FULL DISCLOJUREA review of the Clojure Programming Language

OUTLINING CLOJURE Background – who, when, and why. Overview of Language Design

Names, Binding and Scope Data Structures Expression and Assignment Control Structures Subprograms Abstraction/Encapsulation

OUTLINING CLOJURE (CONT’…) Object Orientation Concurrency Exceptions and Events

Evaluation of Clojure Readability Writability Reliability Cost

Discussion

CLOJURE’S STORY Author Rich Hickey first released in

2007. Essentially a unified JVM-based LISP. Supposed to make programming,

specifically with regard to concurrency, easier.

Aims at combatting the perceived overuse of object-orientation and its implications.

Pragmatism trumps idealism: Not purely functional, OO borrowed from where useful.

NAMES, BINDINGS, AND SCOPE Slightly different naming scheme

As usual case-sensitive and no leading numeric Alphanumerics, “+”, “-”, “*”, “/”, “!”, “?”, “.”,

“_” Everything is a symbol (details later)

Scope is lexical with ‘let’, dynamic with ‘bind’.

Var is Clojure’s variable (but not really). Keywords always evaluate to themselves.

DATA STRUCTURES Data structures are immutable (no changes). “Nil” is Java Null, representing all of nothing. False is either “false” or “nil”, all else is true. Has Java primitives plus the Ratio Type. Operators don’t autopromote overflowing

values, but they have complements that do. Coercion is widening (as with Java).

DATA STRUCTURES (CONT’…) Collections: lists, vectors, array-maps,

sets. Count gets the collection size Conj adds to the collection Seq sequences the collection according to

its type (different behaviors, but all use it) Supports set operations and some

relations.

EXPRESSION AND ASSIGNMENT Forms are Clojure Expressions. Parenthetical prefix notation using

forms. Leftmost form is the function, while the

remainder are arguments of the function.

No declarations – Clojure is Dynamic No interpreter required: items are

examined, and if needed compiled and evaluated.

EXPRESSION AND ASSIGNMENT (CONT’…)

Evaluation resolves literals and keywords to themselves, and symbols to their value.

Symbol resolution is path-dependent Namespace-qualified gives its global value Package-qualified symbol gives it Java-

class Special forms get applied with their rules. Class-mapping, local binding, var, are

checked. Nonexistence will generate an error.

EXPRESSION AND ASSIGNMENT (CONT’…) The form comprises everything.

Literals are forms that are their own values. Symbols are the next lowest, being nonliterals

and nonsyntactic characters. Composites are defined by enclosing brackets

and represent vectors, maps, or lists. Special forms are composites whose

definitions are built directly into Clojure. Macros are used to extend code internally.

CONTROL STRUCTURES “do” executes a sequence of forms in order. “if” takes the comparison, the true-return, and

the optional else-return as parameters. “if-not” is a macro inverting the “if” special

form. “when” is a macro of “if” with an implicit “do”. “when-not” is an inversion of “when”. “cond” approximates if-elseif structure. “condp” approximates the switch structure.

CONTROL STRUCTURES (CONT’…) There are no loops in Clojure. Recursion is the only way to iterate. A function call containing a control

statement allows recursion to emulate typical looping.

Unbounded or infinite loops are achievable using recur to eliminate memory growth, but it must be explicitly called.

SUBPROGRAMS Clojure programs are built up like trees from

functions as nodes and values as leaves. Invites too-clever coding and the associated

inscrutable maladies arising therefrom. All functions are first-class objects, so they

can appear as values to other functions. Functions needn’t be predefined for use. OO behavior can be achieved with layers.

ABSTRACTION/ENCAPSULATION Record, Protocol, and Type enable

custom data-structure definition. Records assign a type to a map or

hash-map using “defrecord” to specify the type.

Protocols resemble Java interfaces, and are implemented using “extends”.

Types are specified with “deftype” which is essentially “defrecord” sans implementation.

ABSTRACTION/ENCAPSULATION (CONT’…) Macros take unevaluated arguments

and instead of values return forms. This allows direct manipulation of code

within the program, by the program itself.

Since Clojure is homoiconic (i.e. data and code have identical structuring) , this enables the powerful metaprogramming for which LISP is known.

OBJECT ORIENTATION Clojure is Anti-OOP, in that it attempts

to deobjectify programming. Polymorphism is maintained, but

decoupled from the state-and-behavior object model.

The desirable OO tools are inherently available, and object-equivalent structures can be synthesized as needed.

Ref and Agent allow for identity modeling.

CONCURRENCY Clojure’s mutable state is rarely-

needed evil. It uses Software Transactional Memory

(STM) to eliminate concurrency issues. Locks, semaphores, and mutexes aren’t

studied, they are automatic to the language.

Concurrent operations are given snapshots, each does its work, and then changes are merged back together, retrying as needed.

CONCURRENCY (CONT’…) Reference types: Ref, Agent, Atom, and

Var A “ref” requires interaction within

transaction. An “agent” value is handed to a

function, and the return is then assigned to the agent.

An “atom” is based on java.util.atomic, the classes aren’t race or locking sensitive.

A “var” is a global variable used for tracking runtime settings and the like.

CONCURRENCY (CONT’…) Multimethods are Clojure’s prime

polymorphism construct. The idea is that a function designator calls

the appropriate function based on the received value.

The equivalent of normal object behavior would have the value equal “this”.

Implementation inheritance is unsupported.

EXCEPTIONS AND EVENTS Java requires explicit exception

handling, and its design works down into the JVM.

Clojure doesn’t require checked exception handling by the programmer, but must deal with the JVM requirements.

To cope, all functions throw root (Runtime) Exception, and the program passes exceptions out from the caller.

EXCEPTIONS AND EVENTS (CONT’…) Naturally, Java’s event-handlers were

designed to deal with mutable state. Dealing with this uses dynamic

bindings. Since the handlers are typically single-

threaded, this rarely threatens concurrency.

Several GUI frameworks for Clojure exist, perhaps most notable Seesaw.

EVALUATING CLOJURE Readability benefits from the syntactic sugar,

but suffers from the unfamiliar LISP style. Writability also benefits from brevity, but again

the alien qualities hinder this as well. Reliability is benefitted by the JVM, pace of

development, and Clojure’s general ideals. Cost is lessened by having the JVM, but the

learning curve can be a training obstacle.

DISCUSSION Thank you for your time, patience, or

interest (whichever you happen to have).

Coming up next: An implementation of a boundary-free “Game of Life”.

Before we go to code, any questions first?