15. Appendix: Slit selection algorithm

This section describes how GMMPS decides which slits are kept for a particular mask.

15.1. Field of View

The algorithm discards an object if more than 10% of the slit is truncated by the edge of the slit placement area, or if the object falls entirely outside. In case of band-shuffling mode, an object is also dropped if the slit is not entirely contained within a science band.

15.2. Acquisition stars

GMMPS will place all acquisition stars defined in the OT in all masks derived from that OT.

15.3. Handling object priorities

Once the acquisition stars are placed, as many priority 1 objects as possible will be placed, followed by priority 2 and then priority 3 targets. An object of higher priority will always be placed on a mask at the expense of any number of lower-priority objects.

15.4. Details of the slit selection algorithm

The follwoing information is not required for using GMMPS. The slit selection algorithm is implemented in C++ and works as follows:

  1. Each target is represented in a Slit object, and an array of Slit objects is created, too.
  2. A conflict graph is constructed for all Slit objects. This is a representation of all the slits on the mask, with the vectors representing each respective object and edges between any two vectors that cannot both be placed on the same mask. This first conflict map is made so that when a slit is placed the program will know which slits it can remove (and therefore remove from consideration for placement) from the main Slits array. This will be useful in the following steps when we are considering slits of the same priority only (but nonetheless cannot place overlapping slits from different priorities).
  3. The acquisition slits are placed on the mask. This step is similar to the following step for slits with priorities 1 to 3, with two exceptions: First, acquisition slits will be placed on all masks. Second, the spectra of acquisition stars are allowed to overlap (i.e. they do not create conflicts amongst each other). In general, a conflict graph is made for all acquisition stars, the objects are then ranked by degree of the object’s graph representation (the number of edges connected to the vertex in the conflict graph) in the conflict graph that was based on objects of the same-priority. Objects are then placed on the graph starting with the lowest-degree object and proceeding until no more objects can be placed on the graph. When an object is placed any other object that conflicted with that object is removed from the Slits array, as well as the local (same priority level) and global (all priority levels) conflict graphs. The local conflict graph is updated during this step to reflect the removal of these objects. When the local conflict graph is empty then no more objects can be placed on the graph, and the program moves on to the next-lower priority level and repeats this process.
  4. If no auto-expansion is desired then the program moves on to the final ODF creation. Otherwise the algorithm sorts all slits by the non-dispersion direction (the x-direction for F2 and the y-direction for GMOS) and expands each slit as much as possible in that direction without causing spectra to overlap (taking into account the minimum slit separation). The modified slit array is stored and the final ODFs are created.