Thread: [Gnuplot-py-users] Resolution of Plots

I have a file of x,y real data pairs with the values to 9 digits of precision. When I use gnuplot to plot the data directly using, say:
plot 'HHAT3260.csv' using 2:1
I get a detailed plot with the values at their original precision. However, when I use gnuplot within Python (having read in the file to float lists x and y) using:
g = Gnuplot.Gnuplot(debug=1)
d = Gnuplot.Data(x, y, with_='points')
g.plot(d)
I get a plot with plotted values rounded or truncated to 7 digits of precision, it seems. Is there a way to preserve the precision of the data when using py-gnuplot?
-- Richard Langley

-----------------------------------------------------------------------------
| Richard B. Langley                            E-mail: la...@un...         |
| Geodetic Research Laboratory                  Web: http://gge.unb.ca      |
| Dept. of Geodesy and Geomatics Engineering    Phone:    +1 506 453-5142   |
| University of New Brunswick                   Fax:      +1 506 453-4943   |
| Fredericton, N.B., Canada  E3B 5A3                                        |
|        Fredericton?  Where's that?  See: http://www.fredericton.ca/       |
-----------------------------------------------------------------------------

On 12/08/2013 02:49 PM, Richard Langley wrote:
> I have a file of x,y real data pairs with the values to 9 digits of
> precision. When I use gnuplot to plot the data directly using, say:
> plot 'HHAT3260.csv' using 2:1
> I get a detailed plot with the values at their original precision.
> However, when I use gnuplot within Python (having read in the file to
> float lists x and y) using:
> g = Gnuplot.Gnuplot(debug=1)
> d = Gnuplot.Data(x, y, with_='points')
> g.plot(d)
> I get a plot with plotted values rounded or truncated to 7 digits of
> precision, it seems. Is there a way to preserve the precision of the
> data when using py-gnuplot?

Gnuplot.py's default is to treat data as 32-bit floating point numbers.
 But if you explicitly pass double-precision data to Gnuplot.Data(),
then I think it will pass the data to gnuplot with the higher precision.
 I.e., do something like

    d = Gnuplot.Data(
        numpy.array(x, dtype=numpy.float64),
        numpy.array(y, dtype=numpy.float64),
        with_='points',
        )

Michael

-- 
Michael Haggerty
mh...@al...
http://softwareswirl.blogspot.com/

Thanks, Michael, but, sadly, that did not work.
-- Richard

On Monday, December 9, 2013,343, at 3:47 AM, Michael Haggerty wrote:

> On 12/08/2013 02:49 PM, Richard Langley wrote:
>> I have a file of x,y real data pairs with the values to 9 digits of
>> precision. When I use gnuplot to plot the data directly using, say:
>> plot 'HHAT3260.csv' using 2:1
>> I get a detailed plot with the values at their original precision.
>> However, when I use gnuplot within Python (having read in the file to
>> float lists x and y) using:
>> g = Gnuplot.Gnuplot(debug=1)
>> d = Gnuplot.Data(x, y, with_='points')
>> g.plot(d)
>> I get a plot with plotted values rounded or truncated to 7 digits of
>> precision, it seems. Is there a way to preserve the precision of the
>> data when using py-gnuplot?
> 
> Gnuplot.py's default is to treat data as 32-bit floating point numbers.
> But if you explicitly pass double-precision data to Gnuplot.Data(),
> then I think it will pass the data to gnuplot with the higher precision.
> I.e., do something like
> 
>    d = Gnuplot.Data(
>        numpy.array(x, dtype=numpy.float64),
>        numpy.array(y, dtype=numpy.float64),
>        with_='points',
>        )
> 
> Michael
> 
> -- 
> Michael Haggerty
> mh...@al...
> http://softwareswirl.blogspot.com/

-----------------------------------------------------------------------------
| Richard B. Langley                            E-mail: la...@un...         |
| Geodetic Research Laboratory                  Web: http://gge.unb.ca/     |
| Dept. of Geodesy and Geomatics Engineering    Phone:    +1 506 453-5142   |
| University of New Brunswick                   Fax:      +1 506 453-4943   |
| Fredericton, N.B., Canada  E3B 5A3                                        |
|        Fredericton?  Where's that?  See: http://www.fredericton.ca/       |
-----------------------------------------------------------------------------

On 12/09/2013 03:40 PM, Richard B. Langley wrote:
> On Monday, December 9, 2013,343, at 3:47 AM, Michael Haggerty wrote:
>> On 12/08/2013 02:49 PM, Richard Langley wrote:
>>> I have a file of x,y real data pairs with the values to 9 digits of
>>> precision. When I use gnuplot to plot the data directly using, say:
>>> plot 'HHAT3260.csv' using 2:1
>>> I get a detailed plot with the values at their original precision.
>>> However, when I use gnuplot within Python (having read in the file to
>>> float lists x and y) using:
>>> g = Gnuplot.Gnuplot(debug=1)
>>> d = Gnuplot.Data(x, y, with_='points')
>>> g.plot(d)
>>> I get a plot with plotted values rounded or truncated to 7 digits of
>>> precision, it seems. Is there a way to preserve the precision of the
>>> data when using py-gnuplot?
>>
>> Gnuplot.py's default is to treat data as 32-bit floating point numbers.
>> But if you explicitly pass double-precision data to Gnuplot.Data(),
>> then I think it will pass the data to gnuplot with the higher precision.
>> I.e., do something like
>>
>>    d = Gnuplot.Data(
>>        numpy.array(x, dtype=numpy.float64),
>>        numpy.array(y, dtype=numpy.float64),
>>        with_='points',
>>        )

> Thanks, Michael, but, sadly, that did not work.
> -- Richard

Hmm, sorry, I thought I had built something like that in.  If so, it
would have been more then 10 years ago, so I hope you will forgive me :-)

The formatting is done by the function write_array() in the file
utils.py.  The individual elements are formatted using '%s'.  If you
change that to '%r' (in two places) or '%.20g' or something like that, I
think you'll get what you want.  Or, of course, you can always write the
data to a file yourself, similar to how PlotItems.Data() does it.

Michael

-- 
Michael Haggerty
mh...@al...
http://softwareswirl.blogspot.com/

Michael:
Neither of those suggestions seemed to work. However, in the float_array function in utils.py, simply changing
"return numpy.asarray(m, numpy.float32)"
to
"return numpy.asarray(m, numpy.float64)"
did the trick for me.
I have another question, but I'll ask it in a separate e-mail.
-- Richard

-----------------------------------------------------------------------------
| Richard B. Langley                            E-mail: la...@un...         |
| Geodetic Research Laboratory                  Web: http://gge.unb.ca      |
| Dept. of Geodesy and Geomatics Engineering    Phone:    +1 506 453-5142   |
| University of New Brunswick                   Fax:      +1 506 453-4943   |
| Fredericton, N.B., Canada  E3B 5A3                                        |
|        Fredericton?  Where's that?  See: http://www.fredericton.ca/       |
-----------------------------------------------------------------------------

________________________________________
From: Michael Haggerty [mh...@al...]
Sent: Monday, December 09, 2013 11:09 AM
To: Richard Langley
Cc: gnu...@li...
Subject: Re: [Gnuplot-py-users] Resolution of Plots

On 12/09/2013 03:40 PM, Richard B. Langley wrote:
> On Monday, December 9, 2013,343, at 3:47 AM, Michael Haggerty wrote:
>> On 12/08/2013 02:49 PM, Richard Langley wrote:
>>> I have a file of x,y real data pairs with the values to 9 digits of
>>> precision. When I use gnuplot to plot the data directly using, say:
>>> plot 'HHAT3260.csv' using 2:1
>>> I get a detailed plot with the values at their original precision.
>>> However, when I use gnuplot within Python (having read in the file to
>>> float lists x and y) using:
>>> g = Gnuplot.Gnuplot(debug=1)
>>> d = Gnuplot.Data(x, y, with_='points')
>>> g.plot(d)
>>> I get a plot with plotted values rounded or truncated to 7 digits of
>>> precision, it seems. Is there a way to preserve the precision of the
>>> data when using py-gnuplot?
>>
>> Gnuplot.py's default is to treat data as 32-bit floating point numbers.
>> But if you explicitly pass double-precision data to Gnuplot.Data(),
>> then I think it will pass the data to gnuplot with the higher precision.
>> I.e., do something like
>>
>>    d = Gnuplot.Data(
>>        numpy.array(x, dtype=numpy.float64),
>>        numpy.array(y, dtype=numpy.float64),
>>        with_='points',
>>        )

> Thanks, Michael, but, sadly, that did not work.
> -- Richard

Hmm, sorry, I thought I had built something like that in.  If so, it
would have been more then 10 years ago, so I hope you will forgive me :-)

The formatting is done by the function write_array() in the file
utils.py.  The individual elements are formatted using '%s'.  If you
change that to '%r' (in two places) or '%.20g' or something like that, I
think you'll get what you want.  Or, of course, you can always write the
data to a file yourself, similar to how PlotItems.Data() does it.

Michael

--
Michael Haggerty
mh...@al...
http://softwareswirl.blogspot.com/