update TODO list (everything on the old list was DONE!)

docs/reference/todo.inc

@@ -4,41 +4,9 @@
 TODO list
 =========
 
-TODO
-
-#. Interact with REGGUI/MyQAiON:
-
-   #. How should I communicate error cases back to REGGUI/MyQAiON
-
-       * disk full
-       * parsing problems
-       * simulation crash (hasn't happened)
-
-   #. Replace shared directory with https
-   #. More in general: define interface
-
-#. Automatic CT protocol recognition based on user-supplied rules
-#. Use stop-on-script *always*, not only for jobs with a statistical uncertainty goal.
-
-   * stop when the *total* number of primaries exceeds the minimum (currently: each job simulates ceil(Nprim/Ncore) primaries)
-   * stop on user request, but still compute the final result
-   * provide estimate of progress, time to completion
-   * always compute & report average statistical uncertainty for final result (also if no goal for this was computed)
-   * always compute & report gamma analysis, iff TPS physical dose is available
-   * any combination of Nprim, Unc and TimeOut is possible:
-
-       * simulation will run at least Nprim primaries, unless TimeOut is reached
-       * simulation will run until the uncertainty is better than Unc, unless TimeOut is reached
-
-   * update simulation goals/constraints (Nprim, TimeOut, Unc) during a job.
-
-DONE
-
-#. Check that coordinates are correct: CT has origin=0,0,0 
-#. Beam line is derived from the *name* of the BDL.
-#. CT calibration is also based on the *name*. New name => new cache.  It would be better to make a md5sum of the contents of the density and composition tables.
-#. CT calibration currently ignores the composition table. The table that I saw is similar to ours, but it lacks the higher HU range. I am using our own Schneider composition table.
-#. REGGUI can handle only one calculation at a time. It would be nice to be able to submit several.  The IDEAL daemon *can* already handle multiple jobs concurrently.
-#. Jobs are identified by the name of the job directory, e.g. `MCsquare_20_08_18_11_33_35`.  We have also seen job directory that are just called "temp".  Should we anticipate that or can we assume that REGGUI will generate unique job directory names?
-#. Radiation type
-
+* Make IDEAL/pyidc into a "pip" project.
+* Support rotating gantry.
+* Support range shifter on movable snout.
+* The job_executor and idc_details classes are too big, lots of implementation details should be delegated into separate classes.
+  + Specifically, the creation of the Gate work directory and the creation of all condor-related files should be wrapped in separate classes. And then it should be more straightforward to support other cluster job management systems, such as SLURM and OpenPBS.
+* There are still many ways in which the user can provide wrong input and then get an error that is uninformative. For instance, using "G4_Water" instead of "G4_WATER" as override material for a ROI results in a KeyError that only says that the "key" is unknown.