gnuplot development

Jacques Le Bourlot wrote:

> * libgd:
>    I had to manually create a link from libgd.so.2 to libgd.so in
> /usr/lib. 

You should never have to do that.  Either ldconfig should do that 
automatically at installation time, or there must be some good reason it 
didn't.

 > It was not found by configure.

That was actually a good thing, since it didn't really work.  configure 
should only find actually working libraries.

 > Also gd was installed, but not
> gd-devel so that "make" crashed when looking for gd.h (which was not
> there). probably configure should look for both libgd and gd.h

Not really.  It only has to look for gdlib-config and trust its output.

Jacques Le Bourlot a =E9crit :
> Bonjour,
>
> I just compiled and installed RC2 on a Linux Laptop with SuSE 9.3 (+ al=
l
> latest upgrades I could find). I had very few issues. Thank you for the
> nice job! Here are the pbs:
>
> <...>
> * Pango:
>    SuSE comes with pango 1.10.2 and I could not find newer packages. I
> compiled 1.10.4 and had no further problems at
> configuration/compilation time. But...
>
> * Wxt:
>    The terminal works (i.e. I get all the nice plots in the window).
> However all texts are ridiculously small (size 1 point?) and useless. I
> could not figure out why.
>  =20
It is precisely the problem I was trying to address when I added the=20
restriction on pango not being 1.10.2.
Here is the corresponding thread:=20
http://comments.gmane.org/gmane.comp.graphics.gnuplot.devel/5319

I don't clearly remember if I tried with pango 1.10.4. Would you mind=20
trying with a newer version of pango, in the 1.12 or 1.14 series ?
I'll try to compile gnuplot with pango 1.10.4, but I won't be able to do=20
it until tomorrow evening.


> * UTF8:
>   Results of "utf8.dem" are unreliable. May come from my fonts
> installation (which I never really understood). More specifically, I ca=
n
> not set encoding to "utf8" (I know it is not in the list...), and could
> not find how to get such characters as "=E9", "=E0", etc... which are
> mandatory in french. Note: I use the standard SuSE setting : LANG =3D
> fr_FR.UTF-8, so neither "default" nor "iso_8859_15" work.
>  =20
Please tell us what terminal you're using since text encoding is=20
terminal specific.

Best regards,

Timoth=E9e Lecomte

On Monday 27 November 2006 01:16 pm, Timoth=E9e Lecomte wrote:
> Jacques Le Bourlot a =E9crit :
> * libgd:
>    I had to manually create a link from libgd.so.2 to libgd.so in
> /usr/lib. It was not found by configure. Also gd was installed, but
> not gd-devel so that "make" crashed when looking for gd.h=20

This is telling you that you should have installed gd-devel.
=46orcing the build process to continue when a required package is
missing is not a good idea.

> > * Pango:
> >    SuSE comes with pango 1.10.2 and I could not find newer
> > packages. I compiled 1.10.4 and had no further problems at
> > configuration/compilation time. But...
> >
> > * Wxt:
> >    The terminal works (i.e. I get all the nice plots in the
> > window). However all texts are ridiculously small (size 1 point?)
> > and useless. I could not figure out why.
>
> It is precisely the problem I was trying to address when I added the
> restriction on pango not being 1.10.2.
> Here is the corresponding thread:
> http://comments.gmane.org/gmane.comp.graphics.gnuplot.devel/5319

The wxt terminal works just fine with libpango 1.10.0

I think the symptom of invisibly small font sizes comes from not
having properly initialized fontconfig.  I originally suffered from
the same problem, but installing and initializing fontconfig fixed it.


> > * UTF8:
> >   Results of "utf8.dem" are unreliable. May come from my fonts
> > installation (which I never really understood). More specifically,
> > I can not set encoding to "utf8"=20

It is not necessary to set encoding to utf8.
The "encoding" tells how to interpret 1-byte characters.
utf8 uses multibyte characters *instead* of using the 1-byte
representations.
=20
> > and could not find how to get such characters as "=E9", "=E0", etc...

=E9 is Unicode code point 00E9
=E0 is Unicode code point 00E0

If your locate is set to fr_FR.UTF-8, then when you type =E0 on the
keyboard it generates a two-byte utf-8 sequence that represents that
Unicode code point.  This is *different* than the 1-byte representation
used by iso_8859_15.  You cannot mix them.  You must pick one or the
other (preferably utf-8).

If you are using KDE, a convenient tool to explore this is kcharselect. =20
I'm sure some equivalent tool must exist on other desktops.

=2D-=20
Ethan A Merritt
Biomolecular Structure Center
University of Washington, Seattle WA

> On Monday 27 November 2006 01:16 pm, Timoth=E9e Lecomte wrote:
>> Jacques Le Bourlot a =E9crit :
>> * libgd:
>>    I had to manually create a link from libgd.so.2 to libgd.so in
>> /usr/lib. It was not found by configure. Also gd was installed, but
>> not gd-devel so that "make" crashed when looking for gd.h
>
> This is telling you that you should have installed gd-devel.
> Forcing the build process to continue when a required package is
> missing is not a good idea.
>
That's what I did immediattely! I did not "force" the build process any
further.

>> > * Pango:
>> >    SuSE comes with pango 1.10.2 and I could not find newer
>> > packages. I compiled 1.10.4 and had no further problems at
>> > configuration/compilation time. But...
>> >
>> > * Wxt:
>> >    The terminal works (i.e. I get all the nice plots in the
>> > window). However all texts are ridiculously small (size 1 point?)
>> > and useless. I could not figure out why.
>>
>> It is precisely the problem I was trying to address when I added the
>> restriction on pango not being 1.10.2.
>> Here is the corresponding thread:
>> http://comments.gmane.org/gmane.comp.graphics.gnuplot.devel/5319
>
> The wxt terminal works just fine with libpango 1.10.0
>
> I think the symptom of invisibly small font sizes comes from not
> having properly initialized fontconfig.  I originally suffered from
> the same problem, but installing and initializing fontconfig fixed it.
>
I will try that. However, It'll have to wait for a few days (I'm an
astrophysicist and teacher at the University: I have very few time to
spare).

>
>> > * UTF8:
>> >   Results of "utf8.dem" are unreliable. May come from my fonts
>> > installation (which I never really understood). More specifically,
>> > I can not set encoding to "utf8"
>
I'll explore also the font issue. I usually use X11 as a term, then turn
to postscript at the very end.

Thank you for the advices.

Jacques Le Bourlot

Joe Koski wrote:
> Ethan, Timoth=E9e,
>
> I tried building gnuplot-4.2.rc2 on my G5 Mac with OS X 10.4.8, Xcode2.=
4
> Developer tools, and same pango/cairo that I had built previously. Ther=
e
> have been no changes to my build routines since I successfully built
> gnuplot-4.2.rc1.
>
> During the build, all goes well until:
>
> <snip> I seem to be missing some parts (a header file?) necessary to bu=
ild the wxt
> terminal.
>  =20
You still need to patch in src/wxterminal with wxmac.diff
I have not committed it yet to the CVS, although I should have...

Thank you again for your efforts in testing this.

Best regards,

Timoth=E9e

on 11/25/06 2:00 PM, Timoth=E9e Lecomte at timothee.lecomte@... wrote:

> Joe Koski wrote:
>> Ethan, Timoth=E9e,
>>=20
>> I tried building gnuplot-4.2.rc2 on my G5 Mac with OS X 10.4.8, Xcode2.4
>> Developer tools, and same pango/cairo that I had built previously. There
>> have been no changes to my build routines since I successfully built
>> gnuplot-4.2.rc1.
>>=20
>> During the build, all goes well until:
>>=20
>> <snip> I seem to be missing some parts (a header file?) necessary to bui=
ld
>> the wxt
>> terminal.
>>  =20
> You still need to patch in src/wxterminal with wxmac.diff
> I have not committed it yet to the CVS, although I should have...
>=20
> Thank you again for your efforts in testing this.
>=20
> Best regards,
>=20
> Timoth=E9e

Timoth=E9e,

Yes, that fixed the problem. Thanks. Yes, I think that the patch can be
submitted, and then we can worry about bundling later.

Joe

Joe Koski wrote:
> on 11/25/06 2:00 PM, Timoth=E9e Lecomte at timothee.lecomte@... wrot=
e:
>
>  =20
>> Joe Koski wrote:
>>    =20
>>> Ethan, Timoth=E9e,
>>>
>>> I tried building gnuplot-4.2.rc2 on my G5 Mac with OS X 10.4.8, Xcode=
2.4
>>> Developer tools, and same pango/cairo that I had built previously. Th=
ere
>>> have been no changes to my build routines since I successfully built
>>> gnuplot-4.2.rc1.
>>>
>>> During the build, all goes well until:
>>>
>>> <snip> I seem to be missing some parts (a header file?) necessary to =
build
>>> the wxt
>>> terminal.
>>>  =20
>>>      =20
>> You still need to patch in src/wxterminal with wxmac.diff
>> I have not committed it yet to the CVS, although I should have...
>>
>> Thank you again for your efforts in testing this.
>>
>> Best regards,
>>
>> Timoth=E9e
>>    =20
>
> Timoth=E9e,
>
> Yes, that fixed the problem. Thanks. Yes, I think that the patch can be
> submitted, and then we can worry about bundling later.
>
> Joe
>  =20
I have just committed the patch to both CVS HEAD and 4.2 branch. Let's=20
hope we'll find how to bundle it correctly now.

Best regards,

Timoth=E9e

Hello,

I want to understanding how gnuplot calculates xtic for a given range  
e.g.
plot x
set xrange [0:100]
replot

will place tic marks at 0, 20, 40, 60, 80 and 100.

Where is this calculated in the code?

How does gnuplot calculate the step size between tic marks?

I have worked on my own algorithm for xtics position for a given  
range and it works well.  However it required the user to input the  
step size between tic marks. e.g. plotting x from 0 to 100 the user  
would have to specify the tic separation was 20.

Regards,

Daniel.

Daniel Farrell wrote:

> Where is this calculated in the code?

src/axis.c: gen_tics and make_tics()

> How does gnuplot calculate the step size between tic marks?

Compute the length of the axis, find a tic size from the usual 1/2/5 
times power-of-ten series that will create a "useful" number of tics 
(between 5 and 10) on that range.  Variations of this scheme are needed 
for logarithmic and time/date scales.

Hi,

Thanks for the pointers. I have it all working now.

I have made an algorithm that 'links' x1 and x2 axis via a equation =20
(it actually just automate the set x2tics(...) function once the x1 =20
and a transformation equation are known.). It would be nice to =20
integrate my work with the gnuplot source when. However, I will now =20
have to learn about how gnuplot interprets equation; so that the user =20=

can link the axes, my code needs to be more general.

I'll post again about this subject after I've had a chance to read =20
some more source.

Cheers,

Daniel.

On 26 Nov 2006, at 21:36, Hans-Bernhard Br=F6ker wrote:

> Daniel Farrell wrote:
>
>> Where is this calculated in the code?
>
> src/axis.c: gen_tics and make_tics()
>
>> How does gnuplot calculate the step size between tic marks?
>
> Compute the length of the axis, find a tic size from the usual =20
> 1/2/5 times power-of-ten series that will create a "useful" number =20
> of tics (between 5 and 10) on that range.  Variations of this =20
> scheme are needed for logarithmic and time/date scales.
>
>

Hi Daniel,

The algorithm is external to all gnuplot source code at the moment.

I don't really know where to start in term of integrating it to the  
project?

I have notices that 'linking' axes is on the 'todo' list, how can I  
get involved?

I guess the command could be:
link {<known axis><linked axes>} {<equation string>}

e.g.
plot 'data.txt' using 1:2 with lines link x1x2 '$1*10'

i.e. if x1 = 0,1,2,3,4,5
x2 = 0,10,20,30,40,50

Regards,

Dan








On 27 Nov 2006, at 11:25, Daniel J Sebald wrote:

>  sourceforge web site to maintain your algorithm.

> I guess the command could be:
> link {<known axis><linked axes>} {<equation string>}

I think it should be
 	set x2range mapping 10*x+1

Have a look to the archive of this conference, I think it was discussed some 
time in the past.

---
PM

Daniel Farrell wrote:
> I have notices that 'linking' axes is on the 'todo' list, how can I  
> get involved?

Do it, post the patch on the tracker, and watch the show.

> I guess the command could be:
> link {<known axis><linked axes>} {<equation string>}

No new top-level command please.  Make it a new option to 'set xtics' 
instead.

As to the evaluation engine: look up the function "evaluate_at", and its 
supporting machinery in eval.c, internal.c, specfun.c and some others.

on 11/23/06 10:09 PM, Mojca Miklavec at mojca.miklavec.lists@...
wrote:

> Just as a warning. This is how the last Mac version of gnuplot 4.0 fails here.
> 
>> /usr/local/bin/gnuplot
> dyld: Library not loaded: /usr/local/lib/libaquaterm.1.0.0.dylib
>   Referenced from: /usr/local/bin/gnuplot
>   Reason: image not found
> Trace/BPT trap
>> locate libaquaterm
> /sw/lib/libaquaterm.1.0.0.dylib
> /sw/lib/libaquaterm.dylib
> /usr/local/lib/libaquaterm.1.0.1.dylib
> /usr/local/lib/libaquaterm.dylib
> 
> Mojca
> 
Mojca,

If you run the AquaTerm installer found via the AquaTerm website, it puts
things into /Library/Frameworks/AquaTerm.framework and /usr/local/lib. You
might try running the "official" AquaTerm-1.0.1 installer, rather than using
the older Fink version of AquaTerm.

Joe

On 11/24/06, Joe Koski wrote:
> on 11/23/06 10:09 PM, Mojca Miklavec wrote:
>
> > Just as a warning. This is how the last Mac version of gnuplot 4.0 fails here.
> >
> >> /usr/local/bin/gnuplot
> > dyld: Library not loaded: /usr/local/lib/libaquaterm.1.0.0.dylib
> >   Referenced from: /usr/local/bin/gnuplot
> >   Reason: image not found
> > Trace/BPT trap
> >> locate libaquaterm
> > /sw/lib/libaquaterm.1.0.0.dylib
> > /sw/lib/libaquaterm.dylib
> > /usr/local/lib/libaquaterm.1.0.1.dylib
> > /usr/local/lib/libaquaterm.dylib
> >
> > Mojca
> >
> Mojca,
>
> If you run the AquaTerm installer found via the AquaTerm website, it puts
> things into /Library/Frameworks/AquaTerm.framework and /usr/local/lib. You
> might try running the "official" AquaTerm-1.0.1 installer, rather than using
> the older Fink version of AquaTerm.

I have that one as well. The problem is that gnuplot has been compiled
with the older version and now it doesn't work any more. I have a
working version of gnuplot at the moment, but such situations should
be avoided in future if possible.

On 11/24/06, Per Persson wrote:
> >Just as a warning. This is how the last Mac version of gnuplot 4.0 fails here.
> >
> >/usr/local/bin/gnuplot
> >dyld: Library not loaded: /usr/local/lib/libaquaterm.1.0.0.dylib
>
> Mojca,
> this is an installer bug in AquaTerm 1.0.1
>
> Unless you are on an Intel Mac, simply use the AquaTerm 1.0.0 installer to downgrade (only difference is that 1.0.1 is a "fat" binary containing both Intel and PPC code).
>
> If you have an Intel Mac, just create a symlink:
> ln -s /Library/Frameworks/AquaTerm.framework/Versions/A/AquaTerm /usr/local/lib/libaquaterm.1.0.0.dylib

I don't think that it makes any sense to downgrade for me since I
managed to create my own binary anyway, but I'm afraid of problems
that other users might have (now or in future) if gnnuplot will remain
highly dependant on versions of libraries used on a certain computer.

I hope that the one in charge for preparing the latest binaries will
also take care about such issues.

Thanks,
Mojca

renjith padmanabha pillai wrote:

> but now i have to generate ps script file for the generated graph. the 
> labels and tics are getting overlaped if i used multiplot.
>  
> how i can set font in the case of post script.

Various ways, depending on which elements you need to do it for:

1) as an option on the 'set terminal postscript' command, to change the 
overall font

2) as a 'font' option in various commands that control text outputs, to 
change the font for those texts

3) using enhanced syntax (see 'help enhanced').

On Wednesday 22 November 2006 02:43 am, Kristian Waagan wrote:
> Hello,
>
> There's a broken link in the FAQ, under the entry
> "Does gnuplot support bar-charts/histograms/boxes?"
> (http://www.gnuplot.info/faq/faq.html#SECTION00064000000000000000).

Thanks.

In any case, histograms/bar-charts are now supported as a distinct
style type. So the link should be updated or supplemented with a
pointer to the current demo

   http://gnuplot.sourceforge.net/demo_4.2/histograms.html

> The broken link is this one:
> http://www.usf.uni-osnabrueck.de/~breiter/tools/gnuplot/barcharts.en.
>html The correct one (I think) can be obtained my removing the last
> 's' ('barcharts' -> 'barchart'):
> http://www.usf.uni-osnabrueck.de/~breiter/tools/gnuplot/barchart.en.h
>tml
>
>
>
> regards,

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