The Shape of Code

P is for Pascal, Prolog, Perl, Python and PL/1.

Pascal was the love of my life when young (what else would cause anybody to think that a C to Pascal translator was a good idea). The language built up a substantial user base, was widely taught in universities and was widely considered to be a good language for writing robust software, yet by the end of the 1980s its market share had crashed. What happened?

I think Pascal lost out to C because it generates a lot more upfront friction. Stronger type checking, runtime array bounds checking; its just unreasonable to expect developers to deal with this stuff and deliver on time. Discovering faults in code is what customers are for. Besides, any language that forces developers to use vendor extensions because the language does not officially support separate compilation, bitwise operations and other ‘useful’ stuff just does not deserve to be taken seriously.

Languages like Pascal will not become mainstream until managers stop allowing developers to put private short term interests before the longer term reliability and cost interests of groups.

Prolog is a member of the set of languages that are very different from all the other languages. looks so easy to use, when reading examples in books. In practice the Experience of using almost any existing languages (i.e., those based on control flow) means a whole way of doing things needs to be unlearned; declarative languages don’t have control flow, they work by developers expressing goals + relationships and leaving it to the implementation runtime to figure out how to reach the appropriate goals.

Perl shares with Lisp the ability to attract beards and Dungeons&Dragon enthusiasts, but differs in not repelling non-clever people (which is a serious design flaw given the amount of Internet infrastructure based on it). At times it feels like you can reach out and touch the Perl community, why is this so strong in Perl? Perhaps mastering the language syntax results in ascending to an astral plane where it is only possible to communicate with other initiates. Here is a self-documenting group of people just waiting to be studied for social anthropology PhD.

Perl 6 continues to provide a useful reminder that while developers hate language warts from early versions, they hate it even more when the warts are removed and the ability of existing code to continue to work becomes questionable.

Python is the language I use to work with the young generation of developers.

PL/1 was the first corporate language and because computer languages were still very new back then they could not adopt something a couple of kids had hacked together based on existing languages. Instead, there were committee meetings, academic work on a formal definition and magazine profiles by Very Important People.

In an attempt to be user friendly PL/1 was designed to try very hard to implicitly convert values to whatever type was required. Given its built-in support for Cobol business data-types and Fortran engineering data-types (in a bid to attract users from both those languages), plus other data-types that neither language had decided to include (e.g., bit strings), developers regularly encountered a level of surprising behavior that modern languages rarely achieve.

Things to read

Algorithms + Data Structures = Programs by Niklaus Wirth was a revelation in its day and is now basic introductory software engineering. Unless resources are very limited, developers use constructs that encapsulate the details illustrated in this book, such as iterators.