read in the object ids from the instance catalogs as strings

[jchiang87]

Currently they are assumed to be ints. Looks like there are just two lines to change:
https://github.com/LSSTDESC/imSim/blob/master/python/desc/imsim/imSim.py#L186 and line 195.

phosim already allows for string ids in v3.7.x, and the SNe for DC2 may have string ids.

[jchiang87]

In sims_GalSimInterface, this line also explicitly expects the object ids to be an int:
https://github.com/lsst/sims_GalSimInterface/blob/master/python/lsst/sims/GalSimInterface/galSimInterpreter.py#L599

[cwwalter]

Recording information from CWW/JC in-person conversation.

This likely will have an effect on the RandomWalk galaxies and also could affect downstream programs as the code for the object type is encoded in the lower 10 bits of the integer uniqueID.

[cwwalter]

Is this still happening, or on hold/canceled due to the other ramifications of the change? Or still needs discussion with Rahul and all?

[jchiang87]

I think we need to make this change to support the WFD SNe. Those are the only object_ids that will not be cast-able as ints, so any downstream programs that read the instance catalogs will still be able to extract things like galaxy_id from the lower 10 bits.

[cwwalter]

OK. So it will always be an added sub-string to the original uniqueID (now as a string)? We already need this for 2.0i even without the sprinkler. Is that correct?

@EiffL just flagging you on this in case it is relevant to the knot class.

[jchiang87]

So it will always be an added sub-string to the original uniqueID (now as a string)?

I don't know the final format but it will be something like <galaxy_id>+<sn_id> = 1000001682_MS_564_6

We already need this for 2.0i even without the sprinkler. Is that correct?

Yes, this is for SNe in the WFD region which will be included even with the sprinkler disabled.

[EiffL]

Argh, yes, changing the uniqueID will impact the RandomWalk (because the random seed used to sample the knots is based on the id), but that should be easy to accommodate. Just let me know if/when the change happens, and I can make sure the RandomWalk code gets properly updated

[cwwalter]

Thanks Francois, don't let us close this issue unless it is working for you too...

[jchiang87]

@EiffL Regarding uniqueID, the only difference is that we'll store them as strings instead of ints, i.e., here
So if your knot objects continue to have ids that are cast-able as ints, the code that sets the random seed,
https://github.com/lsst/sims_GalSimInterface/blob/4c67ec1e7e6e55df837d5630ae9b75003d68c473/python/lsst/sims/GalSimInterface/galSimInterpreter.py#L454
should be fine.

[EiffL]

agreed, it should be fine as long as the objectids cast to ints correctly

[cwwalter]

Is there a PR for this one? Or is it in the commit of some of the others? I can't find it in imSim or sims_galSimInterface..

[jchiang87]

For imSim, it's integrated in the dc2_run2.0_rc branch directly (2786b1f), and there's this PR in sims_GalSimInterface

[cwwalter]

OK got it; thanks. I had the centroid one, but I thought there were more. The Random knot changes are somewhere else?

[jchiang87]

The RandomWalk code isn't affected by this change since it had already cast the uniqueId to an int when passing it as a seed to the rng. As long as the instance catalog code doesn't use uncastable string ids for the knots, no changes are needed for the RandomWalk code.

[cwwalter]

This change was made in the 2.0i development work.