When using a custom visibility column (e.g. DIR1_DATA), the code doesn't know up front that the dims are (row, chan, corr). The data mappers, in particular, need this info. It can be supplied to the table_schema argument of xds_from_table (as table_shema=["MS", dict(...)].
Suggest adding an optional --column-dims argument. If not given, assume a shape of (chan, corr) for columns ending in _DATA, or complain for other unknown array columns.
@Athanaseus was bitten by this one, so time to prioritize it up. In the meantime @Athanaseus please try the issue-80 branch which has a kludgy terrible temporary workaround.
When using a custom visibility column (e.g.
DIR1_DATA
), the code doesn't know up front that the dims are (row, chan, corr). The data mappers, in particular, need this info. It can be supplied to thetable_schema
argument ofxds_from_table
(astable_shema=["MS", dict(...)]
.Suggest adding an optional
--column-dims
argument. If not given, assume a shape of (chan, corr) for columns ending in_DATA
, or complain for other unknown array columns.Standard known columns are in this list: https://github.com/ska-sa/dask-ms/blob/master/daskms/table_schemas.py