gnuplot-info Mailing List for gnuplot (Page 11)

Dave Horsfall wrote:

>> But actually IRC isn't dead at all, there are currently
>> 43 979 users globally (on 22 529 channels) on Libera.chat -
>> use the /motd command - Libera, which today is the IRC
>> network of the FOSS world.
>
> Thanks for all that info! I might give it a go.

To Emacs users and I'm sure they are quite a few here I can
recommend `erc' (for ERC, Emacs IRC) if anyone is unaware
of it.

Also built-in in vanilla GNU Emacs is `rcirc' which you get
with M-x irc RET. I have no experience from it but my friends
on gmane.emacs.help says it's good, and I believe them,
however ERC seems more common these days.

Here is a screenshot from ERC where a user is executing a bot
command to show ASCII art :) from the 90s, right?

  https://dataswamp.org/~incal/figures/fireworks-over-toronto.png

On Thu, 30 Jun 2022, Emanuel Berg wrote:

[...]

> But actually IRC isn't dead at all, there are currently 43 979 users 
> globally (on 22 529 channels) on Libera.chat - use the /motd command - 
> Libera, which today is the IRC network of the FOSS world.

Thanks for all that info!  I might give it a go.

-- Dave

Dave Horsfall wrote:

>> I think this is a mistyping of IRC - Internet Relay Chat,
>> a chat tool that was used extensively by Noah but not so
>> much since the 17th century ?. Hence the "Ha-ha" comment
>> it received.
>
> Indeed

But actually IRC isn't dead at all, there are currently
43 979 users globally (on 22 529 channels) on Libera.chat
- use the /motd command - Libera, which today is the IRC
network of the FOSS world.

There are just 16 users on #gnuplot tho.

The most 10 most popular channels are 

  2078 #linux 
  1820 #libera 
  1776 #fedora 
  1606 #python                
  1491 #kde_ru                
  1401 #archlinux             
  1394 #archlinux-cn          
  1204 #archlinux-cn-offtopic 
  1002 #neovim                
   907 #ansible                

> unless the messages are archived then it's useless.

But that's super easy! The IRC clients can do logging so you
just do that and then rsync the logs to the web pile.

https://libera.chat/

-- 
underground experts united
https://dataswamp.org/~incal

On Wed, 29 Jun 2022, Peter Rockett via gnuplot-info wrote:

> I think the idea of a single mailing list probably reflects gnuplot's 
> age. Other projects I follow tend to have forums with multiple 
> categories like development, using blah-blah, and so on.

Only when they reach a certain threshold of topic divergence.

> FWIW: I think this mailing list is actually pretty quiet. This flurry of 
> activity on 12M or whatever is the first for quite some while.

It's about the quietest list I'm on (and I'm on lots); it's just that it 
wakes up every now and then, and should a newbie then join...

> I think this is a mistyping of IRC - Internet Relay Chat, a chat tool 
> that was used extensively by Noah but not so much since the 17th century 
> ?. Hence the "Ha-ha" comment it received.

Indeed; unless the messages are archived then it's useless.

-- Dave

On 29/06/2022 15:12, Jeremy Nicoll - ml gnuplot wrote:
> On 2022-06-28 16:07, Glenn Golden wrote:
>
>> Imo, yes.  I would like to respectfully suggest that this ongoing
>> tutorial discussion be relocated ...
>
> This sort of thing is normal on every other -users mail-list I'm
> on.
I think the idea of a single mailing list probably reflects gnuplot's 
age. Other projects I follow tend to have forums with multiple 
categories like development, using blah-blah, and so on.
>
>
>> It's pretty noisy and does not seem to contain much of general
>> interest that can't be glened from the docs.
FWIW: I think this mailing list is actually pretty quiet. This flurry of 
activity on 12M or whatever is the first for quite some while.
>
> As a newcomer to gnuplot, I'm finding it useful, because it gives
> me concrete things to try, to help me to understand gnuplot better.
>
> If no new user ever posts here, what sort of thing would you expect
> to see?
>
>
> Is IRQ (which I think means "interrupt request" - do you mean some
> sort of chat thing?) searchable in the same way that mail list
> archives tend to be?  If not, it's no use to anyone other than the
> participants at the instant in which a discussion takes place.

I think this is a mistyping of IRC - Internet Relay Chat, a chat tool 
that was used extensively by Noah but not so much since the 17th century 
😁. Hence the "Ha-ha" comment it received.

Peter

On 2022-06-28 16:07, Glenn Golden wrote:

> Imo, yes.  I would like to respectfully suggest that this ongoing
> tutorial discussion be relocated ...

This sort of thing is normal on every other -users mail-list I'm
on.


> It's pretty noisy and does not seem to contain much of general
> interest that can't be glened from the docs.

As a newcomer to gnuplot, I'm finding it useful, because it gives
me concrete things to try, to help me to understand gnuplot better.

If no new user ever posts here, what sort of thing would you expect
to see?


Is IRQ (which I think means "interrupt request" - do you mean some
sort of chat thing?) searchable in the same way that mail list
archives tend to be?  If not, it's no use to anyone other than the
participants at the instant in which a discussion takes place.

-- 
Jeremy Nicoll - my opinions are my own

Glenn Golden wrote:

> Imo, yes. I would like to respectfully suggest that this
> ongoing tutorial discussion be relocated to IRQ or

IRQ haha :)

> It's pretty noisy and does not seem to contain much of
> general interest [...]

This is the user ML or am I mistaken?
gmane.comp.graphics.gnuplot.user

-- 
underground experts united
https://dataswamp.org/~incal

On Tue, Jun 28, 2022, at 08:52, Emanuel Berg wrote:
> Balick, Rebekah C [US] (ES) wrote:
>
>> Can whomever is in charge of this email list take me off of
>> it? I have unsubscribed twice and followed the instructions
>> but I am still receiving emails. Please remove me from the
>> list. Thank you!
>
> ... was it _that_ bad?
>

Imo, yes.  I would like to respectfully suggest that this ongoing 
tutorial discussion be relocated to IRQ or elsewhere. It's pretty
noisy and does not seem to contain much of general interest that
can't be glened from the docs.

Just my 2c, not trying to start a fight. 

Balick, Rebekah C [US] (ES) wrote:

> Can whomever is in charge of this email list take me off of
> it? I have unsubscribed twice and followed the instructions
> but I am still receiving emails. Please remove me from the
> list. Thank you!

... was it _that_ bad?

-- 
underground experts united
https://dataswamp.org/~incal

Can whomever is in charge of this email list take me off of it? I have unsubscribed twice and followed the instructions but I am still receiving emails. Please remove me from the list. Thank you!

-----Original Message-----
From: Emanuel Berg <in...@da...> 
Sent: Monday, June 27, 2022 9:34 PM
To: gnu...@li...
Subject: EXT :Re: [Gnuplot-info] 12M? or 12 000 000? (not 12000000)

> I just read of columnhead, however that's still hard-coded (the number 
> of lines)

array Cols = ["cyan", "magenta", "yellow", "green", "white", "red"]

data="/home/incal/public_html/hits/hits.dat"
plot for [i=1:*] \
 data u ($0+1):i w l lc Cols[(i % 6) + 1] lw 2 t columnhead

OK, found the for loop (but the color array doesn't work).

If I remove the sort in the source file and have the program add the columnhead I think that will work (be automated) unless users are removed or added, right?

--
underground experts united
https://dataswamp.org/~incal



_______________________________________________
gnuplot-info mailing list
gnu...@li...
Membership management via: https://lists.sourceforge.net/lists/listinfo/gnuplot-info

> I just read of columnhead, however that's still hard-coded
> (the number of lines)

array Cols = ["cyan", "magenta", "yellow", "green", "white", "red"]

data="/home/incal/public_html/hits/hits.dat"
plot for [i=1:*] \
 data u ($0+1):i w l lc Cols[(i % 6) + 1] lw 2 t columnhead

OK, found the for loop (but the color array doesn't work).

If I remove the sort in the source file and have the program
add the columnhead I think that will work (be automated)
unless users are removed or added, right?

-- 
underground experts united
https://dataswamp.org/~incal

> plot \
>  data u ($0+1):1 w l lc "cyan"    lw 2 t "solene",   \
>  data u ($0+1):2 w l lc "magenta" lw 2 t "incal",    \
>  data u ($0+1):3 w l lc "yellow"  lw 2 t "lich",     \
>  data u ($0+1):4 w l lc "green"   lw 2 t "josk",     \
>  data u ($0+1):5 w l lc "white"   lw 2 t "rjc",      \
>  data u ($0+1):6 w l lc "red"     lw 2 t "katzeilla"

I just read of columnhead, however that's still hard-coded
(the number of lines)

data="/home/incal/public_html/hits/hits.dat"
plot \
 data u ($0+1):1 w l lc "cyan"    lw 2 t columnhead, \
 data u ($0+1):2 w l lc "magenta" lw 2 t columnhead, \
 data u ($0+1):3 w l lc "yellow"  lw 2 t columnhead, \
 data u ($0+1):4 w l lc "green"   lw 2 t columnhead, \
 data u ($0+1):5 w l lc "white"   lw 2 t columnhead, \
 data u ($0+1):6 w l lc "red"     lw 2 t columnhead

I realize I have sorted the data so the columnhead solution
doesn't work if one dude passes the other. So right, I'll
unsort that but ... I don't see how it will work, if one user
is added or deleted for example? I guess tagging the
individual data won't work. Maybe put it in a column of
its own and refer to it explicitly?

-- 
underground experts united
https://dataswamp.org/~incal

> The only thing that isn't automated [...] are the user
> names, those are known tho (from the log file names) so they
> just need to be passed to gnuplot - maybe the columns in the
> data files can be assigned names (a multi-column header)?

So the data for day x is now at line x in the .dat file, e.g.
day 147 at line 147 says

  11814601 737184 610310 13198 4223 1885

Can I maybe tag these data, like this

  solene:11814601 incal:737184 lich:610310 josk:13198 rjc:4223 katzeilla:1885

or in some other way so that gnuplot will associate the graph
with that name?

So I don't have to do this hard-coding of the names?

data="/home/incal/public_html/hits/hits.dat"
plot \
 data u ($0+1):1 w l lc "cyan"    lw 2 t "solene",   \
 data u ($0+1):2 w l lc "magenta" lw 2 t "incal",    \
 data u ($0+1):3 w l lc "yellow"  lw 2 t "lich",     \
 data u ($0+1):4 w l lc "green"   lw 2 t "josk",     \
 data u ($0+1):5 w l lc "white"   lw 2 t "rjc",      \
 data u ($0+1):6 w l lc "red"     lw 2 t "katzeilla"

?

The hard-coding isn't just bad because it is hard-coding in
general, which is bad (almost always), here it is also the
last step that isn't automated ... 

https://dataswamp.org/~incal/hits/hits.gpi
https://dataswamp.org/~incal/hits/hits.png

-- 
underground experts united
https://dataswamp.org/~incal

Norwid Behrnd via gnuplot-info wrote:

>> It should start either at 2022-01-10 and then +7 days for
>> every xtick, or it can start with 148 also with +7 days
>> per step.
>
> From this I bet your interest is the automated analysis of
> time series with gnuplot [...]
> 
> Ben Hoey's example is one both to easy to decipher and
> suitable as point of departure about how to interact from
> bash with gnuplot.

It is automated, the way you put it it seems more advanced
than it is tho ...

What happens is a zsh script [1] is invoked from cron once
every day, the script goes on to call Emacs which uses an
Elisp program [2] to fetch data from the logs already present
in the system - files named

  /var/www/logs/access-${user}.log

where one line equals one HTTP request or one hit for
our purposes.

All in all the zsh script including the Elisp extracts and
appends data to the data file [3] and then the gnuplot
commands [4] execute on that.

The result I've shown you many times by now ... thanks for all
the help. I think it looks cool!

The only thing that isn't automated IINM are the user names,
those are known tho (from the log file names) so they just
need to be passed to gnuplot - maybe the columns in the data
files can be assigned names (a multi-column header)?

  https://dataswamp.org/~incal/hits/hits.png

[1] https://dataswamp.org/~incal/hits/piles-all
[2] https://dataswamp.org/~incal/emacs-init/piles.el
[3] https://dataswamp.org/~incal/hits/ooa/hits.dat
[4] https://dataswamp.org/~incal/hits/hits.gpi

-- 
underground experts united
https://dataswamp.org/~incal

theozh wrote:

> if you give only one column for a line plot it will use the
> pseudocolumn 0 as x-value (check "help pseudocolumns") which
> is basically the line number (index zero-based).

Understood, thanks!

> So, instead use:
>
> data="/home/incal/public_html/hits/hits.dat"
>  plot \
>    data u ($0+1):1 w l lc "cyan"    lw 2 t "solene",   ...

You are right, that did it!

Now everything is correct, the days, the weeks (start at
Mondays).

The only thing that is strange is it appears to be a week
starting at day 147 if you look at the image?

  https://dataswamp.org/~incal/hits/hits.png

Source: https://dataswamp.org/~incal/hits/hits.gpi

-- 
underground experts united
https://dataswamp.org/~incal

if you give only one column for a line plot it will use the pseudocolumn 0 as x-value (check "help pseudocolumns") which is basically the line number (index zero-based).
So, instead use:

data="/home/incal/public_html/hits/hits.dat"
  plot \
    data u ($0+1):1 w l lc "cyan"    lw 2 t "solene",   ...

Now it's almost good except it still counts from 0, so day 170
is actually day 171. I don't know how that can be changed, I'm
tempted to add a bogus line to the data ... but no. Don't know
what else to do tho?

Take a look:
  https://dataswamp.org/~incal/hits/hits.png

TIA

#! /usr/local/bin/gnuplot
#
# this file:
#   https://dataswamp.org/~incal/hits/hits.gpi
#
# the graph:
#   https://dataswamp.org/~incal/hits/hits.png

reset

set term png background rgb "black"
set out "/home/incal/public_html/hits/hits.png"

set tics font "verdana,10"

set format y "%.0s%c"
set logscale y
set yrange[1000:1e9]

set format x "%s" timedate
set xtics 2,7
set xrange [146:] noextend

set grid   lc "yellow"
set border lc "gray"

set key outside top right spacing 1.4 tc "white" font "verdana,10" horizontal

data="/home/incal/public_html/hits/hits.dat"
plot \
 data u 1 w l lc "cyan"    lw 2 t "solene",   \
 data u 2 w l lc "magenta" lw 2 t "incal",    \
 data u 3 w l lc "yellow"  lw 2 t "lich",     \
 data u 4 w l lc "green"   lw 2 t "josk",     \
 data u 5 w l lc "white"   lw 2 t "rjc",      \
 data u 6 w l lc "red"     lw 2 t "katzeilla"

-- 
underground experts united
https://dataswamp.org/~incal

@Emanuel and future readers interested in time-series analysis with
gnuplot :

Two days ago, the discussion on fortran-lang about the continuous
visualization of computational results[1] was amended by a link to
the presentation «Realtime display with Gnuplot» by Ge Baolai
(2022/06/15).  As one of the online seminars hosted by the Canadian
sharcnet,[2] three approaches how to address such a task are
presented and compared with each other.  So I think this would fit
well here, too.  To quote some of the abstract:[3]

«In this talk we will discuss the technique for displaying output in
real time with Gnuplot during number crunching and data processing.
Often people might want to see the results while computing to
determine if the program runs normally or to verify if the output
results make sense.  [...]  We will show in this seminar how to use
Gnuplot to achieve the goal of displaying output data in real time
without calling external functions explicitly. The targeted audience
would be those who use general purpose programming languages such as
C/C++, Fortran, shell scripting etc.»

In addition to the video,[4] the presenter's slides[5] are shared.

[1] «How to use Fortran + gnuplot to plot in real time in just one
window?», root entry
https://fortran-lang.discourse.group/t/how-to-use-fortran-gnuplot-to-plot-in-real-time-in-just-one-window/2904
[2] https://helpwiki.sharcnet.ca/wiki/Online_Seminars
[3]
https://helpwiki.sharcnet.ca/wiki/Webinar_2022_Realtime_display_with_Gnuplot
[4] https://youtu.be/A9YPibAEXEE
[5] https://helpwiki.sharcnet.ca/wiki/images/e/ea/Dirt.pdf

On Friday, 24 June 2022 01:30:25 PDT Emanuel Berg wrote:
> 
> But how will one get infamation on the x-axis?
> 
> It should start either at 2022-01-10 and then +7 days for
> every xtick, or it can start with 148 also with +7 days
> per step.
> 
> In the shell (shell-mode) it works using
> 'date +"%Y-%m-%d"' or 'date +"%j"' if that helps ...
> 
> It would be better to automate it than hard code it but
> actually hard-code would also be an improvement at this stage.

I draw your attention to these possibly relevant changes to gnuplot
from last year.  These are in the development code but not in the
current version 5.4 release.    Is there interest in having these changes
back-ported to 5.4?

commit a8a8a88d77a77f3af047f5b4d0444737a6837c67 
Author: Ethan A Merritt <merritt@u.washington.edu> 
Date:   Mon Jun 14 15:03:25 2021 -0700 

   "set mxtics time <N> <units" 
    
   Allow explicit choice of minitic placement along time axis. 
   For example 
       set mxtics time 1 month 
   places a minitic exactly on the 1st day of Feb/Mar/Apr even 
   though this does not correspond to equal intervals measured in days.
commit 529062ae873901010d138e68897d8f6d057a8e7a 
Author: Ethan A Merritt <merritt@u.washington.edu> 
Date:   Tue Jun 15 22:19:08 2021 -0700 

   Revise auto-placement of tics for time data 
    
   For timescales of days or longer, if a ticmark is generated for 
   some date then place it at the start of that same day (0 AM). 
   The spacing between "day" tics is always exactly N days. 
    
   It used to be that the date was rounded up if time > 22:00 but 
   hours/minutes/seconds were left unchanged. This meant that the 
   tic positions were not exactly on a day boundary and might not 
   even be on the same day. Thus the spacing between "dates" was not 
   an integral number of days. 
    
   This discrepancy was particularly obvious when using the new 
   option "set minitics time N days", since the evenly spaced 
   minitics did not align with the major tics.

> 
> But how will one get infamation on the x-axis?
> 
> It should start either at 2022-01-10 and then +7 days for
> every xtick, or it can start with 148 also with +7 days
> per step.
> 

From this I bet your interest is the automated analysis of time series with
gnuplot.  Did you visit the gallery of examples on gnuplot.info, in particular
about axis transformation?[1]

Ben Hoey's example[2] is one both to easy to decipher and suitable as point of
departure about how to interact from bash with gnuplot.  (Your mileage may vary
for other OSes like Windows or Mac.)


For one, he stores the log data in a space separated file named `mydata.txt`:

```
2019-02-01  15
2019-03-01   8
2019-04-01  16
2019-05-01  20
2019-06-01  14
```


For two, the pattern to be applied is defined in a gnuplot script called to
action from the CLI including arguments.  It isn't like Python's argparse where
you have to define and remember the flag's keys, but their counting and
sequence: ARG0 refers to the name of the gnuplot script itself.  ARG1, ARG2,
etc. are for subsequent adjustments to affect gnuplot's work.

Already modified for demonstration, Ben's script I store as `s1.spt` in the same
folder as `mydata.txt` is the following (lines' length limit is 80 chars/line):

```
set xdata time                 # Indicate that x-axis values are time values
set timefmt "%Y-%m-%d"         # Indicate the pattern the time values will be in
# set format x "%m-%y"         # Set how the dates will be displayed on the plot

# set xrange ["2019-01-01":"2019-12-31"]   # Set x-axis range of values

set yrange [0:30]              # Set y-axis range of values
set key off                    # Turn off graph legend
set xtics rotate by -45       # Rotate dates on x-axis 45deg for cleaner display
set title 'Squirrels Spotted'  # Set graph title

# set terminal jpeg            # Set the output format to jpeg
# set output 'output.jpg'      # Set output file to output.jpg

# introduced modifications, start:
set xrange [ARG1:ARG2]         # listen to arguments by the CLI
set format x "%Y-%m"           # coarse-to-fine time format on abscissa

set grid
set terminal pngcairo
set output "test.png"
# introduced modifications, end.

plot 'mydata.txt' using 1:2 with linespoints linetype 6 linewidth 2
```


Three, gnuplot is called from the CLI with `-c` (like /collect/ the arguments)
in a pattern like

```
gnuplot -c s1.spt 2019-01-01 2019-12-31
gnuplot -c s1.spt 2019-03-14 2019-07-14
```

with ARG1 defining the start, and ARG2 the end of the canvas.  In the present
example, the two have to be provided in the pattern YYYY-MM-DD.

If the data contain more than two columns, either adjust `using` (short `u`) to
indicate gnuplot the independent and dependent variable(s), or reorganize the
log file submitted to gnuplot (e.g., AWK).  On the other hand, it looks like the
major bits and bolts now are at hand to automate your report generation.

[1] http://gnuplot.sourceforge.net/demo_5.4/timedat.html
[2] https://bhoey.com/blog/simple-time-series-graphs-with-gnuplot/

Norwid Behrnd via gnuplot-info wrote:

> For two, the user defined additional tic only was set to
> show that this possibility (still) is at your disposition.
> E.g., to mark a critical threshold like in control charts
> (analytical chemistry / chemical engineering), adding
> lower/upper control levels before a process runs away. Thus,
> it was an alternative to an one-liner like

Right, I would have expected such a feature to exist and now
you have showed it beyond doubt :)

But how will one get infamation on the x-axis?

It should start either at 2022-01-10 and then +7 days for
every xtick, or it can start with 148 also with +7 days
per step.

In the shell (shell-mode) it works using
'date +"%Y-%m-%d"' or 'date +"%j"' if that helps ...

It would be better to automate it than hard code it but
actually hard-code would also be an improvement at this stage.

https://dataswamp.org/~incal/hits/hits.gpi
https://dataswamp.org/~incal/hits/hits.png

-- 
underground experts united
https://dataswamp.org/~incal

Emanuel Berg wrote:

> 
> > set ytics font "verdana,10" add \
> > ("25 M"   2.5e7)  # only shown for the ordinate scaling by lg  
> 
> Looks even better without it?
> 

For one, I agree with you there are situations where indeed it does
look better without it.

For two, the user defined additional tic only was set to show that
this possibility (still) is at your disposition.  E.g., to mark a
critical threshold like in control charts (analytical chemistry /
chemical engineering), adding lower/upper control levels before a
process runs away.  Thus, it was an alternative to an one-liner like

```
set arrow from graph 0,first 2.5e7 to graph 1, first 2.5e7 nohead lt
0 lc "red"
```

as in «our infrastructure reliably processes up to 25M events/s».

Norwid Behrnd via gnuplot-info wrote:

> set ytics font "verdana,10" add \
> ("25 M"   2.5e7)  # only shown for the ordinate scaling by lg

Looks even better without it?

  https://dataswamp.org/~incal/hits/hits.gpi
  https://dataswamp.org/~incal/hits/hits.png

-- 
underground experts united
https://dataswamp.org/~incal

Norwid Behrnd via gnuplot-info wrote:

> + gnuplot actually may add SI prefixes (k, M, etc.) on the
>   scales for you and show user-defined tics. For this to
>   happen, drop "set format y "%.0f" to try instead ``` set
>   format y '%.0s %c'
>
> set ytics font "verdana,10" add \
> ("25 M"   2.5e7)  # only shown for the ordinate scaling by lg
>
> set logscale y  # i.e., the implicit form of "set logscale y 10"

Oh, cool!

> Perhaps altogether with an instruction like "set
> yrange[1000:1e9]"

Done.

> this may justify to move/reorganize the keys a little bit.

https://dataswamp.org/~incal/hits/hits.png B)

-- 
underground experts united
https://dataswamp.org/~incal

theozh wrote:

> Please read `help colorspec` carefully.
>
> This would be one valid option of many options:
>
> green = "#00009600"
> set title font "verdana,12" tc rgbcolor green [...]

OK, but I thought it worked with just 'tc "#00009600"' so
I don't understand why rgbcolor is needed all of a sudden?
Maybe it didn't work, actually.

But anyway I decided to stick with the default/built-in colors
so now it looks like this:

  https://dataswamp.org/~incal/hits/hits.png

Thank anyway ...

-- 
underground experts united
https://dataswamp.org/~incal