The use of APL versus other languages—a six year trend analysis
John W. Myrna, Peter G. DiChellis · ACM SIGAPL APL Quote Quad · 1985
Like everyone in the APL community, we've heard that all-too-familiar chorus: APL is a very small piece of the language market — just 1% and declining. According to this school of thought, APL never got off the ground and probably never will. It's considered something of a “poor cousin” to well-known, successful languages like BASIC, FORTRAN, and COBOL. Well, we've monitored the use of APL for some time now by reviewing the results of an unbiased, independent tracking survey on language usage, and we'd like to share some results that may surprise you. Results that show, for example, that the use of APL is not declining at all, and is even growing among one of the most important user segments in the market — the “casual programmer.” Within this segment — in a mainframe environment — APL enjoys ten times that pessimistic 1% share and is running neck-and-neck with BASIC. Admittedly, our data doesn't cover the entire data processing universe. It covers “primary use” in a particular high-quality mainframe environment (more on that in a minute). Still, we feel the data gives us an important look at where APL has been and where we, at least, think it's going. The data we're using has been collected as part of an annual membership survey by the GUIDE organization. GUIDE is an independent group whose members are organizations using IBM mainframes, usually “small” machines such as the 4341. GUIDE's annual membership polls are very large, generally about 1500 installations. The data presented in this paper is based on responses to the following question for the years 1978 through 1983: “A full-time programmer is one who earns his/her livelihood by programming. A casual programmer is one who does not earn his/her livelihood primarily by programming, but whose job does require some programming occasionally. Indicate the primary (most frequently used) language for (Check one language only per column). ” [I.e., one for full-time and one for casual]. So let's look at the numbers, beginning with full-time programmers. As Exhibit 1 shows, the use of APL as a primary language among full-time programmers has indeed hovered around 1% to 2% over the past 6 years. But those aren't the only numbers in the research, and we'd like to make a couple of points about what we think this data is really trying to tell us. First, the data simply does not lend any credence to the premise that the use of APL is declining. APL use is at least holding steady and perhaps increasing, although the size of the increase in the survey is not really large enough statistically to say that usage is increasing. Secondly, BASIC — one of the all-time “big three” — does no better than APL, and hasn't for the past six years (which is as far back as we have data). Another world-beater — FORTRAN — has been piddling around at about 4% ever since we've been tracking. That's not much of a margin over APL's 1% to 2% when you consider that FORTRAN is apt to be available in more of these installations than APL. In Exhibit 2, the data does show very clearly that COBOL is the dominant language among these full-time programmers, and we won't argue that point. But let's scrap once and for all the idea that APL can't hold it's own against languages like BASIC and FORTRAN among professional programmers. Now let's look at casual programmers — people who don't earn their livelihood by programming, but whose jobs do require some programming, Exhibit 3 shows that the use of APL as a primary language among this group has doubled over the past 6 years, climbing steadily to 10%. As was also true among full-time programmers, the data shows that APL is equally as popular among these casual programmers as BASIC. The difference with casual programmers, of course, is that use of APL has grown rather impressively over the years, while BASIC has stagnated. Meanwhile, FORTRAN's popularity seems to be declining at about the same rate that APL is climbing. Exhibit 3 shows FORTRAN declining 6 percentage points over the past 6 years to its 1983 level of 18% primary usage. Maybe we'll catch them, too. Exhibit 3 also shows that COBOL is again the dominant language, though by nowhere near the margin it holds among full-time programmers. Margin of leadership is commonly used by business analysts to calculate a ratio called “relative share.” The idea is that 15%, for example, is top-notch if the market leader only has 16%. Anyway, it seems to us that APL's relative share of over one-third (10.1 / 27.7) among casual programmers is pretty respectable when you consider some of the popular assumptions about the size of the APL niche. Exhibit 4 tracks APL's relative share among casual programmers since 1978. So what does it all mean? First of all, we feel the data shows APL in a pretty favorable light. Our conclusion is that if a quality APL is available (as is true in the segment of the universe represented by this data), users will adopt it and stick with it. The six years of data we've reviewed here bears that out. Because of this, we firmly believe that quality APL products should be ported to a variety of new environments. Secondly, we think the positive implication of APL's growth among “casual programmers” is fairly obvious considering the spread of computer literacy in the burgeoning white collar labor force. The current industry phrase — “knowledge workers” — says it all about this segment. They aren't computer jocks, but they get their jobs done by using a computer to convert a variety of data into usable knowledge. These are the users who are really driving the microcomputer explosion, and we're waiting to see what effect IBM's banner-year PC shipments will have on our 1984 and 1985 data. However, based on the casual-user data we've seen, we believe that with quality APL systems now available on the current generation of 16-and 32- bit microcomputers, we will see more APL usage among the growing base of casual microcomputer programmers, and 10% APL penetration as the primary language of these users is a benchmark against which we should measure ourselves. The challenge in this segment will be competing with the spreadsheets, DBMS, and other slick front-end software that is setting new standards in user interface and expectations. We think the casual programmer will demand more of these types of interfaces and features. To continue to succeed, APL systems will have to evolve along these lines. Tables of the raw survey data are included in the appendix, incidentally, so you're free to do your own analysis in addition to ours. The authors would like to express their thanks to Tom Gull, who compiled the GUIDE data used in this paper.