[IPOL discuss] matplotlib? + decisions + postscript experiment

[Miguel Colom]

As a summary, this is my proposition for plotting:

1) Authors can generate their own figures from their C code using the
libraries installed on the server. This way the author can generate the
figures with absolute freedom and IPOL's publication policy is respected.

2) For those authors whose figures are standard plots (for example: 2D
plots [1]) we offer a standard script that generates the figure for the
given data. For 2D plots, a text file with the coordinates of the points
in the IPOL format is enough. The Python demo code can configure the
script to set the title, legends, colors, line styles, etc.
I think it's important to provide such an standard way to generate 2D
figures, because most of the authors will agree on the standard result and
frees them from writing code has nothing to do with their algorithm. If
not, alternatives 1) and 3) are available.

3) For those authors than don't want to generate their plot from the C
code (option 1) and the output of the script isn't adapted enough to their
needs, the have the possibility to use gnuplot to produce figures.
However, this method should be avoided and advised against, because if
gnuplot API changed, the demo wouldn't work right.

Best,
Miguel

[1]
http://dev.ipol.im/~colom/ipol_demo/noise_estimationAuto/tmp/2B01424346155B7662B581F0A65730B3/output.png)