We need to think a bit to the insert temporality here.
sps_visit and mcs_exposure are inserted when the fits file is actually created which make sense.
For sps_sequence(iic_sequence), visit_set have two options regarding the insert timing:

• before the exposure and after gen2 returns getVisit, earliest option. in that case some of the existing constraints would need to be relieved because right now visit_set cannot be inserted before sps_visit, note that INSTRM-1305 raises a similar problem.
• after the exposure, works with the current schema, but it wouldn't be ideal in the obslog purpose.

dragging Kiyoto Yabe here, because INSTRM-1293 might not be sufficient.

I think the constraints might be fine. The sps_sequence creates the new visit_set_id and provides all the command/sequence info, right? So create that before taking any exposures. Each sps exposure will then get a new pfs_visit_id from gen2: when the exposure is done insert the (visit_set_id, pfs_visit_id) row in visit_set. No?

Actually, if I'm reading models.py right, the primary key on visit_set should be sps_sequence.visit_se_id instead of sps_visit.pfs_visit_id.

right. In anycase, the join will only be possible when the exposure is finished, but looking at the obslog it might be okay.

merged to master