tile QA defaults / usability

[sbailey]

This PR adds some convenience features to simplify the daily tile VI procedure:

• updated desi_tile_vi defaults so that you can just run "desi_tile_vi" from a surveyops/ops svn directory without having to specify extra arguments:
  • --infile defaults to tiles-specstatus.ecsv instead of tiles.csv (which was both wrong and now crashes anyway since QA columns have been removed from tiles.csv)
  • --user defaults to unix username
• desi_update_tiles_specstatus
  • checks if input tiles-specstatus.ecsv is up-to-date or needs an svn update first; use --force to override (at your own peril... svn merge conflicts are likely if you aren't up-to-date first)
  • --specstatus now defaults to "tiles-specstatus.ecsv"
  • new --dry-run option to calculate what would be updated without saving any files

Standard usage would now be:

export SPECPROD=daily
cd (directory where you have surveyops/ops checkout)
svn update .
desi_update_tiles_specstatus --dry-run
desi_update_tiles_specstatus
desi_tile_vi
svn commit -m "tile QA for YEARMMDD"

The desi_update_tiles_specstatus --dry-run is optional, but a useful cross check before changing anything, and both desi_update_tiles_spectstatus commands will complain if you forgot to svn update first.

@schlafly @araichoor @julienguy

[coveralls]

Coverage decreased (-0.005%) to 26.623% when pulling 221b85a on specstatus into 9c654e6 on master.

[schlafly]

I promised I'd suggest some additional usability improvements:

• support for people with long names (already addressed)
• "only new exposure" mode in desi_tile_vi. I think we want people doing daily maintenance of newly observed exposures to only see the new exposures, and not old "unsure" observations; if someone's said they're unsure, we want an expert to look at it, not for these likely problematic tiles to be mixed in with a bunch of likely good tiles until a non-expert just validates them and whatever issue was present gets lost.
• I think desi_tile_vi should print a summary at the end that would be appropriate for copy/pasting to #survey-ops. e.g.

type	valid	unsure	invalid
new dark
new bright
new other
old dark
old bright
old other
very old dark
very old bright
very old other

Here's I'm imagining separating the new ones (defined as "formerly had 'none' as QA"), the old ones (not none, and lastnight within 2 weeks?), and the very old ones (not none, and lastnight older than 2 weeks?).

[julienguy]

desi_tile_vi already has the option --new which I think does exactly what you want.
I'm adding the option in the command line on the wiki page.

[schlafly]

Thanks, that looks exactly right, sorry I missed it.

[sbailey]

@schlafly I like the idea of the summary table and have also wanted that when running desi_tile_vi. Could you take a pass at adding that? If it is at all tricky and/or if you don't have the time to do it ~today, I suggest that we accept 2/3 bonus features as good enough and merge this and then spawn that request into a separate ticket so that it doesn't block the other features of this PR.

[schlafly]

I can take a pass at this but it won't be today. I don't think we need to conserve merges and am happy if you want to merge now.

[schlafly]

I added my features:

• save every time
• remove save / exit options (can just Ctrl-D or Ctrl-C to exit, since it saves every time)
• pass is renamed "unsure"
• print a summary table at end
  I was able to use this for my VI run earlier today, but I can't say that I've tested it more than that.

[sbailey]

Live use for real QA is good enough testing. Merging now. Thanks for adding the additional features.