Unique Features of Various Programming Languages

A language that doesn't affect the way you think about programming, is not worth knowing.

– Alan Perlis

I like learning new languages to get myself to think about problems in different ways.

Here’s a list of a few well-known languages I’ve worked with and some of the more interesting features I’ve encountered:

C

• Undefined behavior
  • The idea that a language specification would explicitly specify certain constructs as having undefined behavior is interesting. Most specifications leave things undefined by omission, not commission.
  • Although this allows various compiler optimizations, many developers rely on their particular compiler’s implementation of undefined behavior without even realizing it.¹

Java

• Objects are the primary unit of composition
  • There are no standalone functions.
  • Design Patterns can be used to describe high level object coordination patterns.
  • However sometimes the community’s focus on objects can be a bit extreme.²
• Checked exceptions
  • Great when used sparingly, as it forces the caller to handle expected error conditions.
    • I do think that it was a usability error to make Exception checked and RuntimeException unchecked. Rather it should be Exception that is unchecked and a new CheckedException should be the base for all checked exceptions. This makes it clear that unchecked exceptions should be the default.
  • Managing checked exceptions correctly is quite difficult.
• Unicode strings
  • Tons of other languages use “bytestrings” as the main string type which causes all kinds of problems when working with international characters.
• Documentation focus
  • Every library is expected to have JavaDoc documentation, which is generated from documentation comments in the source
  • This is really powerful, since it makes it easy for developers to write documentation, and to do so at the same time the implementation is coded, when the desired behavior is most fresh in the mind.
• Compatibility focus
  • I have never seen a community so focused on maintaining backward compatibility. I can’t think of a single deprecated method in the Java library that was actually deleted.
  • Even the Java Language Specification has an entire chapter devoted to binary compatibility.
  • This focus may be partially attributable to the early mantra of Write once, run anywhere.
• Garbage Collection
  • Being freed from the confines of manual memory management makes it a ton easier to focus on more important things.

C#

Very similar to Java.³

• Assemblies
  • This is a level of encapsulation above the typical namespaces or modules in most languages. Assemblies are similar to the idea of static/dynamic libraries in C or JAR files in Java.
  • Notably, you can mark members as internal, which makes them public within the same assembly, but private to everybody outside the assembly. This is quite useful.
• Cross-language compatibility is first-class (not just for C)
  • C# runs in the Common Language Runtime, which was designed from the beginning to support interoperability between languages.
• Properties are first-class
  • No longer need to write explicit getter and (optional) setter methods.
• Listeners are first-class
  • Classes can declare an event Foo with addFooListener and removeFooListener functionality built in.
  • Unfortunately the implementation has some annoyances⁴.
• Foreign Methods⁵ are first-class
  • C# calls these extension methods.
• Partial Classes
  • Allows a class’s members to be defined in multiple files.
  • Useful to add functionality to a generated class (for example, from a parser generator) without those modifications getting lost when the class is next regenerated.

Python

• Indentation is significant
  • Everyone notices this pretty fast.
  • Overall I think requiring correct indentation is a Good Thing™, as it contributes directly to readable code.
  • However this makes lambda expressions much less powerful than in other languages, since you can’t easily nest statements inside a lambda expression without explicit braces (or similar delimiters).
  • Another side effect of having significant indentation is that the choice of tabs vs. spaces really matters. If you mix them, your program probably won’t run.
• Interactive interpreter (REPL)
  • Amazingly useful for prototyping quickly, running experiments, and learning the language.
• Collections are first class
  • You can type [1,2,3] to get a list, {'key': 'value'} to get a dictionary, and {1,2,3} to get a set. So much faster than new ArrayList<Integer>(...), new HashMap<Integer>(...), or new HashSet<Integer>(...).
• Functions are first class
  • You can declare them as literals, pass them around, and create higher-order functions that take functions as parameters.
  • Functions can also live on their own without an enclosing class, in contrast to Java. This is often the simplest approach for a given implementation.
• “Magic” marked explicitly
  • Anything “magic” that the language treats specially has names surrounded by double underscores.
  • For example:
    • An object’s constructor is called __init__.
    • The method implementing an operator overload for plus is called __add__.
    • A class’s metaclass is held by the __metaclass__ field.
• Generators and Coroutines
  • Generators enable straightforward pulling of values from a complex data source (like a parsed data structure).
  • Coroutines enable straightforward pushing of values to a complex data sink.
  • Python calls both constructs a generator.

JavaScript

• No blocking I/O
  • All I/O is non-blocking and asynchronous. This results in heavy use of continuation passing style.
• Successful despite huge flaws
  • Ease of deployment and ubiquity (i.e. business considerations) trump ease of use. (PHP also wins for the same reason.)
  • A few flaws:
    • Everything is in the one global namespace.
    • No user-defined namespaces, modules, or importing of other files.⁶
    • No built-in facility for classes.⁷
    • Bizarre loose semantics for ==.⁸
    • Multiple illegal value sentinels: null and undefined.⁹
• JSON (JavaScript Object Notation)
  • A fantastically compact, readable, and portable notation for representing all kinds of data structures. Great for data interchange.

Haskell

• Lazy evaluation
  • Expressions are only evaluated when some primitive operation (like print or add) requires the value of the expression.
  • Allows you to glue programs together in new ways. In particular, execution of multiple functions can be interleaved trivially and termination conditions can be separated from looped computation.
    • This is explained in more detail in Why Functional Programming (1990), §4 “Gluing Programs Together”.
  • Allows you to extract complex expressions without fear of introducing a performance hit (since the expression will only actually be evaluated if it is needed).
  • Allows you to define your own control flow operators.¹⁰
• Side effects banned by default
  • Mutation of data structures and I/O, both of which have order-sensitive side effects, are not allowed except within the confines of monad.
  • A monad is a construct that explicitly controls evaluation order, in contrast to the usual unpredictable lazy evaluation behavior.
• null banned by default.
  • Unlike many languages, there is no special null value in Haskell that can be substituted anywhere.
  • Instead if a function wants to return a value of type T or null, you would declare the function as returning type Option<T>, which could either have the value Some(tValue) or None. When declared in this way, callers are required to handle both possibilies.
• Type inference (+ static typing)
  • The benefits of static typing without the need to specify the types for everything?
    Count me in!

Lisp (Common Lisp, Scheme, Clojure)¹¹

• Homoiconic
  • When you write a Lisp program, the notation you use (the grammar) is equivalent to what a compiler would see (an abstract syntax tree or AST).
  • Furthermore this Lisp code is represented as a nested structure of lists, symbols, and literals, all of which can be directly generated and manipulated in Lisp itself!
  • This allows Lisp code to generate list structures which can then be run as Lisp code directly.
    • Generation can be done at compile time with macros.
    • Generation can be done at runtime as well, and then invoked with eval.
  • However the highly uniform structure of Lisp code, devoid of operator and syntactic diversity, makes for lousy typography and thus low readability.
• Macros
  • A function that transforms the AST of its operands at compile time to new code.
  • Macros can be used to generate arbitrary new statements and control structures.
    • Domain specific languages, in particular, are very easy to implement in Lisp thanks to macros.
  • Macros can also be used to perform code optimizations at compile time¹² (similar to “template metaprogramming” in C++).
  • Fluent use of macros requires the host language to be homoiconic, which is rare. Thus Lisp remains the only well-known language family that has macros.
• Lisp Conditions and Restarts
  • Allows bidirectional communication between different parts of the call stack. More powerful than exceptions, since conditions can not only unwind the stack but also wind it back again via a restart.
• Call-with-current-continuation
  • Allows you to save the current execution state of the program in a variable and jump back to it later. Multiple times, even. It’s like a friggin' time machine. You can implement fairly complex control flow operators with this function.

Other

• Fexprs
  • A function whose operands are passed to it at runtime without being evaluated.
  • Similar to macros and lazy evaluation in terms of power.
  • Has fallen out of favor since the 1980s due to being difficult for compilers to optimize. Also it is hard to provide good error messages.
  • Furthermore most fexpr functions tend to rely on eval to continue evaluating its operands. The use of eval has its own problems…

Related Articles

• Visual Guide to Programming Language Properties
  • Shows programming language features, how they are related, and which features are present in popular languages.
  • Visualizes some of the features discussed in this article.
• Notes on Racket
  • Describes Racket, a dialect of Lisp with batteries included.
    Useful for implementing other languages and creating cross-platform GUI programs.
• Notes on Prolog
  • Describes Prolog, a highly declarative language.
    Useful for exploring and verifying proofs.
• Learnings from SICP (and Lisp)
  • Discusses computer science concepts and how they manifest in various programming languages.