Invalid EPS export of boxplots

[farleylai]

I recently need to draw boxplots to show inter-packet delay(ms). There are three configurations to plot respectively but two of them always end up with invalid file formats. I tried to compute the statistics to see what's wrong with the dataset as follows but cannot figured out why.

Config 1(34870 rows): // this works and the output eps can be opened
statistics [min, medium, Q1, Q2, Q3, max] in order
column0 [0.0, 0.0, 0.0, 0.0, 0.24414062, 1.0309203, 358.88672]
column1 [0.0, 0.48828125, 0.24414062, 0.48828125, 0.7324219, 2.9891071, 1748.5352]
column2 [0.0, 0.7324219, 0.24414062, 0.7324219, 2.1972656, 5.3231072, 3301.0254]
column3 [0.0, 0.7324219, 0.24414062, 0.7324219, 2.9296875, 7.8209295, 4803.2227]

Config 2(34829 rows): // invalid eps to open, Config 3 is similar and not shown
statistics [min, medium, Q1, Q2, Q3, max] in order
column0 [0.0, 0.0, 0.0, 0.0, 0.0, 0.86110216, 722.65625]
column1 [0.0, 0.0, 0.0, 0.0, 0.0, 2.6439805, 1781.25]
column2 [0.0, 0.0, 0.0, 0.0, 0.0, 5.0793586, 3304.6875]
column3 [0.0, 0.0, 0.0, 0.0, 0.0, 6.8676205, 4894.5312]

The acrobat distiller shows the error messages as follows:

%%[ Error: undefined; OffendingCommand: NaN ]%%
Stack:
0.0
%%[ Flushing: rest of job (to end-of-file) will be ignored ]%%
%%[ Warning: PostScript error. No PDF file produced. ] %%

Any ideas?

boxplots.zip

[farleylai]

Using double instead of float, the values are not zeros any more and the boxplot exports in EPS work. But still, too many zeros are unacceptable?

[eseifert]

Hi @farleylai. Thanks for reporting the issue and providing example output files. I haven't looked into it in detail but my guess is, that there's some division by zero happening because of the zero values. We'll try to fix this soon.

[farleylai]

I think so. The 'NaN' seems like the clue.