create_visibility_from_ms

ska_sdp_datamodels.visibility.create_visibility_from_ms(msname, channum=None, start_chan=None, end_chan=None, ack=False, datacolumn='DATA', selected_sources=None, selected_dds=None, average_channels=False)[source]

Minimal MS to Visibility converter

The MS format is much more general than the RASCIL Visibility so we cut many corners. This requires casacore to be installed. If not an exception ModuleNotFoundError is raised.

Creates a list of Visibility’s, split by field and spectral window

Reading of a subset of channels is possible using either start_chan and end_chan or channnum. Using start_chan and end_chan is preferred since it only reads the channels required. Channum is more flexible and can be used to read a random list of channels.

Parameters:

  • msname – File name of MS

  • channum – range of channels e.g. range(17,32), default is None meaning all

  • start_chan – Starting channel to read

  • end_chan – End channel to read

  • ack – Ask casacore to acknowledge each table operation

  • datacolumn – MS data column to read DATA, CORRECTED_DATA, or MODEL_DATA

  • selected_sources – Sources to select

  • selected_dds – Data descriptors to select

  • average_channels – Average all channels read

Returns:

List of Visibility

For example:

selected_sources = ['1302+5748', '1252+5634']
bvis_list = create_visibility_from_ms('../../data/3C277.1_avg.ms',
    datacolumn='CORRECTED_DATA',
    selected_sources=selected_sources)
sources = numpy.unique([bv.source for bv in bvis_list])
print(sources)
['1252+5634' '1302+5748']

Issues with Polarisation bug YAN-1616

The MS Format permits the order of the polarisations to essentially be arbitrary. As long as they are explicitly defined in the CORR_TYPE column.

This loader made the assumption that the input order is XX,YX,XY,YY even if it is not as it used “sorted” on the CORR_TYPE column.

This created bug YAN-1616 which has been fixed by removing the sorted() and introducing the linearFITs frame.

I have left this comment here as: * This change did not break the existing unit tests. * I am concerned there is a use case that relies on

the sorted() and will now break.

Steve Ord - 2024 12 December