The Future of Prolog




DUCUNT VOLENTEM FATA NOLENTEM TRAHUNT

(Attributed to Lucius Annaeus Seneca)


Lessons from the past

Prolog is harder to learn than many other programming languages.

In part, this is due to inherent reasons: Prolog is simpler and more powerful than many other programming languages, and so it takes longer to get used to this if you are more familiar with low-level languages.

However, a far more important cause is found in limitations of existing teaching material: In many cases, Prolog is being taught exactly like it was taught in the 1980s and 1990s. Outdated teaching material leads to broken promises: The use of low-level and impure language features complicates or prevents logical reasoning about Prolog programs, forcing students to think procedurally about their programs instead of declaratively.

In the past, this has resulted in a lot of frustration, wrong impressions and coding horror. For beginners, it is easy to get sidetracked and start wrestling with low-level issues while losing sight of much more important principles. Do not fall into this trap!

Prolog is also hard to teach, since it is hard to keep all informal statements you make about Prolog programs sufficiently general to cover all possible usage modes. Conscious effort is necessary to convey this generality to beginners.

Current developments

In the Prolog community, most current developments are focused on increasing logical purity, and making pure code more feature-rich and efficient.

Video: Preparing Prolog


The ISO standardization process is an important means to improve compatibility between different Prolog systems.
Video: Prolog Conformity Testing


Web services and data analysis are increasingly relevant application areas for Prolog.

One of the most exciting developments for efficient data and text analysis with Prolog is the implementation of a very compact representation of strings as lists of characters so that DCGs can be applied efficiently to files and streams. Scryer Prolog is the first system that implements this representation, and it is to be hoped that other Prolog systems will also adopt it. The ultimate goal of this development is a Prolog system's ability to directly apply a DCG to file contents by using the system-call mmap for very efficient access. Trealla Prolog is the first—and currently still only—Prolog system that provides this.

Most recent and ongoing developments of Prolog happen in universities and in institutions that are closely associated with academia. Important contemporary research areas include probabilistic logic programming, answer set programming (ASP) and language dialects that satisfy certain safety criteria or other interesting declarative properties.

As an application programmer and Prolog practitioner, you can contribute to these developments by getting acquainted with modern Prolog features, applying them in your own programs, and discussing opportunities for improvements with your Prolog vendor. Continuous and wide-spread usage helps to ensure that these features are constantly improving.

The begin

The advent and wide availability of new, declarative language features justify and moreover necessitate that we teach Prolog differently than we did in the past.

Video: Applying the Koch method to Prolog


Nowadays, it is becoming increasingly inappropriate to dedicate a significant portion of your lectures to explaining the intricacies of low-level language constructs that have long been superseded by more general replacements. The time is better spent elsewhere: in explaining the foundations of logic programming in such a way that modern language constructs are naturally covered. This retains full generality and allows both declarative and procedural readings instead of precluding one or the other.

To really appreciate modern Prolog features, you also have to take into account their history: Previous generations of Prolog instructors were acutely aware of the declarative limitations associated with the constructs they were teaching, but had no other options. In many cases, the instructors themselves were also among the pioneers in the difficult search for more elegant and more general constructs.

For instance, consider the following section from a manuscript by Richard O'Keefe, written no later than 1984: 
Instantiation Faults

The fifth kind of error is when a question has too many variables in it.
Now this logically speaking is no error at all.  The question "plus(X,X,Y)"
has a perfectly good set of solutions {(X,Y)|X is an integer and Y=2*X}.
It is evident from this draft that very declarative features have been anticipated for several decades. It has simply taken a certain amount of time until the techniques that were envisioned and even outlined by many pioneers were fully realized. Today, we get in full accordance with what the draft envisaged: 
?- X + X #= Y.
   clpz:(2*X#=Y).
By now, declarative features like constraints are widely spread and available at least to some extent in all popular Prolog systems. For many people working in declarative language research, this is a dream come true, and still materializing. These features are meant to be used. I mean not only as interesting curiosities, but as integral parts of the language. Prolog systems that do not yet include some of these features will adopt them as they become more popular.

In the future, the most influential Prolog books will be those which most explicitly expose the pure core of Prolog and the unique advantages and application opportunities of logic programming and constraints, such as: By fully embracing such concepts and using them to their utmost advantage, future generations may at last behold and benefit from the true power of Prolog.


More about Prolog


Main page