to  
get correct send to awk


   system "tail -n1 '../htdocs/data/adc.data'| awk 'format= \" c3=%g\n  
c4=%g\n c5=%g\n\"  {printf format,$3,$4,$5}' >temps.tmp  "

awk throws an error back to gnuplot:

gnuplot> load "control.gnu"
awk: format= " c3=%g
awk:         ^ unterminated string


For some reason it seems not to see the second \"
I have tried to rearrange this in different ways but it always seems to  
hit the same root cause.

Is this a possible bug in the parser?

Thx, Peter.

Re: screen terminal scrolling

[Petr M. <mi...@ph...>]

> BTW if there are no zoom variables exposed (which appears to be the case)  
> what file is the zoom code in, I may try to take a look if I can find some  
> of that mythical free time people sometimes refer to.

I've sent you two emails, one with answer, one with a solution. Your email 
address/mail server has permanent errors. You should fix them, or read the 
archive.

See also:
	show variables all

---
PM

Re: screen terminal scrolling

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

On Sat, 29 Mar 2008 19:57:50 +0100, <pl...@pi...> wrote:

> On Mon, 24 Mar 2008 08:22:57 +0100, <pl...@pi...> wrote:
>
>> Hi,
>>
>> unless I have missed something it seems there is no way to scroll a  
>> zoomed view of a plot with the on screen terminals.
>>
>> I invariably find I need to back out to a previous zoom level and  
>> redefine the zoom. I would considerably speed things up if the arrow  
>> keys allowed to scroll the same sized zoom in x or y.
>>
>> To prevent this being very slow (since it is basically a replot) it  
>> would probably be most useful to move one major tic at a time.
>>
>> If I am studying some data in detail I do want to have to break off  
>> mentally to zoom out , recall which bit I was looking at on the larger  
>> scale and then zoom back in a bit to the left.
>>
>> One keypress on the arrow keys would mean I did not even need to take  
>> my eyes off the screen.
>>
>> This would be a major plus in usability for on screen data viewing.
>>
>> Thx, Peter.
>
> Looking at this again , I wonder if it can be done with macros.
>
> It would be easy to bind something to "Left" (or whatever key is  
> prefered) but I dont see any exposed variables relating to the zoom  
> window that could be tweeked by a gnuplot command.
>
> Can anyone suggest if there is a way to adjust the current zoom window  
> parameters before calling replot?
>
> TIA. Peter.
>
>

BTW if there are no zoom variables exposed (which appears to be the case)  
what file is the zoom code in, I may try to take a look if I can find some  
of that mythical free time people sometimes refer to.

regards, Peter.

Re: 3d terminals - any hope?

[Michail V. <ma...@ia...>]

Dear Ethan,

On Sun, 30 Mar 2008, Ethan A Merritt wrote:

>> is there any plans to make real 3d terminal API that would provide
>> terminal driver with x,y,z as opposed to current x,y?
>
> No.
> That's not to say we would be opposed to such a plan, were someone to
> present a concrete proposal. But as of this moment there is no such plan.
> The bigger issue is simply that gnuplot was never designed for 3D use. 
> Some fairly major changes to the infrastructure would probably be  needed;
> it may not be worth the effort.

Thank you.
I suspected that, but wanted to hear from someone who knows gnuplot internals.

BTW, do you know anything gnuplot-like that does create files in some 3d format
(free, scriptable, maintained, mature, open-source, not a Godzilla-sized monster)?

       Sincerely, Michail

Re: logscale base bug

[Ethan A M. <merritt@u.washington.edu>]

On Sunday 30 March 2008 08:24, Juergen Wieferink wrote:
> > Am Sonntag, 30. März 2008 schrieb Allin Cottrell:
> > If the intent is to support non-integral bases (e.g. natural
> > logs), this should be
> >
> > newbase = fabs(real_expression());
> >
> > since abs() returns an int.
> 
> Mea culpa.  Please commit the proposed change.

Done.
 
-- 
Ethan A Merritt

Re: 3d terminals - any hope?

[Ethan A M. <merritt@u.washington.edu>]

On Sunday 30 March 2008 08:40, Michail Vidiassov wrote:
> Dear All,
> 
> is there any plans to make real 3d terminal API that would provide 
> terminal driver with x,y,z as opposed to current x,y?

No.
That's not to say we would be opposed to such a plan, were someone to
present a concrete proposal. But as of this moment there is no such
plan.

 
> IMHO, there were attempts to introduce VRML and POV-Ray output, but they
> had to be hacks and were not welcomed by the core team.
> 
> The reason for my interest is the fact that now Adobe Reader has 3D 
> capabilities, it can show embedded 3D models without any plugins.
> That may solve one of the problems of 3D adoption - the lack of a widely 
> installed viewer.
> There is free software to make such models from VRML-like text files and 
> embed them into PDF.
> (a gallery of 3D pdfs is at http://www.pdf3d.co.uk/gallery.php
> scientific use example can be fount at  http://astronomy.swin.edu.au/s2plot/3dpdf/ )
> 
> Hope that will make the gnuplot team reevaluate the prospects of 3D support.

As always in a collaborative open source project, the best way to move a project
in a particular direction is to contribute code.  Let me note, however, that I
don't the that lack of viewers is a major problem.  The bigger issue is simply
that gnuplot was never designed for 3D use.  Some fairly major changes to the
infrastructure would probably be needed;  it may not be worth the effort.


-- 
Ethan A Merritt
Biomolecular Structure Center
University of Washington, Seattle 98195-7742

3d terminals - any hope?

[Michail V. <ma...@ia...>]

Dear All,

is there any plans to make real 3d terminal API that would provide 
terminal driver with x,y,z as opposed to current x,y?

IMHO, there were attempts to introduce VRML and POV-Ray output, but they
had to be hacks and were not welcomed by the core team.

The reason for my interest is the fact that now Adobe Reader has 3D 
capabilities, it can show embedded 3D models without any plugins.
That may solve one of the problems of 3D adoption - the lack of a widely 
installed viewer.
There is free software to make such models from VRML-like text files and 
embed them into PDF.
(a gallery of 3D pdfs is at http://www.pdf3d.co.uk/gallery.php
scientific use example can be fount at  http://astronomy.swin.edu.au/s2plot/3dpdf/ )

Hope that will make the gnuplot team reevaluate the prospects of 3D support.

      Sincerely, Michail

Re: logscale base bug

[Juergen W. <wie...@fr...>]

Am Sonntag, 30. März 2008 schrieb Allin Cottrell:
> This happens because on line 2054 of set.c we have
>
> newbase = abs(real_expression());
>
> If the intent is to support non-integral bases (e.g. natural
> logs), this should be
>
> newbase = fabs(real_expression());
>
> since abs() returns an int.
>
> Note: the "abs()" formulation is new as of revision 1.241 of
> set.c (Jan 2007); before that newbase was set as
>
> newbase = magnitude(const_express(&a));
>
> which preserved the non-integral part.

Mea culpa.  Please commit the proposed change.

Juergen

logscale base bug

[Allin C. <cot...@wf...>]

The help for logscale doesn't state that the base must be 
integral, but at present user input is truncated to an integer
and you can have this kinda funny dialog with gnuplot:

gnuplot> set logscale y 1.5
                           ^
         log base must be >= 1.1; logscale unchanged

This happens because on line 2054 of set.c we have

newbase = abs(real_expression());

If the intent is to support non-integral bases (e.g. natural 
logs), this should be

newbase = fabs(real_expression());

since abs() returns an int.

Note: the "abs()" formulation is new as of revision 1.241 of 
set.c (Jan 2007); before that newbase was set as

newbase = magnitude(const_express(&a));

which preserved the non-integral part.

-- 
Allin Cottrell
Department of Economics
Wake Forest University, NC

Re: helpfile breakage

[Ethan A M. <merritt@u.washington.edu>]

On Sunday 30 March 2008 07:44, Allin Cottrell wrote:
> The change from revision 1.167 to 1.168 of command.c has broken 
> gnuplot helpfile access.  

Got it.  Thanks.


-- 
Ethan A Merritt

helpfile breakage

[Allin C. <cot...@wf...>]

The change from revision 1.167 to 1.168 of command.c has broken 
gnuplot helpfile access.  

Specifically, you get a segfault on trying to access help.  This 
is because the help() function is being called with the pointer 
help_ptr == NULL, and that in turn is because the removal of the 
atari-specific code was not sufficiently surgical: it also
removed the default of

help_ptr = HELPFILE;

for platforms other than OS/2.

Patch attached.

-- 
Allin Cottrell
Department of Economics
Wake Forest University, NC

Re: screen terminal scrolling

[Petr M. <mi...@ph...>]

> Looking at this again , I wonder if it can be done with macros.
> 
> It would be easy to bind something to "Left" (or whatever key is prefered)  
> but I dont see any exposed variables relating to the zoom window that  
> could be tweeked by a gnuplot command.

I think I've sent a recipe how to do it. Anyway, here is a full 
implementation:

# 4.2.x: use "replot" command
# 4.3:   use "refresh" command

# Note: only for linear scale; for GPVAL_X_LOG!=0, do it accordingly
DELTA_X(a)=a*(GPVAL_X_MAX-GPVAL_X_MIN)*0.5
DELTA_Y(a)=a*(GPVAL_Y_MAX-GPVAL_Y_MIN)*0.5

# enlarge limits independently:
bind "Ctrl-Right" \
  "set xrange [GPVAL_X_MIN+DELTA_X(1):GPVAL_X_MAX+DELTA_X(1)]; replot"
bind "Ctrl-Left"  \
  "set xrange [GPVAL_X_MIN-DELTA_X(1):GPVAL_X_MAX-DELTA_X(1)]; replot"
# scroll:
bind "Right"      "set xrange [GPVAL_X_MIN:GPVAL_X_MAX+DELTA_X(2)]; replot"
bind "Left"       "set xrange [GPVAL_X_MIN-DELTA_X(2):GPVAL_X_MAX]; replot"
# squeeze limits independently: not needed, use zoom by mouse instead

# enlarge limits independently:
bind "Ctrl-Up"    "set yrange 
[GPVAL_Y_MIN+DELTA_Y(1):GPVAL_Y_MAX+DELTA_Y(1)]; replot"
bind "Ctrl-Down"  "set yrange 
[GPVAL_Y_MIN-DELTA_Y(1):GPVAL_Y_MAX-DELTA_Y(1)]; replot"
# scroll:
bind "Up"         "set yrange [GPVAL_Y_MIN:GPVAL_Y_MAX+DELTA_Y(2)]; replot"
bind "Down"       "set yrange [GPVAL_Y_MIN-DELTA_Y(2):GPVAL_Y_MAX]; replot"
# squeeze limits independently: not needed, use zoom by mouse instead

Re: screen terminal scrolling

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

On Mon, 24 Mar 2008 08:22:57 +0100, <pl...@pi...> wrote:

> Hi,
>
> unless I have missed something it seems there is no way to scroll a  
> zoomed view of a plot with the on screen terminals.
>
> I invariably find I need to back out to a previous zoom level and  
> redefine the zoom. I would considerably speed things up if the arrow  
> keys allowed to scroll the same sized zoom in x or y.
>
> To prevent this being very slow (since it is basically a replot) it  
> would probably be most useful to move one major tic at a time.
>
> If I am studying some data in detail I do want to have to break off  
> mentally to zoom out , recall which bit I was looking at on the larger  
> scale and then zoom back in a bit to the left.
>
> One keypress on the arrow keys would mean I did not even need to take my  
> eyes off the screen.
>
> This would be a major plus in usability for on screen data viewing.
>
> Thx, Peter.

Looking at this again , I wonder if it can be done with macros.

It would be easy to bind something to "Left" (or whatever key is prefered)  
but I dont see any exposed variables relating to the zoom window that  
could be tweeked by a gnuplot command.

Can anyone suggest if there is a way to adjust the current zoom window  
parameters before calling replot?

TIA. Peter.

Re: getting date into a label

[Thomas S. <t.s...@fz...>]

> I'm astounded everybody else forgot that gnuplot has had this feature 
> since just about forever:
>
>	set timestamp '%Y-%m-%d' top

plotter wanted to have the date in the title.
(a timestamp is not a title - it can't be centered)

-- 
View this message in context: http://www.nabble.com/getting-date-into-a-label-tp16296555p16324295.html
Sent from the Gnuplot - Dev mailing list archive at Nabble.com.

Re: getting date into a label

[Hans-Bernhard B. <HBB...@t-...>]

pl...@pi... wrote:

> I was wondering if there was not a gnuplot command to do this but help  
> date did not help.

"help time", OTOH, might have ;-P

I'm astounded everybody else forgot that gnuplot has had this feature 
since just about forever:

	set timestamp '%Y-%m-%d' top

Re: Announcing the book on the website

[Philipp K. J. <ja...@ie...>]

That's cool!

Could you please make the file (index.html)
group-writable? Currently, nobody but you can 
touch it.

Best,

		Ph.


On Wednesday 26 March 2008 01:06, you wrote:
> > > 	http://philipp-janert.com/gnuplot/www/index4.html
>
> I've added "News" section, see
> 	gnuplot.sourceforge.net
>
> ---
> PM

Re: getting date into a label

[Juergen W. <wie...@fr...>]

Am Mittwoch, 26. März 2008 schrieb Thomas Sefzick:
> > eg
> > today = date("+%D");
> > title=title.today
>
> titletext="plotting time: "
> today=`date +\"%D\"`
> set title titletext.today
>
> or
>
> titletext="plotting time: "
> set title titletext.`date +\"%D\"`

Or

set title "plotting time: " . system("date +%D")


Juergen

Re: getting date into a label

[Thomas S. <t.s...@fz...>]

> eg
> today = date("+%D");
> title=title.today

titletext="plotting time: "
today=`date +\"%D\"`
set title titletext.today

or

titletext="plotting time: "
set title titletext.`date +\"%D\"`

-- 
View this message in context: http://www.nabble.com/getting-date-into-a-label-tp16296555p16297579.html
Sent from the Gnuplot - Dev mailing list archive at Nabble.com.

Re: Announcing the book on the website

[Petr M. <mi...@ph...>]

> > 	http://philipp-janert.com/gnuplot/www/index4.html

I've added "News" section, see
	gnuplot.sourceforge.net

---
PM

getting date into a label

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

Hi,

I was wondering if there was not a gnuplot command to do this but help  
date did not help.

I am trying to automatically add the date to the title of a plot.
for some reason the following does not respect the -n option and adds  
newline.

datefile="date.txt";
system "echo -n 'date= \"' >".datefile." ; date '+%D'>>".datefile."\""
load datefile
title=title.date

finally got what I want with this:

datefile="date.txt";
system "today=\'date= \"\'`date '+%D'`\'\"\'; echo $today>".datefile
load datefile


If this is not just a feature I cannot find, could it be added? Gnuplot  
already parses all the std date format args, it would presumably be  
trivial to add.

eg
today = date("+%D");
title=title.today


Thx.

Re: web site markup errors

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

On Tue, 25 Mar 2008 19:36:35 +0100, <pl...@pi...> wrote:

> On Mon, 24 Mar 2008 22:57:42 +0100, Ethan Merritt
> <merritt@u.washington.edu> wrote:
>
>> Not sure what errors you were getting.
>> The only validation errors I get for the gnuplot home page are warnings
>> like "left-border: 3%" is deprecated syntax.  There is no font-size
>> markup
>> other than the header and footer.
>>
>
>
> It should be pretty clear what errors there were , I posted a link to the
> validiator at w3c I used and cut and pasted most of the errors it showed.
>
> The fact the text was coming out other than intended could be due to any
> mark up error screwing up the parsing near the text concerned. It may
> misinterpret quite a bit before it gets back in sync.
>
> I saw no point in working out extactly why it was wrong until the
> maintainer of the page got it to validate.
>
> Some of the more recent links in this thread seem to be fixed since that
> text is coming out much more readably and "gnuplot homepage" is no longer
> wrapping.
>
> I'll wait for the dust to settle on these changes and post back if I  
> still
> see anything out of place.
>
> Use of the w3c validator should get things in order anyway.
>
> regards.
>

BTW you currently have at least three things calling themselves "gnuplot  
homepage":

http://gnuplot.sourceforge.net/index.html
http://www.gnuplot.info/
http://www.gnuplot.info/documentation.html

My comments were based on 2) the top google result , 3) would seems to be  
an incorrect title but is coming up 2nd place on google.

The one you all seem to be calling the homepage is not even in the top 30  
google results , so I guess you need to be on the team to know it's there.

While there is some work being done on the site content you may want to  
rationalise some of the confusion there.

regards, Peter.

web site markup errors

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

On Mon, 24 Mar 2008 22:57:42 +0100, Ethan Merritt  
<merritt@u.washington.edu> wrote:

> Not sure what errors you were getting.
> The only validation errors I get for the gnuplot home page are warnings
> like "left-border: 3%" is deprecated syntax.  There is no font-size  
> markup
> other than the header and footer.
>


It should be pretty clear what errors there were , I posted a link to the  
validiator at w3c I used and cut and pasted most of the errors it showed.

The fact the text was coming out other than intended could be due to any  
mark up error screwing up the parsing near the text concerned. It may  
misinterpret quite a bit before it gets back in sync.

I saw no point in working out extactly why it was wrong until the  
maintainer of the page got it to validate.

Some of the more recent links in this thread seem to be fixed since that  
text is coming out much more readably and "gnuplot homepage" is no longer  
wrapping.

I'll wait for the dust to settle on these changes and post back if I still  
see anything out of place.

Use of the w3c validator should get things in order anyway.

regards.

Re: Contributions to web pages and FAQ

[Ethan M. <merritt@u.washington.edu>]

On Tuesday 25 March 2008 10:11, Tim Hoffmann wrote:
> > I remember there was an announcement of JavaPlot, but I had no time to edit 
> > web pages. Any volunteer for contributions to web pages, as well as to the 
> > FAQ, is welcome.
> This is already included for a while. Accidentally it went in the 
> front-end section rather than in the library section. Corrected:
> http://home.arcor.de/lord_astfgl/links.html
> 
> Is there a reason why h3 is colored green in the css?

[shrug]
I took the css file from one of my own web pages, where there is a
consistent color scheme. But the gnuplot pages don't use the same
conventions for category headings and div colors.

-- 
Ethan A Merritt            Courier Deliveries: 1959 NE Pacific
Dept of Biochemistry
Health Sciences Building
University of Washington - Seattle WA 98195-7742

Re: Contributions to web pages and FAQ

[Petr M. <mi...@ph...>]

> Is there a reason why h3 is colored green in the css?

No. It look strange. Patch for css welcome.

> I'm not sure if it's worth the effort, but it might be reasonable to have the
> website sourcecode included in the CVS.

It might ... somebody would have to do it...