Use the minimum of image.nobjects and the number of objects available to be rendered.

[jchiang87]

Mike's suggested fix seems to work. However, I don't know how to write an isolated unit test for this. The only way I can see to test it would be to run a full simulation and checking the outputs. Even running with just a couple of objects takes a long time. Suggestions welcome.

[rmjarvis]

I think you can use the truth catalog output to test.
Case 1:
instcat with say 10 objects, and output.nobjects=3 -> check that there are 3 items in the output truth catalog.
Case 2:
instcat with 3 objects and output.nobjects=10 -> check that there are again 3 items in the truth catalog.
Case 3:
instcat with 3 objects and no output.nobjects specified -> same output as case 2.
And if the first 3 objects in case 1 are the same as the 3 in cases 2,3, all three output files should be identical. Both the truth catalog and image.

[jchiang87]

Sure, but I think that still involves running full simulations. I was hoping there would be something that could be done to test the setting of LSST_ImageBuilder.nobjects directly for the different yaml inputs.

[rmjarvis]

There are ways to do that with mock, but I always prefer just running a full integration test if possible. You can make the fluxes quite small, so the run time could be very quick.