Plotting tools

[rs028]

Python script:

• import too many libraries?
• generally slow
• inefficient way to read file header and data (open file twice)
• spacing between panels

gnuplot script:

• does not read variable names from file header
• unequal dimension of panels
• spacing between panels

Matlab/Octave script:

• inefficient way to read file header and data (open file twice)
• does not save to pdf

[rs028]

Python version fixed by PR #394, but still slow. The version with pandas is only marginally quicker than the version with numpy. (keep both?)

[spco]

For reference, it's not the data processing that's really expensive - it's the many calls to savefig() and tight_layout() that do the damage, as seen in this flame graph (this is for the case I'm investigating in #384). save_fig() is taking 0.7s per PDF page (of which there are 12 = 8.4s), while tight_layout() is taking 0.192s per graph (there are 62 of these = 11.9s) :

[rs028]

PR #490 solves this issue -- except for the Python scripts being slow (not sure it can be easily solved, but it's largely just an annoying thing).