Consolidate testing data

[russelldj]

It would be nice if there was a centralized location for unit test data. As a concrete example of my frustration, I want to write a unit test in ocv which uses the same image as the unit tests in vxl. While I might be able to hack together a mechanism to point it at that data, I'm not even sure that's portable and that seems sub-optimal long term.

[aaron-bray]

I started to use data.kitware to store example data
https://data.kitware.com/#collection/5991c8df8d777f7d33e9c2d1

And use CMake to pull the files to your build directory for use
https://github.com/Kitware/kwiver/tree/master/examples/external

This works pretty well to store large amounts of data

[russelldj]

Thanks Aaron, this is useful. However, it's not exactly what I was talking about. I was mentioning that there are becoming duplicate data files for different arrow tests because each one has its own unit test data directory.

[aaron-bray]

That makes sense
I think we could probably add a data foler with a hierarchy of folders under it that contain all text/example data
Folders could hold both the actual file in the repo (LFS?) as well as sha512 files
And maybe create config parameters the hold the data root directories you could reference in examples/tests
such as <REPO_DATA_ROOT>\arrow\ocv\test_image1.png and <EXTERNAL_DATA_ROOT>\arrow\ocv\test_movie.mp4

Where REPO_DATA_ROOT is in source and EXTERNAL_DATA_ROOT is in build, and they resolve to paths specific to your machine

something like that..
food for thought...