casagui 0.3.110

visualization toolkit and apps for casa

This is a pre-alpha, prototype package. It is not useful for external users, and all applications being built with it are currently in various phases of prototyping.

Introduction

For some time, the GUIs provided by CASA have been based upon Qt. While Qt works well, the compiled nature of C++ code made building and distributing the GUIs for each architecture a hurdle. This in turn caused the GUIs we developed to tend toward large, monolithic applications which were difficult to integrate and control from Python. We first used DBus to control our Qt application. Qt provides a nice interface to DBus, but it became clear that DBus development had slowed and that DBus was unlikely to make major inroads outside of the Linux Desktop. At that point, we switched to gRPC. gRPC supports a variety of platforms and languages. It also has significant support behind it. However despite the improved technology, it was still difficult to incorporate a scripting interface which allowed a stand-alone C++/Qt process to be controlled by a separate Python process at a low enough level to be practically useful for control at the level of granularity we desire.

Similar to the CASA visualization development experience, the CASA framework as a whole has experienced the ups and downs of the large C++ development experience. Experience with a Python parallelization trade study which CASA conducted indicated that the loss of CPU throughput in a switch from C++ to pure Python can be made up for in gains made in the selection of parallelization framework like Dask along with just in time compilation with something like Numba. In addition to the focus of the trade study, additional gains are possible by mixing in GPU resources.

These experiences have led CASA to begin a multi-year transition from being a large C++ framework attached to Python to being a pure-Python framework for processing radio astronomy data. This package is visualization portion of that transition.

After an abbreviated trade study where we considered a few pure-Python visualization frameworks, we selected Bokeh as the basis for creating new visualization infrastructure for CASA. The choice of Bokeh was made based upon its extensibility, its community support (including NumFocus), and its limited external dependencies (just JavaScript and a modern web browser). A stand-alone application can be created by using the Bokeh server. These options allow for GUIs to be created and used interactively from a Python command line session, as a stand-alone mini web server, integrated into a desktop application (using Electron) or as part of a Jupyter Notebook.

Beyond this architectural flexibility, our intention is to create a toolbox of Bokeh based components which can be combined to create a collection of visualization tools which can be used in each of these settings (Python command line, Notebook and desktop application) so that we maintain smaller, reusable tools instead of very large monolithic applications. Interactive clean is our path-finder application of this approach and is currently the only example available.

Installation

casagui is available from PyPI.

Requirements

• Python 3.8 or greater

• casatools and casatasks built from CAS-13743

Install

• bash$ casa-CAS-13743-2-py3.8/bin/pip3 install casagui

Caveats

• Remote access is slow, later a desktop application will be developed (using the same Bokeh toolbox) to improve this situation, but for now if running remotely, it is best to pre-start your preferred browser on the host where you will be running interactive clean. For example

  • bash$ export BROWSER=/opt/local/bin/firefox

  • bash$ $BROWSER > /dev/null 2>&1 &

• Konqueror does not work. We only test with Chrome and Firefox.

• node.js version 14.0.0 or higher is required

Simple Usage Example

A simple example of the use of interactive clean is:

CASA <1>: from casagui.apps import InteractiveClean
CASA <2>: InteractiveClean( vis=ms_path, imagename=img, imsize=512, cell='12.0arcsec',
                  specmode='cube', interpolation='nearest', nchan=5, start='1.0GHz',
                  width='0.2GHz', pblimit=-1e-05, deconvolver='hogbom', threshold='0.001Jy',
                  niter=50, cycleniter=10, cyclefactor=3, scales=[0,3,10] )( )

In general, the InteractiveClean constructor takes a subset of parameters accepted by tclean. All of the masks used in running interactive clean are available from the InteractiveClean object. To get access to the list of masks, you would create the object as a separate statement:

CASA <2>: ic = InteractiveClean( vis=ms_path, imagename=img, imsize=512, cell='12.0arcsec',
                  specmode='cube', interpolation='nearest', nchan=5, start='1.0GHz',
                  width='0.2GHz', pblimit=-1e-05, deconvolver='hogbom', threshold='0.001Jy',
                  niter=50, cycleniter=10, cyclefactor=3, scales=[0,3,10] )( )
CASA <2>: ic( )
CASA <3>: print(ic.masks( ))