Installing¶
cubevis needs Python 3.12 or later. Each application has its own extra,
so that installing one does not pull in what only the other needs:
To use |
Install |
Brings in |
|---|---|---|
|
|
xarray-ms, zarr, dask, distributed, datashader |
|
|
casatasks and casatools, arcae, xarray |
both |
|
all of the above |
only the window, with the data on another computer |
|
what is needed to start and talk to a remote kernel |
The visplot and iclean extras include remote.
Add the notebook extra to use iclean in a Jupyter notebook:
pip install "cubevis[iclean,notebook]".
Where visplot can read data¶
visplot reads MSv2 measurement sets and MSv4 / Processing Sets through
xarray-ms and its casacore bindings, arcae. arcae is published for
Linux (x86_64 and aarch64) and macOS (Apple Silicon on macOS 14 or later,
Intel on macOS 15 or later). There is no Windows build.
On a computer where the data cannot be read, or when the data is on another computer, the data access runs there and only the window runs on your computer: see Setting up remote execution.
How iclean reads images¶
iclean runs CASA’s tclean through casatasks, but reads the residual
and reads and writes the mask with arcae, not casatools. It reads CASA
images (paged, or concatenated with ia.imageconcat) and
xradio images saved in Zarr. arcae has a
build everywhere casatools does, so nothing needs to be set.
To read the images with casatools instead, set an environment variable
where the images are (for a remote kernel, in that kernel’s environment):
export CUBEVIS_IMAGE_STORE=casatools
With CUBEVIS_DEBUG=1, the log says once which one is in use. If arcae
or xarray cannot be imported, casatools is used and the log says so.
The same image access works on its own, with no CASA installed
(pip install "cubevis[image]"): xr.open_dataset(path, engine="casa-image")
after import cubevis.xarray.image, or cubevis.imaging.ImageReader.
Checking the installation¶
from cubevis import visplot # or: from cubevis import iclean
The applications are loaded on first use, so an import error for a missing package appears here and names the package.
cubevis.docs() then opens the documentation for the version installed.