--
NicholasElias - 2010-06-18
Miriam Krauss (mkrauss at nrao.edu)
Summer school issues evaluated for inclusion in/establishment of JIRA
tickets.
Key:
TICKET -> a new JIRA ticket has been/will be created
COMMENT (CAS-*) -> a comment will be added to CAS-*
CABAL -> needs to be circulated to CSSC
EXPERT -> perhaps a ticket should be created, but if so, this should be
done by an expert on the topic
RESOLVED -> already resolved outside of JIRA
NONISSUE -> not really an issue
CAS-* -> already in JIRA with this issue number
clean:
CABAL -> clean will (quietly) ignore parameters it's been
given (such as imsize) if it finds an old set of files with the
same imagename, which leads to substantial user confusion. Unlike
what is expected (create a new image of a different size or
resolution), it will simply continue to clean the old image. This
behavior is not documented in the help file.
Preferred behavior:
- Add a boolean 'continue' parameter.
- Add the boolean 'overwrite' parameter as per CAS-2288.
- Both parameters will be 'False' by default.
- The behavior should be (assuming that the named output files
alread exist):
If overwrite=F: exit with an error that the files exist and
overwrite=F.
If overwrite=T and continue=F: start all imaging & deconvolution
from scratch with given parameters, overwrite existing files.
If overwrite=T and continue=T: check that the input parameters
for cell & image size match the files on disk. If yes, then
continue deconvolution and overwrite as needed. If no, then exit
with an error that the parameters do not match.
- Note that all checks relating to 'continue' and 'overwrite'
should be performed *before* the task does anything else
Note that this is related to CAS-1690, though I believe the
description above proposes a more general solution to the problem.
It should also be noted that many people have run into problems
with the current behavior, so the priority should probably be 'Major'.
NONISSUE -> Threshold is the same for all image planes, which may result
in some not getting cleaned.
This likely arose because clean boxes were not made for all image
planes, resulting in only 'I' actually getting cleaned.
TICKET (CAS-2304) -> clean currently assumes no clean box if it is run
interactively. Therefore, no cleaning is performed until a box is
provided. This behavior should be changed to clean the whole image
if no box is drawn.
TICKET (CAS-2305) -> clean boxes must be separately drawn for each
Stokes plane, or cleaning will only occur in Stokes I. This
behavior is already implemented for frequency channels, with radio
buttons to select "Displayed Plane" and "All Channels". To account
for both channels and Stokes planes, this should be expanded to
four buttons, two for channel and two for Stokes planes. Recommend
"Displayed Channel", "All Channels", "Displayed Stokes Plane", "All
Stokes Planes".
ACTION -> need to clarify 'clean' behavior for multiple image planes
when run non-interactively
CAS-1938 -> Need to be able to select by channel range.
plotting & imaging:
TICKET (CAS-2306) -> plotants has flag buttons, which flag entries in
the ANTENNA subtable, but this has no effect on the data flags
themselves, as some users expect. These buttons need to be
disabled until (and if) we support data flagging with the ANTENNA
table
TICKET (CAS-2307) -> Improvements for the editing and saving of flags
in plotms:
1) It would be good to have flagging behavior like what is in the
viewer, where flagging edits are not written to the MS until a
"Save Edits" is requested. Also, the ability to "Undo Last Unsaved
Edit" and "Undo All Unsaved Edits" would be very useful.
Currently, the "Undo Flags" button behavior does not work for this,
since it will undo *any* flags that apply to the selected region,
whether or not it was applied by the current plotms session. The
only workaround is to judiciously us 'flagmanager' to save flags to
disk while editing the flags with plotms.
If these improvements are to use the current icon scheme at the
bottom of the GUI, the red flag icon should create an edit (but not
save to disk), the 'unflag' icon should undo the last edit, and
there should be a save-flags-to-disk (MS) icon. There should also
be an 'undo all unsaved flags' icon, perhaps several flags with a
slash through them.
2) There should be a way to save the current MS flags to a backup
within plotms. This could be in the "Flagging" tab, and should
have a place to enter the desired filename.
CAS-1144 -> ORIGINAL ITEM: It would be helpful to be able to
save and load text-formatted regions with viewer.
NOTE: this is already being worked on.
RESOLVED (CAS-202) -> viewer: image statistics should have units.
Related to CAS-202.
RESOLVED -> immath: it is not clear how the image label is propagated;
appears to be the label for the first image in the expr string. If
multiple labels are in the input images, the output image should
probably not have any label (since it would be very tricky to
create a new label appropriate to the function in 'expr').
COMMENT (CAS-1032) -> Would be nice to have a task interface for toolkit
functionality to rotate & bin image cubes to get, e.g.,
source-averaged position-velocity diagrams.
NONISSUE -> plotms colorization by antenna1/2 is only useful for a
subset of antennas with numbers >< the reference antenna. Would be
useful to have a scheme that gives a unique color for each baseline
to the reference antenna. Note: this is what the 'baseline'
coloration will do, given a single antenna name.
script:
RESOLVED -> The screenshot to describe making clean boxes says to
click on button circled in green; should be 'red'.
RESOLVED -> NOTED: The two gaincals should [COULD] be one.
RESOLVED -> Should add a word of caution about self-cal on a mosaic,
since it's not guaranteed that all fields will have good soln's.
RESOLVED -> polcal should be run with careful gainfield selection for
the gain calibration on the polcal source.
flagging:
NONISSUE -> quack with mode='beg' will also chop off the last bit of a
previous scan if the fields are identical.
Could not reproduce.
TICKET (CAS-2308) -> There should be a 'flagbackup' parameter in
applycal, so one can save a version of the flags before calibration
(and possibly new flags) are applied. The inputs could be the same
as what is currently implemented in flagdata, and the output files
could be named 'applycal_1, applycal_2' etc. Priority is set to
'Major' because many users do not realize that data are flagged
when they run applycal: even if they choose not to save a version
of the flags, having a 'flagbackup' parameter present will raise
awareness of the process. Along these lines, it would also be
useful to add information to the help file, e.g. "Applycal will
overwrite existing corrected data, _and flag data for which
corresponding calibration products are not present_".
EXPERT -> Flag versions should be internal to the MS.
TICKET (CAS-2309) -> Flagdata should record the flagging commands in
the MS history. The recorded commands should contain enough
information to reproduce the flags that were applied, and this
information should also be retrievable by flagmanager with a
mode='history' command, using the MS name as input.
-> Consistency with tools writing useful information to MS
history.
calibration:
NONISSUE -> solnorm=T does nothing in phase-only gaincal
Does not seem that this is unanticipated behavior, since the help
file states that solnorm will act on amplitude (not phase)
solutions.
TICKET (CAS-2310) -> The choice of 'refant' (used by bandpass,
gaincal, polcal, [fringecal]) should be cleverer in the case where
an alternate reference antenna is necessary. The alternate should
be near the requested one, which could be accomplished by using the
positions listed in the ANTENNA table. Similarly, it would be
ideal to have the default reference antenna be one near the center
of the array rather than '0'.
documentation:
TICKET (CAS-2311) -> The help file for MS selection syntax (see memo
"http://www.aoc.nrao.edu/~sbhatnag/misc/msselection/msselection.html"),
par.selectdata, could use a minor update to add the
auto/cross-correlation syntax ('&&&' and '&&') and perhaps the use
of '>' and '<' in the spw and scan parameters. Also, all the help
files for par.* which could use the MS selection syntax (par.scan,
par.spw, etc.) should cross-reference the par.selectdata help file.
For example, par.scan currently reads:
scan()
Scan number range - underdevelopment.
Note that par.uvrange already provides this reference.
casa (general):
RESOLVED (CAS-2288) -> Should have a universal 'clobber' parameter for
tasks that write output file(s). Currently some have this (e.g.,
overwrite in exportfits), but most don't. The behavior for any
task that will write a file should be:
1) Check for output file(s) with the same name that the task will
try to write.
2) If any are present check the clobber parameter. If True,
continue. If False, exit with an error message that the user has
requested the same file name but clobber is set to False.
3) Perform the task and create output files as needed.