Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.graphics.apps.gnuplot > #2433 > unrolled thread

timefmt, changing the day of the week

Back to article view | Back to comp.graphics.apps.gnuplot

Contents

  timefmt, changing the day of the week Jean-Pierre Coulon <coulon@cacas.pam.obs-nice.fr> - 2014-05-02 15:10 +0200
    Re: timefmt, changing the day of the week Ethan A Merritt <EAMerritt@gmail.com> - 2014-05-02 09:15 -0700
      Re: timefmt, changing the day of the week Jean-Pierre Coulon <coulon@cacas.pam.obs-nice.fr> - 2014-05-02 19:34 +0200
        Re: timefmt, changing the day of the week Hans-Bernhard Bröker <HBBroeker@t-online.de> - 2014-05-02 22:16 +0200
          Re: timefmt, changing the day of the week Jean-Pierre Coulon <coulon@cacas.pam.obs-nice.fr> - 2014-05-03 08:16 +0200
            Re: timefmt, changing the day of the week Hans-Bernhard Bröker <HBBroeker@t-online.de> - 2014-05-03 14:48 +0200
              Re: timefmt, changing the day of the week Jean-Pierre Coulon <coulon@cacas.pam.obs-nice.fr> - 2014-05-05 08:37 +0200
                Re: timefmt, changing the day of the week Karl <mail.kfr@gmx.net> - 2014-05-05 13:16 +0200
                  Re: timefmt, changing the day of the week Jean-Pierre Coulon <coulon@cacas.pam.obs-nice.fr> - 2014-05-05 14:09 +0200
                Re: timefmt, changing the day of the week Hans-Bernhard Bröker <HBBroeker@t-online.de> - 2014-05-05 18:49 +0200
                  Re: timefmt, changing the day of the week Jean-Pierre Coulon <coulon@cacas.pam.obs-nice.fr> - 2014-05-06 07:39 +0200
    Re: timefmt, changing the day of the week Hans-Bernhard Bröker <HBBroeker@t-online.de> - 2014-05-02 18:22 +0200

#2433 — timefmt, changing the day of the week

I am plotting an array with x in universal time. I know that 1-Jan-2004 
was TU=1072569615. and I am plotting using ($1-1072569615.) I said: set 
format x "%a-%d-%b\n%H:%M:%Y". I obtain the right month and day-of-the 
month on the X axis scale, but day-of-the-week is false . I guess 
Gnuplot guesses that 1-Jan-2004 was a Monday or a Sunday. How can I force 
day-of-the-week to start on the right day?

Jean-Pierre Coulon

[toc] | [next] | [standalone]

#2434

Jean-Pierre Coulon wrote:

> I am plotting an array with x in universal time. I know that
> 1-Jan-2004 was TU=1072569615. and I am plotting using ($1-1072569615.)
> I said: set format x "%a-%d-%b\n%H:%M:%Y". I obtain the right month
> and day-of-the month on the X axis scale, but day-of-the-week is false
> . I guess Gnuplot guesses that 1-Jan-2004 was a Monday or a Sunday.
> How can I force day-of-the-week to start on the right day?
> 
> Jean-Pierre Coulon

Your time offset from the epoch in seconds is not correct.

bash-4.2$  TZ=UT date --date='@1072569615' 
Sun Dec 28 00:00:15  2003

bash-4.2$  TZ=UT date --date='@1072915201' 
Thu Jan  1 00:00:01  2004

Gnuplot agrees:

gnuplot> print strftime("%a %d-%b-%Y", 1072915201.)
Thu 01-Jan-2004

Note: This is shown for the development version of gnuplot, in which
the standard epoch 1-Jan-1970 is used.  Earlier gnuplot versions used
an idiomatic epoch 1-Jan-2000 for which the offset is 946684800.
So for gnuplot_4.6:

gnuplot> EPOCH = 946684800.0
gnuplot> print strftime("%a %d-%B-%Y", 1072915201. - EPOCH  )
Thu 01-January-2004

So far as I know, gnuplot does correctly assign days of the week.

[toc] | [prev] | [next] | [standalone]

#2436

On Fri, 2 May 2014, Ethan A Merritt wrote:

> Your time offset from the epoch in seconds is not correct.
>
> bash-4.2$  TZ=UT date --date='@1072569615'
> Sun Dec 28 00:00:15  2003

Then this is what I had feared: Gnuplot and the application that wrote my 
input file do not adhere to the same UT school! If I simply say plot 
<file> using ($1):2 I am brought on Tue-25-April 2034! I'm sure my 
experiment took place Wed-30-April 2014. So at least counting from 
2004-01-01 gives the correct day-of-the month and month. But showing the 
right day-of-the-week would be a plus!

So there are 2 possiblilties:

1: I cause Gnuplot to adhere to the same UT school as may application. If 
there are many schools I hope I'll find the right one by trial an error!

2: I "cheat" by "forcing" the 1rst day-of-the week.

Jean-Pierre Coulon

[toc] | [prev] | [next] | [standalone]

#2437

On 02.05.2014 19:34, Jean-Pierre Coulon wrote:

> If I simply say plot
> <file> using ($1):2 I am brought on Tue-25-April 2034!

> I'm sure my
> experiment took place Wed-30-April 2014.

I'm pretty sure one of those statements is off by a decade.

 > So at least counting from
> 2004-01-01 gives the correct day-of-the month and month.  But showing the
> right day-of-the-week would be a plus!

Well, of course it does!  You've beamed your data into entirely the 
wrong year, Sherlock!

> 1: I cause Gnuplot to adhere to the same UT school as may application.

No need to.  All you have to do is to actually _tell_ it that your data 
is in Unix epoch time.  Instead you're throwing random offsets into the 
mix and expecting miracles.

[toc] | [prev] | [next] | [standalone]

#2438

On Fri, 2 May 2014, Hans-Bernhard Bröker wrote:

>>  If I simply say plot
>>  <file> using ($1):2 I am brought on Tue-25-April 2034!
>
>>  I'm sure my
>>  experiment took place Wed-30-April 2014.
>
> I'm pretty sure one of those statements is off by a decade.

My data file goes from 1082875699.5 to 1082879698.5. Is there an "offical" 
UT site providing a [UT to "common language" date and time] conversion 
after you type a UT into a dialog box?

Then I'll make a complaint to developers of the application that provided 
this data last week.

Is set timefmt "%s" correct? And plot using (timecolumn(1)):2 brings me in 
2004!

Jean-Pierre Coulon

[toc] | [prev] | [next] | [standalone]

#2439

On 03.05.2014 08:16, Jean-Pierre Coulon wrote:

> My data file goes from 1082875699.5 to 1082879698.5. Is there an
> "offical" UT site providing a [UT to "common language" date and time]
> conversion after you type a UT into a dialog box?

There can't be, because is no such thing as an "official UT" like the 
data you have, in the first place.  There's UTC, but it's not stated in 
seconds like you're looking at: it's given in date, hours, minutes and 
seconds.

Then there's conventional Unix/POSIX timestamps, which are measured in 
seconds since 1970-01-01, but you're not using that either, which is why 
your use of gnuplot "%s" format threw you into the wrong decade.

Your data actually appears to be in seconds since 1980-01-05, which 
almost certainly means those are GPS timestamps.  Calling that "UT" 
without any further description is indeed a misrepresentation of facts.

[toc] | [prev] | [next] | [standalone]

#2453

On Sat, 3 May 2014, Hans-Bernhard Bröker wrote:

> Your data actually appears to be in seconds since 1980-01-05, which almost 
> certainly means those are GPS timestamps.  Calling that "UT" without any 
> further description is indeed a misrepresentation of facts.

Right. Sat 5-Jan-1980, 23:59:45 to be precise. But how can I tell this to 
Gnuplot?

I can obtain the correct day-of-the-month and month my saying plot using 
($1-<the number correponding to 1-Jan-2014>) but then day-of-the-week is 
wrong. Can I adjust the day of the first point of my plot? Or can I cause 
"using ($1-something)" to act on the day-of-the-week only?

-- 
Jean-Pierre Coulon

[toc] | [prev] | [next] | [standalone]

#2454

Am 05.05.2014 08:37, schrieb Jean-Pierre Coulon:
> Right. Sat 5-Jan-1980, 23:59:45 to be precise. But how can I tell this
> to Gnuplot?
> 
> I can obtain the correct day-of-the-month and month my saying plot using
> ($1-<the number correponding to 1-Jan-2014>) but then day-of-the-week is

Why 2014? You have to substract the number of seconds between tzero in
gnuplot (1Jan2000) and tzero of your GPS time. The way you´re doing it,
you don´t get the wrong day-of-week, you´re in the wrong year!

> wrong. Can I adjust the day of the first point of my plot? Or can I
> cause "using ($1-something)" to act on the day-of-the-week only?

Gnuplot just calculates the day-of-week from the date, there´s no way to
influence that. If you´re still convinced it does that wrongly, can you
post the gnuplot script and a short snippet from your data? Best use
inline data, i.e. something like

  plot "-" using (whatever)
  123012938 23.4
  123040033 45.5
  123084232 12.3
  e

that way it´s easiest for others to reproduce your problem.

 Karl

[toc] | [prev] | [next] | [standalone]

#2456

On Mon, 5 May 2014, Karl wrote:

> Why 2014? You have to substract the number of seconds between tzero in
> gnuplot (1Jan2000) and tzero of your GPS time. The way you´re doing it,
> you don´t get the wrong day-of-week, you´re in the wrong year!

Works fine. Thanks.

Jean-Pierre Coulon

[toc] | [prev] | [next] | [standalone]

#2458

On 05.05.2014 08:37, Jean-Pierre Coulon wrote:
> On Sat, 3 May 2014, Hans-Bernhard Bröker wrote:
>
>> Your data actually appears to be in seconds since 1980-01-05, which
>> almost certainly means those are GPS timestamps.  Calling that "UT"
>> without any further description is indeed a misrepresentation of facts.
>
> Right. Sat 5-Jan-1980, 23:59:45 to be precise. But how can I tell this
> to Gnuplot?

GPS_offset = strptime('%Y-%m-%dT%H:%M:%S','1980-01-05T23:59:45')

plot 'data' u ($1+GPS_Offset)

[toc] | [prev] | [next] | [standalone]

#2460

On Mon, 5 May 2014, Hans-Bernhard Bröker wrote:

> GPS_offset = strptime('%Y-%m-%dT%H:%M:%S','1980-01-05T23:59:45')
>
> plot 'data' u ($1+GPS_Offset)

Perfect! Thanks.
-- 
Jean-Pierre Coulon

[toc] | [prev] | [next] | [standalone]

#2435

On 02.05.2014 15:10, Jean-Pierre Coulon wrote:

> I am plotting an array with x in universal time. I know that 1-Jan-2004
> was TU=1072569615. and I am plotting using ($1-1072569615.)

On what basis did you decide that 2004-01-01 had any relevance to this 
task?  If x is indeed UTC, you could just

	set xdata time
	set timefmt x "%s"

in current versions of gnuplot.

>I said: set
> format x "%a-%d-%b\n%H:%M:%Y". I obtain the right month and day-of-the
> month on the X axis scale, but day-of-the-week is false .

So what about the year?

> I guess
> Gnuplot guesses that 1-Jan-2004 was a Monday or a Sunday.

On what basis?

[toc] | [prev] | [standalone]

Back to top | Article view | comp.graphics.apps.gnuplot

csiph-web