Re: Development document for adding support for matplotlib graphing to relax. -- March 27, 2014 - 11:54

Hi,

This all works!  To have this, we can add the 'format' argument to the
matplotlib.write user function.  It can have the following values
which can all be documented in the docstring:

- 'Script', this produces a single Python script containing everything
- all the data and matplotlib commands needed, and has a single show()
call at the end,
- 'PDF',
- 'PS',
- 'EPS',
- 'PNG',
- 'SVG'.

Then the matplotlib.view user function could have the 'file',
'interpreter', and 'python_exe' options.  If the file is one of the
vector or bitmap graphics, then the OS dependent image viewer can be
opened.  But if it is the Python script, then 'interpreter' argument
can be used.  It could have two values 'internal' and 'external'.
'internal' will just run plt.show() in the current relax instance,
i.e. the matplotlib script will be imported as a module.  'external'
would then take the 'python_exe' argument value and run the matplotlib
script in any Python version the user chooses.  I think this would
enable everything you mention.

Regards,

Edward


On 27 March 2014 11:30, Troels Emtekær Linnet <tlinnet@xxxxxxxxx> wrote:
> Hi Edward.
>
> Regarding #3.
> I think the best support would be make an external python file.
>
> The question is, if the data should be store in different variable arrays,
> containing x,y data etc.
> or as a ordered dictionary.
>
> The matplotlib.view(), should then just call: "python spin_file.py".
> Then relax is not dependent on matplotlib installation, but a functional
> external python installation with matplotlib.
> Here "python" could point to for example to the python packaged in the
> Enthought distribution.
> That distribution is bundled with matplotlib. But it could be any of these
> easy python distributions.
>
> That should save a default .png file, but a plt.show() in the script will
> call the graph window to appear, and the graph can me modified by hand and
> saved as wished.
>
> For this method, relax don't need a module check.
> (There is also no module check for xmgrace ? )
>
> Again about the data.
> If the relax state files are simple xml files, then the relax could instead
> write a special matplotlib state file with the interpolated values saved.
> And the plot script would access this state, extract x,y values and plot
> them?
>
> Just some thoughts.
>
> Troels Emtekær Linnet
>
> Ah, there is one step I forgot about.  The Grace graph generation
> happens in lib.software.grace module whereas the assembly of data from
> the relax data pipe happens in the pipe_control.grace module.
> Therefore the missing step is:
>
>
> 6)  Matplotlib interface.
>
> Create a new module lib.software.matplotlib.  This is where matplotlib
> is imported and the plots are produced and saved.  The
> lib.software.grace module should not be copied as the functions in
> lib.software.matplotlib would be completely different.  But the
> write_xy_data() and write_xy_header() functions can be used as
> inspiration. As matplotlib does not appear to have a header from data
> separation, then maybe on write_xy() function can be written for 2D xy
> data plots.  A different function for 3D data could be created as
> maybe write_3D().
>
>
> Note that at some point in the future, the plotting modules in
> lib.software might shift to lib.plotting.
>
> Regards,
>
> Edward
>
>
> On 27 March 2014 10:28, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:
> > Hi,
> >
> > The following is planning document for adding support for matplotlib
> > graphing to relax.  The plan will be broken into a few simple steps:
> >
> > 1)  Module detection.
> >
> > Firstly the dep_check module would need to be modified to detect the
> > presence of the matplotlib Python package.  In addition, the
> > package_info() method in the 'info' module should be changed to print
> > out if matplotlib is present and which version is present
> > (matplotlib.__version__).  Then matplotlib will be shown when running
> > relax with the --info command line option.
> >
> >
> > 2)  Icons
> >
> > For the GUI and the user manual, icons are needed.  These can be taken
> > from the matlplotlib sources and added to the graphics/relax_icons/
> > directories.  The 16x16, 24x24, 32x32, and 48x48 icons could be taken
> > from the ./doc/_static/favicon.ico file, specifically layers #4, #5,
> > and #6.  The 128x128 icon used in the manual could be regenerated
> > using the matplotlib graphic script ./examples/api/logo2.py, but with
> > the text removed.
> >
> >
> > 3)  Uf frontend.
> >
> > For this, the user_functions.grace module can be copied and modified
> > to change all instances of 'grace' with 'matplotlib'.  So the user
> > functions would be matplotlib.view and matplotlib.write.  New user
> > functions could be added later on if additional functionality from
> > matplotlib is desired.
> >
> >
> > 4)  Uf backend.
> >
> > The pipe_control.grace module can be copied to pipe_control.matplotlib
> > and changed as needed.  The key functions are view() and write() as
> > these are used by the user function frontends.  All other functions
> > can be added as needed to simplify the matplotlib graph generation
> > (and the old ones form grace deleted).
> >
> >
> > 5)  Saving format.
> >
> > Do we output in SVG or PNG?  I don't know matplotlib, but is there a
> > way of saving a matplotlib native file format so it can be modified
> > later?  The options should be available as arguments to the
> > matplotlib.write user function.
> >
> >
> > Completion of these steps is sufficient for full matplotlib support
> > within relax.  Then adding the plotting as default for the
> > auto-analyses would be trivial.
> >
> > Regards,
> >
> > Edward
>
> _______________________________________________
> relax (http://www.nmr-relax.com)
>
> This is the relax-devel mailing list
> relax-devel@xxxxxxx
>
> To unsubscribe from this list, get a password
> reminder, or change your subscription options,
> visit the list information page at
> https://mail.gna.org/listinfo/relax-devel