rcv icon indicating copy to clipboard operation
rcv copied to clipboard

Clean up test folder

Open HEdingfield opened this issue 7 years ago • 4 comments

It would be nice to standardize the file names and clean up the test folder. I like the format date-location-office. Maybe we could consistently prepend cvr_* and config_* to those.

There also might be some test files in there that aren't needed anymore or, if they are, could also have more consistent names (e.g. "minimum-threshold-test-cvr.xlsx", "test_continue_tabulation_config.json", etc.).

HEdingfield avatar Jul 04 '18 18:07 HEdingfield

Re-opening because it seems like tests added since this was completed don't follow these standards. In particular, should address:

  • Everything that's run in TabulatorTests.java should follow the consistent test_* format (I see lots of folders with *_test and some that don't even have the word test.
  • It seems like some CVR files, like 2013_minneapolis_park_cvr.xlsx are used multiple times (e.g. 2013_minneapolis_park_hare_cvr.xlsx, 2013_minneapolis_park_bottoms_up_cvr.xlsx, 2013_minneapolis_park_sequential_cvr.xlsx all appear to be the exact same file; @tarheel please confirm); maybe we should have a single copy and each of those test configs reference the same location?
  • This is a minor / partial blocker to #281, because I think we need to distinguish between files that are included as regression tests, and those that are there for manual testing / playing with by users. Perhaps the latter type should be stored in a separate directory.
  • Which leads to a final thing we should add: a setup that lets users easily trigger / play with the interactive tiebreaker.

HEdingfield avatar Jun 14 '19 17:06 HEdingfield

Yes, we do have a number of duplicate CVR files in the test data. If we wanted to avoid that, we would need to create an exception to the naming scheme, I guess. Also, 2013_minneapolis_mayor_scale has 13 copies of the same file because we don't allow a config to list the same source file multiple times. I agree it's dumb to have duplicate files in the repo, though. Feel free to address these issues in a way that you think is reasonable.

tarheel avatar Jun 21 '19 04:06 tarheel

It's not worth it to come up with a scheme to exclude a mere 25MB of test files. If we make a distro which includes test files, we should include them all.

moldover avatar Jun 21 '19 05:06 moldover

Began the process of distinguishing between "test" and "sample" input files in PR #316. Verified that, in that PR, only the sample_interactive_tiebreak folder isn't used in a unit test, so all other folders should start with the prefix "test".

HEdingfield avatar Jun 28 '19 07:06 HEdingfield