Version 1.9 Build 000

Latest News FAQ Search Home

AIPS++ Development Plan

Introduction

This document describes and records the development plan for AIPS++ from March 1995 (when new management of AIPS++ was instituted) until mid-1996. The plan will evolve and change with time. Revisions since the last version are shown in bold. Items subject to continuing discussion are given in italics.

First I give some some general observations that drive the plan:

• The goal of AIPS++ is to be the next generation software system for astronomical data analysis. The strategy to acheive this goal is to follow three phases:
  1. Develop the infrastructure library,
  2. Test and extend the library. We will do this by tightly focussed applications development. To attract early adopters of AIPS++, we will preferentially add key, unique applications that do not exist in other packages,
  3. Expand out to completeness.
  Phase 1 is mostly complete. We are embarking upon the second phase in which the infrastructure will be tested and extended. This will occur principally at the two main NRAO sites (Socorro and Charlottesville). Once the libraries and central design concepts are shown to be robust, applications development (Phase 3) can accelerate and spread to more AIPS++ sites.
• Adoption of AIPS++ by applications programmers requires robust, reliable, well-designed, well-documented libraries. It would be irresponsible to develop designs that are not documented. This will a key part of the AIPS++ intellectual heritage. The completeness of the documentation of the early design of IRAF serves as a good example.
• I expect that the libraries will change and grow significantly as applications are written. This requires close communication between application developers and those responsible for the library. The distinction between infrastructure and application code is hard to make at this early stage since most applications should produce code that will eventually end up in the library.
• To maintain focus and to provide simple feedback concerning the library, applications development should proceed in a few tightly focussed areas. These applications will be drawn from single dish support (specifically GBT commissioning and observing), and synthesis support. Once these applications have been demonstrated, we expect that further application development will proceed more rapidly and with the active involvement of other sites.

AIPS++ is conceptually broken down into various subcomponents: Infrastructure, System, Documentation, Single Dish Processing,Synthesis Processing, Image Analysis,Visualization, Observing support, and Management.

Infrastructure

• Object persistence: A mechanism for allowing object persistence is required.
• User Interface: We will continue to use Glish as the CLI. There is a separate development plan for Glish.
• Tasking: We will develop further and implement the design for Tasking described in Glendenning's memo on the architecture of AIPS++. As part of this work, the Inputs Classes must be redesigned to support more complex applications e.g. convert input file name to an image directly.
• Graphical User Interface: Paul Shannon will be responsible for the adoption of an existing GUI or the development of a prototype GUI. Harvey Liszt and Rick Fisher will advise. This will be done in partnership with the GBT Monitor and Control group.
• Coordinates: The ATNF group will be responsible for the completion of the design and implementation of Coordinate and Measure classes.
• Image Class: The Image class will be integrated with the new Coordinate and Measure classes. Masks and weights will be added.
• Plotting Capabilities: 1-D plotting is needed in some form for the GBT work. For the GBT Systems integration work, we will use a commercial widget, Xrt/Graph. We will evaluate alternative, public domain solutions. Paul Shannon will be responsible for this with advice from Harvey Liszt and Rick Fisher.
• Table development: Storage managers and optimization will be the key to efficient use of the Table system, particularly in the "Big Table" model of Measurement Sets. This work will be under the supervision of Ger van Diepen. The key remaining development is of a hypertiled storage manager.
• FITS Classes: The FITS classes will be augmented to support random access i/o, to avoid the downcasting now required, and to improve the interface.

System

• Code Management System: We will add support for automated testing.
• Compiler Support: We will evaluate the beta version of the Sun native compiler. As a second priority, we will also investigate use of the IBM compiler.

Documentation

• Overall Documentation: Much of the existing documentation is out-of-date. We will attempt to update key documentation. We will distinguish between up-to-date and outdated documentation.
• User Documentation: User documentation is proceeding well under the direction of Alan Bridle. Decisions must be made on directory structure and development of tools.
• Programmer Documentation: One person will be given overall responsibility for programmer documentation and will work in close collaboration with Alan Bridle to provide a unified approach to user and programmer documentation. Design documents must be regarded as a necessary part of the development of a new class or of a substantial revision of an exisiting class. For major existing classes, we will write post hoc design descriptions. Defunct design documents will be removed from active directories.

Single Dish Processing

• On The Fly Mapping: We will finish the OTFTool developed by Bob Garwood for the 12m and support it for use at the telescope. However, substantial additions or augmentations will take second priority to the GBT support discussed below.
• GBT Commissioning support: AIPS++ will support the GBT systems integration work on the 140' telescope. This requires filling into an AIPS++ Table, plotting and some simple manipulations of engineering and astronomical data, all by July 1995.
• SDCalc: After support for the commissioning of the GBT, we will start development of a Single Dish Calculator of the sort envisaged by Harvey Liszt and Rick Fisher. This will require development of a GUI and plotting capabilities.
• SDImager: After SDCalc, we will develop imaging capabilities for Single Dish Observing in both spectral line and continuum modes. These will be analogous to the advanced imaging capabilities important in interferometry.

Synthesis Processing

• Advanced Imaging: We will develop tools for advanced imaging using the Briggs NNLS deconvolution algorithm in conjunction with self-calibration and careful editing of visibility data.
• Calibration: We will develop a complete calibration package for the ATCA. Calibration for other telescopes is deferred until a workable scheme for the ATCA has been developed. Calibration for BIMA, VLBA, WSRT is tentatively planned to be addressed early 1996.

Image Analysis

• Image Analysis will be deferred until the initial applications in Single Dish and Synthesis processing are complete.This is likely to be until early 1996.

Visualization

• Visualization will continue to be developed at BIMA sites. The code will be checked into the AIPS++ Code Management System. AIPS++ Classes will be used where possible. A second round of revisions and augmentations to AIPSView will be made with the goal of supporting the planned applications development (for example, AIPSView will supporting editing of visibility data)

Observing support

• The VLBA Observe program (also known as Gnomes) will be ported to AIPS++ starting Fall 1995.

Management

• This plan is predicated upon the following manpower being available and dedicated to AIPS++ with high predictability but not necessarily 100% time allocation.
• The AIPS++ group inside NRAO will be composed as follows:
  • Socorro: Cornwell (Project Manager), Glendenning (Deputy), Roberts and two new hires will work on Management, Infrastructure, and Synthesis Applications. Hjellming is a scientific advisor. I expect several long-term visitors to be present in Socorro at any time over the next few years. Mark Wieringa will visit the AOC for at least 6 months starting in September 1995 to work on ATCA calibration. In Fall 1995, Young will join the AIPS++ group. In early 1996, Beasley wil begin work on VLBA calibration inside AIPS++.
  • Charlottesville: Garwood, Shannon, and Schiebel will work on Single Dish and Infrastructure. Bridle is working on User Documentation. Liszt is a scientific advisor of Single Dish.
  • Green Bank: Payne will support Single Dish. Fisher is a scientific advisor for Single Dish and Infrastructure.
  • Tucson: There is currently no AIPS++ support person at Tucson. Darrel Emerson is a scientific advisor for Single Dish.
• The AIPS++ group at ATNF will be: Killeen (Local Manager), Brouw, Calabretta, and Wieringa.
• The AIPS++ group at BIMA/NCSA will be: Crutcher (Local Manager), Baker, Baxter, Ravlin, and Pixton.
• The AIPS++ group at NFRA will be: Noordam (Local Manager), van Diepen, and Olnon.
• The other AIPS++ sites (HIA/DRAO, NRAL, NCRA) are expected to provide substantially less than 1 person each to this plan.
• The Project Management will report quarterly to the AIPS++ Board on progress in the last quarter and plans for the next quarter.
• Plans for AIPS++ development will be described in this document which will be updated as required.
• Tracking of progress in AIPS++ will be via a Target Dates list that will be updated and distributed to
  aips2-workers
  at the beginning of every week.

Copyright © 1995,1999,2000 Associated Universities Inc., Washington, D.C.