Science
Articles:

CO in
Serpens

AIPS++ in
the RET
Program

Multi-Scale
Clean for
Extended
Structure

A Layman's
Experience
of Glish/AIPS++

The image shown is the nearby (distance=220 pc) stellar factory in Serpens as seen in its integrated C¹⁸O J=1-0 emission, as observed by the NRAO 12-meter telescope. This molecule traces the overall column density structure in the region. The backbone of the cloud is an arc curving through roughly 180 degrees, with a length of 30'. Two main peaks in the C¹⁸O emission are located near the center of the arc, separated by 4'. The Submillimeter and far infrared continuum emission from the region are dominated by sources that lie primarily *between* the two C¹⁸O peaks, including a cluster of 50-140 young stars. A first moment map of the C¹⁸O emission displays a broad velocity gradient running from a velocity centroid of 9.0 km/s at the northwest end to 7.5 km/s at the southeast end. It is unclear whether this gradient arises from rotation, two clouds at different velocities or interaction of cloud material with recent star formation.

In the paper on this source published earlier this year (McMullin et al. 2000, Ap. J., 536, 845), five different software packages were used to examine the array of data: 1) AIPS for reduction of NRAO 12-m OTF data, 2) CLASS and 3) COMB for analysis of single dish spectra from the CSO and NRAO 12-m, 4) MIRIAD for analysis of the BIMA interferometric data and 5) WIP for fashioning the final, publication-quality figures. While the calibration had been done, I threw in a sixth package, AIPS++, to explore better, more streamlined avenues for reducing a diverse set of data, which is typically the norm rather than the exception in observational astronomy.

Overall, I was pleased with the capabilities of the images and viewer modules in particular. It was gratifying to obtain the same answers as I had measured in MIRIAD (which I'm quite fond of). In addition, I found these tools better for exploring the data; interactively examining the structure on both the large and small scales is one of the more versatile tools handed to you in the AIPS++ GUI. The ability to script (or construct a command line record of your sundry GUI button clicks) was also a valuable addition for streamlining future reductions. In general it was my experience that the basic toolkit has grown to house quite an array of analysis tools, so it's more a question of finding the name of the task you need than whether you can perform an operation.

I look forward to completing my next data set as the previous battery of packages can now be replaced, from calibration->imaging->figure production, with AIPS++.

Science
Articles:

CO in
Serpens

AIPS++ in
the RET
Program

Multi-Scale
Clean for
Extended
Structure

A Layman's
Experience
of Glish/AIPS++

The HI luminosity-linewidth profile method is a valuable tool in radio astronomy. Among other things, it helps astronomers determine distances, rotation velocities, and hydrogen composition for distant galaxies. During my summer Research Experience for Teachers (RET), I worked with J. Richard Fisher at the National Radio Astronomy Observatory in Green Bank. This past October, Dr. Fisher used the Arecibo Radio Telescope to collect data for over 500 galaxies; my job was to reduce this data. The data helps to support research completed with the Infrared Astronomical Satellite and provide a comparison of luminosity-linewidth and supernovae Ia distance scales. In order to complete the task I had to first acquire a working knowledge of AIPS++ and Glish, the command and scripting language of AIPS++; Glish allows users easy access to all AIPS++ analysis routines and provides a means of writing customized applications by bundling together collections of commands into new functions. This past summer, Dr. Fisher and I wrote AIPS++ scripts that would implement data correction factors, display data scans, and assist in making precise measurements.

As a volunteer researcher, I plan to continue my work with this project during the coming school year. I will be installing AIPS++ on a Linux machine at my school in Oil City, PA using the latest version of AIPS++ on CD. Selected students will train and assist me in completing the data reduction process for these galaxies. In any science education program whether it be elementary school or college, it is crucial that students be provided the opportunity to experience science rather than to simply be told about it. If anything this summer, I've become even more convinced that science is active, not passive. Science is a vehicle by which man contributes to his understanding of the universe around him. Thanks to my RET experience, my students will be going on a little "drive" this year.

Science
Articles:

CO in
Serpens

AIPS++ in
the RET
Program

Multi-Scale
Clean for
Extended
Structure

A Layman's
Experience
of Glish/AIPS++

We have used the Multi-Scale Clean algorithm in AIPS++ in an effort to characterize the faint emission from radio halos. The negative bowl caused by the missing short spacings problem can have a serious impact on the analysis of such emission. Figure S2 shows a minor axis profile across the egde-on galaxy NGC 4631 from VLA C-array data at 1.4 GHz. It is obvious from the plot that the Multi-Scale Clean algorithm is superior to the standard Clark Clean in compensating for the effects of the negative bowl surrounding the source. The benefits gained in the outer regions of the halo are not insignificant, amounting to an increase in scale height of the extended component by almost 25 percent over the Clark algorithm. The addition of D-array data substantially improves the situation by adding more short spacings, but Multi-Scale Clean is still a marked improvement over the Clark Clean, all but eliminating the negative bowl.

Science
Articles:

CO in
Serpens

AIPS++ in
the RET
Program

Multi-Scale
Clean for
Extended
Structure

A Layman's
Experience
of Glish/AIPS++

Editor's note: This article traces the path of a non-programmer astronomer learning how to program in AIPS++ at the Glish level in order to perform customized scientific calculations. It frankly lays out both the frustrations one might expect to encounter on this path and the ease and power of Glish/AIPS++ programming once you have learned how to use these tools. The bottom line is best summarized by Jay's concluding remark: "I have found Glish/AIPS++ a major advance over other software tools that I have used and recommend it to anyone looking for a general purpose package for analysis of astronomical data."

Some months ago while trying to model a large HI cloud seen in 21cm emission and self-absorption against the galactic HI background, I was faced with the need to make some simple but rather extensive calculations. At Green Bank, I am surrounded by people using glish and developing AIPS++, so it was a natural temptation to play with these new tools for what I thought would be just a few day's worth of modelling. By the time the project was finished, months later, I had written several large glish scripts, each with hundreds of lines of code, and had analyzed several large spectral-line data sets using glish/AIPS++ functions. The results will appear in a forthcoming issue of the ApJ; here I have been asked to write about the experience and what lessons might be offered to others who would use the AIPS++ system as a general computing tool. I must warn readers that I have a very unclear understanding of how the tools I used fit into the greater scheme of AIPS++ and likely much of what I think I understand is incorrect. In particular, I will refer to glish and AIPS++ without a clear distinction between them because that is how I used them. I "backed into" AIPS++ by adopting a "bottom-up" approach, that is, by using glish in a primitive way, and only incorporating AIPS++ functions where necessary. But perhaps the lesson is that this system can be quite useful even if it is not understood at all.

I began the project in total ignorance of AIPS++ and glish, but armed with Rick Fisher's excellent Note Getting Started with Glish for AIPS++ and as important, with Rick himself just down the hall, within a short time I was manipulating arrays in the effortless way that glish encourages. In fact, if there is one major impression I have of working in glish, its that most things are either effortless or impossible. But more on this later. I was quite impressed on the ease with which simple radiative transfer problems could be coded in glish, and the results displayed using the built-in pgplot functions. It was a short step to incorporating AIPS++ functions in the code, creating AIPS++ images from arrays, convolving the images to mimic observations with a radio telescope, and so on. A few months later the results were in hand, and it had been a satisfactory experience. I now find myself beginning a totally different project, again using the Glish environment. It has become my default language for data analysis and modeling.

Here are a few general impressions derived from this experience:

• Glish is compact and powerful. With its basic functionality it is possible to do an enormous amount quite quickly. I was continually impressed with how straightforward and logical it was to do more or less exactly what I wanted. For example, given an array of spectral line measurements in various directions, one might want to form a smaller array of only the data with detections. It is simple in glish to say detections:=positions[Peak_T>3.0*sigma] and get an array of, well, of the good data! Rick Fisher's memo conveys this power quite well. When programming in glish, the AIPS++ functions are easily available and these make processes like convolution quite simple.
• The style of glish/AIPS++ is quite different from anything I had experienced before, and perhaps because I lack any understanding of the intellectual basis behind it, in places it seems somewhat weird. But it is, for the most part, consistent, and I was gratified to discover that once I learned a pattern it could be applied in many different circumstances. This is to say that glish/AIPS++ has a reasonably consistent set of rules.
• The online AIPS++ documentation is quite extensive though I found it almost uniformly obscure. Luckily, it usually contains enough examples to allow me to deduce the rules governing whatever function I need while sparing me the necessity of learning any comp-sci terminology.
• That glish is an interpreting language (rather than one which is compiled) has both advantages and disadvantages. On the plus side, I quickly developed the habit of trying out swatches of code in the interactive environment, testing to make sure that each variable had the correct value, then embedding the code into a function which operated in a sort-of batch mode during actual use. For most of what I did the loss of speed due to the continued interpretation of the glish scripts was not significant -- it takes so much time simply to get glish started on my workstation that I typically logged on to the fastest machine available and worked entirely there. However, if my programs get any more complex than the current ones, the speed of glish will probably become a factor, and I may need to optimize the run-time of certain small blocks of code.
• Errors: Whatever you do, don't make a mistake in Glish. It's error-reporting is crude and often less than helpful. For example, if you forget to supply all the parameters in a function call, it lists the entire function as part of the error message. Also, an error which involves an array can generate a number of error messages equal to the length of the array. I am making this sound worse that it actually is, for in a relatively short time I developed a bit of sense for deducing the true error from the Glish spasm. Still, some errors can still be maddeningly difficult to track down. Particularly insidious, though, are errors which glish does not consider to be errors at all. This is the topic of the next section.

  Editor's note: Glish code checked into the system is pretty good at handling errors. Because of Glish's flexible typing, a variable which usually has an array in it might be set to a "fail" value if an error was made, the is_fail() function can detect the failure of the array, and the glish script can orchestrate an appropriate response. More advanced user scripts should utilize this sort of error checking as well. On the topic of Glish's habit of printing out lots of stuff, this is often used as a "feature": for example, if you can't remember the name of some function of the Table tool, just typing the name of your Table tool will list all of the functions defined for Table. Similarly, typing the tabletoolname.functionname() will show you the parameters that function requires.

• Gotcha! Sometimes things didn't work out the way I expected, and this was typically the largest source of wasted time in my experience with glish. An example that I ran into just today occurred while I was trying to integrate a number of spectra over some minimum to maximum velocity, with the constraint that the minimum should be at least 100 km/s . Given an array of minimum velocities I wrote min_vel:=max(min_vel,100). Well, the variable min_vel goes into this statement as an array but comes out as a scalar. This legerdemain is actually discussed in the glish documentation, but unless you are the person (I am told there is one) who reads all relevant documentation before beginning to write code, you might feel, as I do, that this is a fault rather than a virtue of the language.
• I/O. This can be quite clunky. You should see what I have had to go through to read the ASCII text output of a spread sheet into useful arrays in glish. It is my impression that glish is comfortable only with string I/O, which is ok if all you pass is messages, but less useful if data needs to come and go.

  Editor's note: while ASCII I/O is not one of Glish's strong points, the tablefromascii() and toascii() Table functions are extremely useful for dealing with ASCII data. For this path to be fruitful, the casual user will need to learn how to manipulate AIPS++ Tables, but given the power of the Table system, the good documentation, and the "cut, paste, and test" aspects of Glish, this is not a great hurdle.

• Bug reporting: Everyone should have the experience of submitting a formal bug report to the AIPS++ community, for it gives one the impression of having pulled a cosmic fire alarm. Email messages fly back and forth, progress reports appear, and I could not shake the feeling that somewhere, deep under a mountain in Colorado, lights on a gigantic panel had turned from green to red. Within a few days the bug was reported to be fixed. This experience is one which I can recommend quite highly.

If you are considering giving glish/AIPS++ a try, here are a few suggestions that may ease your way: