Potential vcs test script

jypeter commented 7 years ago

Following @durack1 request in #23 , I have tried to write a self-contained script that produces that produces the figure at the top of https://github.com/UV-CDAT/uvcdat/issues/1080

The script needs the gsw package to run (https://pypi.python.org/pypi/gsw/). I have just done the following in my old versions of CDAT (don't know what people should do for that in conda)

pip install gsw

The script, png and pdf outputs are available there: https://files.lsce.ipsl.fr/public.php?service=files&t=81f4d60326e1f3804b93efc05702b480

You'll find below what I get when I run the same script with different versions of CDAT (I do have some quite old versions)

UVCDAT-1.1.0 deep_ocean_test_uvcdat-1 1 0

UVCDAT-1.5.1 deep_ocean_test_uvcdat-1 5 1

UVCDAT-2.1.0 deep_ocean_test_uvcdat-2 1 0

VCDAT-2.8.0 deep_ocean_test_vcdat-2 8 0

Of course, the output I like best is the one I get from CDAT 1.1.1 ... Ideally, I would like to have the very same plot with any recent (vtk-based) version of CDAT

I guess I have a black png output for 2.1.0 because I'm running vcdat in a virtual machine and vtk/gl based stuff does weird things sometimes (pdf output is OK, but not quite right). I don't have 2.8.0 installed, so a colleague ran it on his Mac

Some thoughts about the 1.1.0 vs 2.8.0

pdf output file of 2.8.0 is ... 87 times bigger than 1.1.0 . It seems that using x.pdf(out_name) instead of x.pdf(out_name, width=311, height=38) does not make it smaller
the relative positioning of objects is good, even if I have a font size problem, so that's pretty good news
A part of the legend text is greenish instead of black. But maybe I did something too non-standard
I have the isoline values in the png, but not on the pdf
The marker sizes are not good, but this probably requires some adjustment, like the font size
probably other differences, but I have to run to catch my bus

sankhesh commented 7 years ago

Thanks for posting the test script @jypeter

@durack1, I don't think this issue is related to the patterns code. We can look at how markers are handled next.

durack1 commented 7 years ago

@jypeter thanks for this! @sankhesh completely agree this is a separate issue, however @jypeter has created some fairly specialised plots in various versions of UV-CDAT and these are perfect "real world" candidates for the test suite. The behavior changes through each version of UV-CDAT is also something useful to be aware of

@sankhesh to make things easier, here's the script deep_ocean_plot_standalone_v2.py.txt

doutriaux1 commented 5 years ago

@scottwittenburg please add this as test in your marker PR. deep_ocean_plot_standalone_v2.py.txt

doutriaux1 commented 5 years ago

goes with #82

CDAT / vcs

Potential vcs test script #126