Add paper "Performance of Python runtimes on a non-numeric scientific co...

[riccardomurri]

This is the paper associated to the "Py vs. Py" poster.

The .rst is converted to LaTeX without errors, but I cannot compile the generated LaTeX without hand-editing the file. In particular:

• the rst2latex converter inserts blank lines here and there, where TeX does not expect them;
• tables are particularly affected by this, as blank lines (hence, \par) are inserted amid the items on a single line;
• figures need to be better positioned and resized.

I can produce the figures in any other format supported by Matplotlib.

[pdebuyl]

Hi,

Please try with grid tables. I uploaded a modification to the branch murri of euroscipy_proceedings.

I also corrected a few bugs.

[riccardomurri]

Hi,

thanks for the fixes - I've been sick for a few days and could not attend to this. Do I need to do anything else on the .rst now?

Cheers, Riccardo

[pdebuyl]

Nothing immediate is needed. The tables are too wide but we'll take care of that part of the layout later.

[pya]

There are several formatting issues in the final PDF:

• The cProfile table is only partially visible. A smeller font may help.
• Table 2 does not fit on the page and is distorted. Making the header "Cython 0.19.1 (w/ hints) line break before the parenthesis may help.
• All figures a too small. The text is not legible when printed on A4 paper. Either make the graphs bigger or increse the font size relative to the graph.

[pya]

While not the major goal, an evaluation of the effort setting up the different run time system might be a useful side result.

A table semi-quantifying critiria such as:

• installation effort
• OS support
• C-complier support
• documentation quality (and quantity)
• learning curve
• other criteria for suitability for everyday use

could be nice addition to the paper.

[pya]

In the section "Experimental setup" it says:

(the difference between the two decorators seems to be that @autojit infers the types at runtime, whereas @jit requires the programmer to specify them);

I would recommend to research this topic to the end and get rid of the conjunctive. If you cannot confirm this, it would be better to drop this statement altogether.

[pdebuyl]

Hi Riccardo,

You may check euroscipy_proceedings "murri" for a fix of the wide tables. I also dropped two columns (filled with 0) of the profile output, see if that suits you or not.

[riccardomurri]

Hello,

many thanks for the suggestions and the fixes -- I'm overwhelmed with work these days and cannot promptly respond. What's the time frame for getting this fixed?

Regarding formatting: I know how to fix the LaTeX, but I'm rather clueless at how to influence formatting in reST. Do you need the fixes to be in the reST sources?

Thanks, Riccardo

[pdebuyl]

Dear Riccardo, I can take care of the rst layout issues. As we use a consistent rst typesetting scheme for the proceedings, we do not edit the LaTeX code directly.

As far as feedback on the content itself is concerned, we'll send reminder when it becomes pressing.

[pdebuyl]

Dear Riccardo,

Your "reviewer number 2" does not seem to take care of the review, so I will do it. I have added minor comments (stylistic) to the pull request. Here are a few general comments on the manuscript:

• You should define, though briefly, what is a homology.
• the fontsize in your plots is unreadable, please increase it.
• figures 1 and 2 are not very clear. consider changing the color and/or linestyles, the plot type (a bar plot may be more suitable).
• a short conclusion would be nice to have.

[riccardomurri]

Following @pya 's suggestion, I've added a table comparing installation and porting effort for runtimes. The results are of course much more subjective than anything else in the paper.

There are still a few issues with the table:

• I cannot make the width of any column larger than what is needed to fit the first column.
• reST table options :header-rows: and :stub-columns: produce incorrect LaTeX code (blank line in the argument to \textbf{})

[riccardomurri]

Hi @pya and @pdebuyl,

many thanks for your very useful comments! I think I have now addressed most of them; for what concerns me, the paper is now in final form.

Let me know if you want me to do further edits.

Thank you very much, Riccardo

[pdebuyl]

Hi Riccardo,

Thanks for the updates!

I've corrected typos and a few stylistic issues. You can check them at https://github.com/euroscipy/euroscipy_proceedings/tree/murri/papers/murri (this branch also contains all contributions that are already merged, however).

I have also tweaked the graphics to make them fit in single columns, as the data density was quite low.

The table is now using the width of the page. I've emailed you the resulting pdf.

Can you review the changes? Afterwards, I'll merge your paper :-)

[riccardomurri]

Hi Pierre,

I've corrected typos and a few stylistic issues. You can check them at https://github.com/euroscipy/euroscipy_proceedings/tree/murri/papers/murri

Looks good to me, with the exception of a couple corrections that I've indicated in the liune-by-line comments.

I have also tweaked the graphics to make them fit in single columns, as the data density was quite low.

Thanks!

The table is now using the width of the page. I've emailed you the resulting pdf.

I'm afraid I didn't get any email other than this.

Ciao, R

[riccardomurri]

Sorry I couldn't push the final changes by Friday. I have just committed 32e108c, which fixes the arrangement and size of the figures. Could you please merge it too?