Add PBE to Conventional Languages

I read guest editor Henry Lieberman's special section introduction, "Programming By Example," and the following article "Novice Programming Comes of Age" (Mar. 2000) by David Canfield Smith et al. with great interest. I agree there is a problem with current conventional programming languages, which contain arcane syntactic notations and form, leading to misconceptions. However, I question the feasibility of the proposed programming by example (PBE) "Creator," as widely acceptable for the programming curriculum or, for that matter, as a substitute for conventional languages.

In regard to learning programming using conventional languages, I went through an experience similar to the authors'. When I first started programming, I questioned the way programs were represented in a textual and linear form. Programs were divided into several groups of statements situated one after the other, thus giving the impression that every statement flows sequentially. However, in a program, when one group of statements is executed, another group may be skipped. Also, it is possible that one group of statements is executed more than once. The textual representation of programming did not help me realize the actual flow of the program and led to much confusion. A pictorial representation of the program could have eliminated this confusion.

The understanding of control flow was the most challenging part of programming for me. After 23 years of educating novice programmers, I have now found, and continue to find, students experiencing the same difficulty. Not much has been done in conventional programming languages, whether procedural or object-oriented, to tackle this problem.

While PBE is a great idea and Creator is a successful model, the question of how it will achieve its goal of being widely accepted remains. Is it the goal of Creator to take over the beginner's programming curriculum and replace the existing conventional languages?

In the past several years, I have not seen a drastic change with regard to programming languages. Any current changes in programming languages are built on the existing languages. For example, C++, Java, and others are all built on C, which itself is built on Algol and Fortran with the backing of a big enterprise or a major institution. It may be argued that even though the design intention of C was for system programming, it has received wide acceptance and has moved into the business and scientific communities. Why has this occurred? Is it a miracle, good luck, performance and production (Unix), or others reasons? The question we need to asked is, How can Creator achieve similar success and become widely accepted?

If Creator is not winning by revolution, then perhaps a transitional period could be a remedy. As indicated by the author, there is a conservative community in the programming field that simply won't accept such a drastic change. It is also important to note that this conservative community is not fading away, so a gradual or evolutionary change is indicated.

Within programming languages, we have seen poorly chosen notations that were adopted because they were the only option, due to the lack of technology, other restraints, or even personal taste. However, these notations continue to be carried on from language to language, as a bad gene continues from one generation to the next. In the past 50 years, there has been no reexamination of these languages. Based on this observation, it would be prudent to incorporate the good features of the current conventional languages into Creator as an option. This would ease the transition of traditional programming to the new PBE standard.

In addition, there should be a massive campaign advocating the benefits and broad feasibility of Creator as a general programming tool and not just as a simulation tool for children's games. For example, given the ability to solve a common programming problem, such as finding an average, a minimum, or sorting, in both PBE and a conventional language would give readers a better chance to compare both languages.

I disagree that nothing has been done to address the problem of novice programmers. In fact, there has been an ongoing effort to resolve the problem of languages written prior to PBE. Two such examples are Karel the Robot, 1981 (Karel++, object-oriented version, 1997) and my own VPCL (Visual Plan Construct Language), 1989. The empirical studies of novice programmers indicate that students who study these languages do substantially better in conventional languages. Therefore, I fully agree with the assertion that working with Creator results in better understanding of conventional languages. The only drawback is that students dislike the fact that they have to learn additional languages and then learn conventional languages.

Alireza Ebrahimi
Old Westbury, NY