Testing log - 2007.06.25
Steven T. Myers

-----------

Test using Jupiter 6cm Synth2000 tutorial data:
/home/sandrock2/smyers/NAUG2/Data/VLA_CONT/FLUX99-6CM.CBAND
or
http://www.aoc.nrao.edu/~smyers/naug/Data/tut2000/VLA_CONT/FLUX99-6CM.CBAND

Machine: sandrock
/home/sandrock2/smyers/Testing2/Jun07/

-----------

flagautocor : 

    way too verbose to logger - control with verbose=True/False

    do we need a separate flagautocor since you can do the same
    thing in flagdata?  I guess its a handy shortcut?

flagmanager :

    needs to have <msname>.flagversions already there
    (should be able to create if missing)

    is there any way to know what the current one is?
    there should be an entry in the .flagversions table
    that points to the "current" version (either the one
    made with a "save" operation or "restore")

plotxy : 

    should have "Quit" or "Done" button on plotter
    NOTE: killing the window has bad consequences requiring
    restarting casapy

    should not have to return at prompt exit flag mode
    and it seems it doesnt really exit as you can still
    hit Next etc. and keep going!

    in enableflags=False mode, it goes back to the
    casapy prompt in a semi-async mode. Should this
    behavior be controlled by async parameters like other
    tasks?  It can be confusing (since clicking buttons
    like Next will cause the CASA prompt to advance, etc.).

    should list hot-keys (like ESC) to shell or logger
    on startup

    all buttons should have flyover help

    default multicolor=True

    name of plot window should be "CASA Plotxy Window 1" or
      something similar, rather than "CASA Figure 1"

    correlation should be under selectdata
      (eventually replace selectsp with standard spw selection)

    correlation = 'RR,LL' preferred to 'RR LL' (both should parse)
    (2007-06-26 even 'RR LL' no longer works)

    with multicolor = True, should have legend for color meanings
    or at least send to shell or logger

    no reason to have different axes options for xaxis and yaxis

    all flagging tasks should behave the same re: flagversions
       1. when it makes a new <msname>.flagversions it should save the
          original flags (e.g. as versionname='initial')
       2. should be a parameter flagversion=True which if True will
          will save the current set (if not already saved) before 
          flagging, and if the flagging actually flags (or unflags)
          data then at end will save a new version    

    can get rid of
       enableflag
          flagregion (relegate to flagdata, or maybe have a Set Region
             button that asks for a user-input region)
          showflags ==> standard parameter
          unflag (is a button)
       flagmode

    I would be tempted to put most of the long list of plot control 
    parameters under a 'selectplot' parameter
       subplot,plotsymbol,plotcolor,markersize,linewidth,connect,
       plotrange,multicolor,overplot,replacetopplot,removeoldpanels,
       fontsize,windowsize

    what does iteration = 'antenna' do?  Looks like it plots ALL
    baselines to that antenna. NO - it behaves as 'antenna1'
    and only plots baselines from that antenna to higher ones.
    This is ok.  Should we have 'antenna2' option to plot only
    from that antenna to lower ones?

    correlation should be an iteration option

    should (like miriad) have separate iteration and panel controls,
    e.g. iteration='antenna', multipanel='correlation' would have (up to)
    4 panels for the correlations and Next would step over antennas
    (interaction with subplot might be tricky)

    does selectdata=False prevent any selection via the sub-pars or
    just hide them in the interface?  it seems to just hide them
    How do we want it to behave?

    ----------------------------------
    There needs to be a way to extend the flagging to all correlations
    even if only one is plotted.  In particular, the cross-hands RL LR
    are often too weak to flag on, so you flag on RR LL and want the
    corresponding RL LR to go away also.  I suspect flagdata in the
    'extend' mode is meant to do this.  But should add to plotxy:

    extendflags = True
       extendcorr = False    # extend flag to all correlations 
       extendchan = False    # extend flag to all channels (this spw if extendspw=False)
       extendspw = False     # extend to all spw
       extendant = False     # extend to all baselines with this Antenna1 (if iteration='antenna')
       extendtime = False    # extend to all times

    These would be done on an integration by integration basis
    (ie. extend to relevant scope for each integration flagged by
    action) except for extendtime.

    ----------------------------------

flagdata :

    way too verbose!  control with verbose=True/False

    does not seem to use the <msname>.flagversions (or at least create
    one if it doesnt exist).  All flagging tasks (plotxy,flagdata,viewer)
    should behave the same way

    if you set mode='save' it doesnt come up red

    if you set mode='summary' it will coredump after saying 
    "Sorry - not yet implemented !"

    if you run mode='query' w/o any selections set it says
    "Please run setdata with/without arguments before any setmethod"
    (clearly some toolkit messages)
    but then if you do, it coredumps

    setclip/setquack too clunky and obscure.
    ----------------------------------
    could instead have a mode='clip' option (along with mode='quack'
    and maybe mode='flagautocor') which opens up clipexpr,clipminmax,outside
    though this would limit to one (potentially long) type per use.
    Instead, could also move to standard params so that autocor,
    quack, clip, autoflag can be done in parallel (in that order when
    it matters) and mode would control whether it flags, or reports
    what it would do when flagging or whatever.

    flagautocor = True

    setquack = True
       quackinterval = 0   # <=0 use scan indexes
       quacklength = 0     # <=0 first integration of each scan
       quackmode = 'beg'   # 'beg','end','endb','tail' (See AIPS QUACK)
       
    setclip = True
       clipexpr = ''
       clipminmax = []
       outside = True

    autoflag = True
       alg = 'timemedian'  # timemedian,freqmedian,sprej
       threshold = 0       # Flagging threshold (n sigma)

    mode = 'flag'          # flag,summary,query,extend


    I'm not sure what mode='extend' does (supposedly lets you extend
    the scope of flagging), so I dont know if it should be a standard
    parameter or not.
       
    ----------------------------------

    When trying to flag using the setclip outside=False (ie. within a 
    range), e.g. setclip = ['ABS RR',[0.0,1.e-10],False]
    you get
    "*** Error *** RecordInterface: field flagrange is unknown"

viewer :

    why a filetype parameter?  it must surely be able to tell between
    an ms and an image! not sure what a "vector" is.