Cognitive Consonance

When you read about programming language comparisons, the main narrative for comparison is normally about the paradigm(s) supported. Lisp, Haskell, Scala, Clojure fall mainly in the functional realm. Prolog is logic. Smalltalk OO. C and Fortran, imperative. Most of them are not “pure” paradigm (e.g. you can make nice OO designed programs in C – just check GTK’s GLib library if you disagree, imperative coding in Prolog, and so on…), but that is besides the point.

The point is that, when comparing programming languages, the main issue of discussion is the bloody paradigm thing.

Paradigm is not really that important! In fact, as said above, you normally can tweak a language to write in your favorite paradigm. Sure the ability to do that varies from case to case, but in most cases that I can think of, it is really not difficult to cross paradigm boundaries. In fact, I would go as far as to argue that it is easier to do proper OO design with C using GLib then with the highly complex and convoluted C++.

Before going into the fundamental point that I want to make, I would also note that ecology matters: Are there good libraries? Good documentation? Does it run on a virtual machine? Portability? Nice community? User base? That is, when comparing programming languages all that is around the language is more important than the language itself. Just ask all the poor of us poor Prolog/Lisp/Haskell fans why are we doing Java/C++ during most of our day? It puts bread on the table, and, for the most of us, that is the most important criteria (I prefer not to starve!).

But, going to the main point here, I would like to propose that one of the fundamental points in comparing programming languages from a technical standpoint is homoiconicity.

Just to remember, an homoiconic language is a language where the program is represented as the core language data-type. Code is a data type.

If you classify languages according to homoiconicity, then they split in completely different ways:

1. The homoiconic bunch: Lisp, Prolog, Ioke, Clojure, …
2. The non-homoiconic bunch: Cobol, Fortran, C, Java, Goovy, Scala, Haskell, OCaml, [A very long list follows]…

From this point of view, the comparison of say, Clojure to Scala as sister-languages makes little sense, as they fall in different groups.

Homoiconic languages lend themselves to – by construction – metaprogramming and extensibility (think very easy embedded DSLs). And some of these features are difficult (with varying levels of difficulty) to implement in non-homoiconic languages. At best (as “best” I am thinking of some scripting languages like Python), they are awkward to do in a non homoiconic language.

As a side jab, last time a checked, Scala was very very poor on metaprogramming (has that changed?), making it the only “modern” language which seems to be scorning metaprogramming. Scala can still be DSL-extensible (I offer my own example both in Scala and Grovy: Ronald: A Domain-Specific Language to study the interactions between malaria infections and drug treatments.

One could argue of the value of doing programs that reason about themselves (and that idea has very bad karma coming from assembler – an idea so old and so disconnected from current reality that I am not even going to discuss it). I am surely on the side that proper metaprogramming is one of the core features of any elegant, productive and declarative solution.

Also, a very nice side effect of having code as data, is that the syntax of homoiconic languages is normally very, very simple (as in trivial to learn). This is just a side effect, but compare this with the learning curve of, say, C++ syntax. There is also a philosophical issue here: you get a simple, highly flexible environment, where complexity is tacked not by having a complex mammoth that tries to address all possible cases, but by a set of plastic, bendable building blocks.

Homoiconicity is not a black-and-white feature. For instance, Lisp macros are not first-class objects (I am a Clojure newbie, so feel free to correct me) so you cannot metaprogram with them. Prolog seems to come close. In fact, to a Prolog programmer, Lisp macros seem especially inelegant as the are “out of the system”.

I am doing some development in Clojure (a Lisp type language for the JVM). Lisp as in a clone tailored for the JVM, not Lisp as only “functional programming”. I note, by the way, that more than functional programming, Lisp is an homoiconic language.

I developed a simple system to specify Swing menus in clojure, here is an example:

Simple Menu

The following “micro-language” was developed to specify this:

 (getMenuBar actionManager '(
    (menu {
      :text "Project" :key "P"
      :content (
        (item {:text "New" :key "N"})
        (item {:text "Open" :key "O"  })
        (item {:text "Close" :key "O" :id "Close" :enabled false})
        (item {:text "Recent" :key "R"})
        (separator)
        (item {:text "Exit" :key "E"})
      )
    })
    (menu {
      :text "Options" :key "O"
      :content (
        (item {:text "Rendering" :key "R"})
      )
    })
))

The code is very easy to read, I hope: two menu items, with a few menu entries with text, ability to enable/disable and accelerator keys, plus a separator.

Notice the actionManager on top, is it the (very simple) event processing function which receives only a text as parameter (to identify the selection). The text is simply the menu text, or, if specified an id. Not the most general solution, but enough for simple menu structures.

The code? Below is the _complete_ implementation.

 
(ns org.tiago.swing
  ;(:require clojure.contrib.def)
  (:use
    [clojure.contrib.seq-utils :only (flatten)]
    [clojure.contrib.def :only (defnk)]
  )
  (:import
    (java.awt.event ActionListener KeyEvent)
    (javax.swing JFrame JMenu JMenuBar JMenuItem)
  )
)
 
(defnk createFrame [title :menuBar nil]
  (def frame (new JFrame title))
  (. frame setDefaultCloseOperation (. JFrame EXIT_ON_CLOSE))
  (if menuBar (. frame setJMenuBar menuBar))
  (. frame pack)
  (. frame setVisible true)
  frame
)
 
(defmulti addMItem (fn [manager x & rst] (first x)))
(defmethod addMItem 'item [manager content menu]
  (let [params (second content)]
    (def mItem (new JMenuItem (:text params)))
    (if (contains? params :id) (. mItem putClientProperty "id" (:id params)))
    (if (contains? params :key) (. mItem setMnemonic (. (:key params) charAt 0)))
    (. menu add mItem)
    (. mItem addActionListener manager)
 
  )
)
(defmethod addMItem 'separator [manager sep menu]
  (. menu addSeparator)
)
 
(defmulti getMBItem first)
(defmethod getMBItem 'menu [desc]
  (let [params (second desc) manager (last desc)]
    (def menu (new JMenu (:text params)))
    ;Assuming mnemonic is ASCII CODE.
    ;java7 has . KeyEvent getExtendedKeyCodeForChar
    (if (contains? params :key) (. menu setMnemonic (. (:key params) charAt 0)))
    (if (contains? params :id) (. menu putClientProperty "id" (:id params)))
    (dorun (map #(addMItem manager % menu) (:content params)))
    menu
  )
)
(defmethod getMBItem :default [arg] (new JMenu "UNK"))
 
(defn getMenuBar [actionManager menuItems]
  (let [manager (
      proxy [ActionListener]
      []
      (actionPerformed [e] (let [obj (.getSource e)
                                 id (.getClientProperty obj "id")]
        (actionManager (if (nil? id) (. obj getText) id))
      ))
   )]
   (def menuBar (new JMenuBar))
   (dorun (map #(. menuBar add %)
            (map #(getMBItem (concat % (cons manager ()))) menuItems)))
    menuBar
  )
)

OK, comments have to be added .
From a declarative point of view, not bad at all.

My first Lisp program. It completely baffles me that, 25 years of programming with all the languages imaginable (including some functional like Caml or highly declarative like Prolog), I never tried Lisp.

If you search the web you can find some discussions on whether IDEs for dynamic languages can be as helpful as IDEs for static languages. The issue is that static languages like Java have compile-time (thus easy to get at IDE-time) information in order to provide that fundamental code-completion functionality (among many others). If the IDE knows that a certain parameter is a String, than it is simple: it will present to you all the String methods when you type in the dot. For dynamic languages things get more complex are there is formally no (by definition) compile-time information. Some people would argue that there are ways around it (which you can already find in existing IDEs, I remember having some sort of code completion, years ago, on SPE – for Python). I will not add anything to that discussion here, this preamble was mainly for putting the reader in context. I am more interested in discussing good IDEs for DSLs.

With DSLs you get, most of the times, added syntax. Worse than that, you might fall into situations where you have changed (not only added) the initial language syntax; furthermore those syntax changes might even become valid only in runtime (imagine that a method is added to a class that is supplying DSL methods).

One example comes from Ioke and Prolog operator precedence and associativity rules which are changeable (see the previous post). It is not trivial to know if something like 1+2 is even syntactically valid (*). Even if it is syntactically valid things like association rules might change. In languages like Groovy you can add (e.g., through categories) methods to code blocs (from classes that can be dynamically changed). Then there is dynamic dispatching and macros. What is valid in a certain piece of code can be different from what is valid a few lines below. In fact, complete information of what is valid in a certain code block might require code execution. Or, to put in another way, it might be very difficult to have a completely helpful IDE! In this scenario there are 3 considerations that I think are worth being done:

1. One should not be discouraged for not having perfect solutions. Maybe it is not possible to determine all that can be expressed in a certain code block, but sometimes good approximations are enough.
2. On this issue, one good example comes from Prolog: In Prolog, syntax can be changed mainly through the use of the p directive (and through asserts and retracts). The p directive changes operators but is very easy to analyze pre-compilation/interpretation. So, the way DSLs are normally be constructed lend themselves very easily to code analysis which can be used by IDEs. This unfortunately not the case in most real-world languages.
3. It would be cool to have a language where DSL specifications could be automatically used to construct IDEs. The current real-world DSL-able languages (Ruby, Groovy, …) are DSL-enabled through indirect techniques which can be used to build DSLs (Dynamic reception, operator overload, whatever), in fact many of these techniques exist with other objectives than creating DSLs. If there was a declarative and explicit way to create DSLs, that information could be used to inform IDEs on parsing and other issues. An embedded, core way, to explicitly specify DSLs.

(*) I suppose some will see this as an argument for the fact that you can do pretty stupid (or at least unintuitive) things with DSLs. Well, you can do stupid things with everything. The question is not if you can or not, but the extent of bad use cases and how bad uses can creep in easily. Another (interesting) discussion, but not for now.

I was reading Ola Bini’s post about operators in Ioke (Ioke being the new language that Ola is developing).

It is a common saying around LISPers that everything that is being done in “modern” languages is a return to LISP. And the argument holds some ground. The truth is, among the 4 most conceptually influential programming languages that I can think of (Lisp/Functional, Fortran/Imperative, Smalltalk/OO, Prolog/Logic), the bad option (Fortran) won as it is the major philosophical contributor to current programming languages (much more than Smalltalk).

Take the reinvention of operators on Ioke as per the post above. This concept is available in Prolog for decades. It is all there: precedence (i.e. 2*3+4 means (2*3)+4 and not 2*(3+4)). Associativity (left or right – ie. 3-2-1 is 0 (3-2)-1 and not 2 3-(2-1) ). And even more as new operators can be defined and can be made of alphanumeric characters (want to create a new operator called say, “in”? go ahead). In fact people were doing DSLs a long time ago (in the small Prolog community at least) using techniques such as these.

The next thing that you will need (and we are getting there with macros and AST access) is no default interpretation. This is especially important with arithmetic, let me give an example:

Imagine the expression 1+x. Most languages will evaluate this expression and will return the sum of 1 + x. If x is defined and say is 4, then 1+x is 5. If x is not defined then an error (compile or run)-time will be raised. This is an absolute disgrace for DSLs with are essentially declarative (i.e., detached from semantics). “1+x” might be something that you want to evaluate now (and get the result) or might be something that you want to specify in order to evaluate later (say, I want to do a chart of all values of x between 1 and 5, or I want to differentiate), look at this pseudo-code

1
2
3
4
5
6
7

Var x
Exp expression = 1 + x**2
 
chart(expression, [[x,[1, 5]]]) //do a chart, x between 1 and 5
evaluate(expression, [[x,3]]) //Evaluate expression where x is 3 (i.e.  10)
diffe = differentiate(expression, x) //returns the expression 2*x
prettyprint(expression) //Pretty prints the expression.

Most people automatically associate the operation evaluate to 1+x**2. That might be so in an imperative world (can I call it shitty world?). But in an declarative/DSL world 1+x**2 is just that, an expression, it has no meaning attached per se. What you do with it depends on the context. Pretty print it, differentiate it, integrate it, or even evaluate it by instantiating x to 3 and getting the “precious” 10.

Update: I was rereading the post and noticed that it might be read as seeing Ola’s work as less interesting. Not at all: I actually think the way forward is precisely improving the current “imperative” setting in the way Ola is doing.

Preamble: In order to understand this post you should know a little bit (a little is enough, that is how much I know) about ExpandoMetaClass and Categories in Groovy.

DSLs that involve existing classes might be a source of long term sorrow. Let me give an example: Imagine that you want to make a small DSL to handle equations, like

x = new Symbol("x")
(2 * x).differentiate(x) //Result is 2

The problem is that the * operator of Numbers doesn’t know how to handle Symbols, therefore an exception would be raised. The obvious solutions as discussed before on mailing lists and blog posts are:

Categories

Categories would solve the problem, but at the expense of polluting the source with things like

use (Something.Category) {
  //code here
}

Not a disaster, but not pretty too…

Talking about disasters…

Expando over Numbers

The idea here would be to change the behavior of Numbers to be able to handle Symbols. Code would be very clean, no need for uses…

As somebody said on the groovy mailing list: This is disaster in the making. The problem is that I change Numbers, then, for another valid reason you change Numbers, somebody else also changes Numbers… This is chaos. Or at least it would make code from different sources potentially not inter operable or exhibiting very strange, buggy, behavior. This is clearly akin to the “global variable” problem. I believe that in the long term and with big software projects, this approach is a dead end.

Enter Python

Python actually has a workaround (I will not call it a clear, beautiful solution) that might be somewhat useful here. Imagine that you do

1 + x

The default 1 (default class for number) is not able to handle the symbol. For python that is OK, it will try to call a “right add” method of x (Search for __radd__ in this page). So, the default behavior is not to raise an exception if the left object cannot handle the operator, but to try to call the “right” version on the right object (if it fails then raise).

Not perfect, but might be just enough to avoid Expando in anger.

I do believe that people still don’t appreciate the consequences of Expanding core classes and the interop disaster that that can entail.