Closed moustakas closed 3 years ago
The tiles file to use is specified in the desisurvey config.yaml
file, which has a default of py/desisurvey/data/config.yaml
, but can be specified via the --config-file
option to surveyplan
and surveyinit
. You could create an sv-tiles.fits
file with your custom tiles using IN_DESI=1
for all those tiles, and use a custom desisurvey config.yaml
to point to that custom tiles file. If that file is in $DESIMODEL/data/footprint/
then you don't have to specify the path, otherwise (I think) you can put it anywhere if you include the full path in the desisurvey config.yaml
file.
i.e. is the real requirement to be able to use IN_DESI=0 tiles in desi-tiles.fits
, or rather is the requirement to be able to use arbitrary tiles files? I think the latter is more flexible and already supported (albeit via config file alternations rather than command line options). Does that provide the functionality that you need? (and if not, why not?)
I understood all your comments, and indeed I'm using the --config-file
option to surveyinit
, surveysim
, and surveyplan
. Let me rephrase as a two-part question:
$DESIMODEL/data/footprint/desi-tiles.fits
file? My understanding was that -- eventually (but before the start of SV itself) -- they would be.IN_DESI=1
or IN_DESI=0
? If the former, then desisurvey.schedule
will work as-is; if not, then we need the option of simulating observations of "non-DESI" tiles (assuming it will still be useful to do survey simulations, even when we're on-sky).For now I propose that we keep SV tiles out of desi-tiles.fits (i.e. a separate file) and use IN_DESI=1 so that we can point the config files at them and run with it. i.e. "IN_DESI=1" becomes a statement about footprint, not about main-survey vs. other programs.
Previously I had thought that we would then merge those tiles into desi-tiles.fits under a different PROGRAM name, but this thread highlights that that could be problematic if we retain IN_DESI=1 since we don't want the main survey to try scheduling them. Eventually survey* code could have an option for which PROGRAM values to consider, so that they could use the separate or merged tiles files.
I think we have converged on a workable solution. Specifically, SV will have its own tiles file satisfying:
The SV tiles will be specified at run time using the config tiles_tile
option.
Any objections to closing this now?
Closing this as we are now through SV.
desisurvey.schedule
assumes that all tiles haveIN_DESI==1
(see here) but for Survey Validation mock-ups (and perhaps Survey Validation itself) it would be convenient if there was an option of using non-DESI tiles / pointings.