Title:                  Proposal for MeasurementSet changes 
 Person responsible:     Mark Wieringa (mwiering@atnf.csiro.au)
 Originator of proposal: Mark Wieringa (mwiering@atnf.csiro.au)
 Exploders targeted:     aips2-workers
 Time table:
         Date of issue:          1996 September 16        Done
         Comments due:           1996 September 25        Done
         Revised proposal:       1996 September 28        Done        
         Final comments due:     1996 October 3           Done
         Decision date:          1996 October 4           Done
 
 Statement of goals:
 
 To 
 (1) fix some inconsistencies and remove unused items in the current
     MeasurementSet and 
 (2) change the required status of some columns and sub-tables to optional
     to avoid wasted space and/or duplicate or dummy information and
 (3) add required support for Measures for coordinates & time
  
 Proposed changes:
 
 1. Remove DEFAULT_CALIBRATION keyword from main table
 2. Make DATA column optional (like FLOAT_DATA)
 3. Add member function Bool MeasurementSet::makeComplexData() to create
    the DATA column from FLOAT_DATA if needed. 
 4. Remove MAX_ANT keyword from ANTENNA table
 5. Remove ANTENNAS and NUM_ANT columns from ARRAY table, fix
    description of NAME (String) 
 6. Remove NUM_FEED keyword from FEED table
 7. Make BEAM_OFFSET column in FEED table of type Double (wrong in docs
    only)
 8. Make RECEPTOR_ANGLE in FEED table a 1-d array of length NUM_RECEPTORS
 9. Add a TIME column to the FIELD table
 10. Add OBSERVATION_ID column to the OBS_LOG table
 11. Make the SOURCE table optional, remove the SOURCE_ID column from main
     table (i.e., SOURCE table is only pointed to from FIELD table) 
 12. Remove NUM_CHAN column from SYSCAL table
 13. Change WIND_VELOCITY to WIND_SPEED in WEATHER table
 14. Make WEIGHT column in main table scalar (*)
 15. Add MEASURE_REFERENCE keyword to columns to take the Measure enum
     specifying the reference frame: 
     e.g., the keywords for FIELD/POINTING_DIR would be:
        MEASURE_TYPE:   DIRECTION  (as now)
        MEASURE_REFERENCE: J2000   (new) n.b. could also be B1950 or GALACTIC
        UNIT:           rad        (as now)
 16. Add convenience functions to the MS to
     easily turn column values into Measures: 
     e.g. for the above example 
     MDirection MS::directionMeasure(const Table&, const String& columnName)
     called with MS::directionMeasure(ms.field(),"POINTING_DIR");
     will return an MDirection pointing at the J2000 pole. 
     MPosition:set() can then be used to fill the coordinate value and, if
     necessary, additional reference frame info from other parts of
     the MS (e.g. time and position for AZEL measures).
 17. Add optional String columns MOLECULE and TRANSITION to the
     SPECTRAL_WINDOW table.
 
 
     Expected Impact:
 
 Positive: 
 1. No longer used (Correctors ax'd)
 2. Avoid bloat or kludge with zero dimensions in single dish case.
    (Using 2 tiled StorageManagers it should be possible to mix
     float and complex data in one MS if required [Multibeam data])
 3. Make it easy to combine single dish and synthesis data when
    DATA is optional
 4. Not used
 5. Not used, and not always uniquely definable (see example in
    FITS Format for Interferometry Data Interchange by Diamond et al.) 
 6. Not used
 7. Consistency
 8. Needed for X-Y receptors to establish local naming convention
 9. Needed as time origin for the _RATE columns; note that TIME will NOT be
    an index/coordinate
 10. Needed to point back to the OBSERVATION table
 11 Most fillers don't have the info to fill this table available to them,
    so instead of dummy-filling the fields we can make the table optional.
    The SOURCE_ID in the main table is not needed, it can be found via
    the FIELD table (each field belongs to a single source). 
 12. Remove duplicate info (already in SPECTRAL_WINDOW table)
 13. It is a scalar quantity, so speed is the correct name
 14. Presently this is used only by the imaging code, but as a scalar
     quantity. This change makes that correct usage.
 15. Adds much needed info on the reference frame of stored coordinates 
 16. Gives easy access to the stored MEASURE_REFERENCE 
 17. Provides a standard place to store these values (present in FITS)

 Negative:
 2. Some more run-time checking may be needed
 8  Some changes to existing imaging code needed
 2,4,5,6,8-15. Fillers will need updating
 14. VisibilityIterator will need updating (but simplifies) and 
     the direct mapping with FITS is lost. 
 
 I intend to fix up checked in (& unlocked) code.
 
 
 Proposed documentation changes:
 
 The MeasurementSet header files and the MeasurementSet Definition Document
 will be updated to reflect the changes.