code to support newtiles, and S vs. R #135
Conversation
…rect behavior when tiles is None
|
Looks good to me! |
|
Hmm. I apparently had an old copy of the newtiles branch while testing, and the latest version of that data branch fails. I'm chasing those now... |
|
Arg. UPPERCASE vs. lowercase column names. I just updated desi-tiles.fits and .ecsv in the newtiles branch to use UPPERCASE columns to match the datamodel of trunk (plus extra columns...). |
|
Sorry. It looks like this update to SVN lost the metadata in the ECSV file. I can further update that later, though, so let's just press on... |
|
Ah, sorry about the metadata. I updated ecsv by re-writing a subset of the fits columns as ecsv, rather than reading ecsv, updating, and rewriting the ecsv itself. But yes, let's press on... :) Will wait for confirmation from @tskisner about fiberassign impact before merging. |
|
Looks good to me |
This PR provides code to match @schlafly's "new" tiling pattern in svn desimodel/branches/newtiles . That branch of the desimodel data expanded the format of the tiles file which broke some of the desimodel code (e.g. multi-dimensional columns), so this adapts to that. I also updated the tests to support both current trunk and the newtiles branch.
Another change that came along for the ride is splitting out the S vs. R vs. Z vs. N data from DESI-0530 into a separate file (in the newtiles branch) instead of adding S (arclength) as a new column in the platescale.txt file (trunk). The code here is compatible with either, but paves the way to update platescale.txt with the as-built model from DESI-4037 (issue #134) without breaking fiberassign.
@tskisner please review for fiberassign impacts.
@schlafly ok?
This PR replaces PR #95 that I just closed.