Using exceptions to simulate tail recursion in JavaScript

JavaScript is very powerful language but it don’t have tail call elimination. But it turn out that you can simulate it using exception system. Here is very simple recursive function that calculate factorial:

function factorial(n) {
    function recur(n, result) {
        if (n == 0) {
            throw result;
        } else {
            recur(n-1, result*n);
    try {
        recur(n, 1);
    } catch(e) {
        return e;

It turn out that in JavaScript (I read that in Douglas Crockford book JavaScript: The Good Parts) you can use any expression in throw and it will be send to variable in catch statement.

So what the above code does it simple exit from the recursive loop and pass result to catch statement. And this is exactly what tail recursion is, in language scheme this happen by default when inner function (you also need to create inner function in scheme) have recursive call as last expression. Here is scheme version of tail recursive factorial:

(define (factorial n)
  (let recur ((n n) (result 1))
     (if (= n 0)
        (recur (- n 1) (* result n)))))

the code use named let but it can be rewriten with inner function and invocation. (this kind of trick is needed in script-fuGimp extension based on scheme).

, ,


Ready to use JSON-RPC Servlet in Java

TL;DR: You can find the whole code in this gist.

I needed to write some Web based project in Java as an assignment, so I decide to use single page application and JSON-RPC service.

Here is the code I use to create this task, step by step

First I found JSON-RPC library, I choose JSON-RPC 2.0 Base library.

I put both JSON Smart and JSON-RPC 2.0 Base jar files into lib directory (inside WEB-INF) of my java app.

In order to use JSON-RPC I needed to get raw POST data from request, I do this using this method (of the Servlet class):

    public String getInputData(HttpServletRequest request) throws IOException {
        ServletInputStream input = request.getInputStream();
        int buff;
        String output = "";
        while (true) {
            buff =;
            if (buff == -1) {
            output += (char)buff;
        return output;

Then I could parse that POST data (inside doPost Servlet method):

JSONRPC2Request reqIn = JSONRPC2Request.parse(this.getInputData(request));

Now I could extract data from the JSON-RPC request:

     Object id = reqIn.getID();
     Object[] params = reqIn.getPositionalParams().toArray();
     String method_name = reqIn.getMethod();

I decide to use other class as a Service and call method of that class based on JSON-RPC request, so I needed to use Reflection to do that

ClassLoader classLoader = ServletClass.class.getClassLoader();
Object service = classLoader.newInstance();
Class aClass = classLoader.loadClass("pl.jcubic.Service");
Method[] methods = aClass.getMethods();

Method method = null;
for (int i=0; i<methods.length; ++i) {
    if (methods[i].getName().equals(method_name)) {
        method = methods[i];
if (method != null) {
    Object result = method.invoke(service, params);
    JSONRPC2Response respOut = new JSONRPC2Response(result, id);
} else {
    out.println(new JSONRPC2Response(JSONRPC2Error.METHOD_NOT_FOUND, id));

I use getMethods and search for method instead of using getMethod because I didn’t know what the classes I need to use for the params to get the right method.

I use class loader because I wanted to be able to recompile the Service (pl.jcubic.Service) and use it without the need to restart the app. But this loader don’t work as expected, the class file is loaded only once (when called the first time) the same as if I use import statement.

In order to have dynamic class load I need to write custom class loader, it look like this:

    class ServiceReloader extends ClassLoader {
        public ServiceReloader(ClassLoader parent) {

        public Class loadClass(String name) throws ClassNotFoundException {
            if (!"pl.jcubic.Service".equals(name)) {
                return super.loadClass(name);
            try {
                String path = getClassPath(name);

                DataInputStream input = new DataInputStream(new BufferedInputStream(new FileInputStream(new File(path))));

                ByteArrayOutputStream buffer = new ByteArrayOutputStream();
                while (true) {
                    int data =;
                    if (data == -1) {
                byte[] classData = buffer.toByteArray();
                return defineClass(name, classData, 0, classData.length);
            } catch (IOException e) {
                throw new ClassNotFoundException(e.getMessage());

I write it as inner class of my Servlet. Then I use this class to load my Service class.

    ClassLoader parentClassLoader = ServiceReloader.class.getClassLoader();
    ServiceReloader classLoader = new ServiceReloader(parentClassLoader);
    Class aClass = classLoader.loadClass("pl.jcubic.Service");

Now I could recompile the Service class and use it from JavaScript without need to restart the app (don’t even refresh the browser if I write new method and wanted to test it)

One more thing that need to be written is catch exceptions and display proper JSON-RPC error responses.

Now all you need is a class with methods that will become your JSON-RPC methods. Here is list of Java types mapend to JSON-RPC types. All paramters and return values of your methods need to be of those types.

, ,

Leave a comment

Add methods to Number object in JavaScript

In JavaScript everything is an object, even functions and numbers, and all objects even functions are first class citizens (which allow functional programming).

So this mean that you can call methods over numbers, and you can add new methods using Number.prototype. Here is example of range functions similar to that from Python, but bit different. (Python one return number from n to m – or from 0 to n – but this one return n numbers (n is this) starting from number passed as argument)

Number.prototype.range = function(n, result) {
    result = result || [];
    n = typeof n === 'undefined' ? 0 : n;
    if (this <= 0) {
        return result.reverse();
    } else {
        return, n, result.concat([this-1+n]));

And you can call this function/method using:


It will return array [1, 2, 3, 4, 5, 6, 7, 8, 9, 10].

Note that there are two dots, it’s because when you use one it’s interpret as float number and thorw exception "SyntaxError: Unexpected token ILLEGAL" because it wait for a digit as next character. The other options is to use parenthesis like (10).range(1);

Using this you can write factorial using

10..range(1).reduce(function(a, b) { return a*b; });

reduce function was added to Array.prototype in ECMAScript 5, You can check what browsers support them in this page.

You can wrap that code with a function and you can add it as Number prototype

Number.prototype.factorial = function() {
    return this.range(1).reduce(function(a, b) { return a*b; });

You can call it as:


Here is example of times function (similar to that from Ruby)

Number.prototype.times = function(fn, self) {
    return this.range().forEach(fn, self);

You can use this function using:

10..times(function(i) { console.log(i); });

Leave a comment

Working with SVG in jQuery

I recently notice that if you create element like a circle in SVG (inline SVG embedded into HTML), using inspector/firebug or jQuery, your circle is not visible on SVG until you refresh the SVG, I found a hack to force refresh of the SVG, I just get text of the SVG and insert it again into the DOM, and all elements that were not visible like circle mention before will be rendered.

Here is referesh function, as jQuery plugin

$.fn.xml = function() {
    return (new XMLSerializer()).serializeToString(this[0]);

$.fn.DOMRefresh = function() {
    return $($(this.xml()).replaceAll(this));

Those plugins should work with every XML embeded into HTML not only SVG. Don’t look very nice but it work, but I found a better way fix jQuery to work with SVG. I found that when I use function document.createElementNS instead of document.createElement elements added to inline SVG using for instance appendChild function everything work fine. So only thing that need to be done for jQuery to work with inline SVG is to replace this function if element is SVG. First I wrote method in jQuery object that test if element is SVG element, there are only three elements that have the same name in HTML and SVG is a, script, style and title tags so I didn’t put them in.

isSVGElement: function( o ) {
        if (o instanceof SVGElement) {
            return true;
        } else {
            if (typeof o === 'string') {
                return $.inArray(o, ['altGlyph', 'altGlyphDef',
                                     'altGlyphItem', 'animate',
                                     'animateColor', 'animateMotion',
                                     'animateTransform', 'circle',
                                     'clipPath', 'color-profile',
                                     'cursor', 'defs', 'desc', 'ellipse',
                                     'feBlend', 'feColorMatrix',
                                     'feComposite', 'feConvolveMatrix',
                                     'feDistantLight', 'feFlood',
                                     'feFuncA', 'feFuncB', 'feFuncG',
                                     'feFuncR', 'feGaussianBlur',
                                     'feImage', 'feMerge', 'feMergeNode',
                                     'feMorphology', 'feOffset',
                                     'feSpecularLighting', 'feSpotLight',
                                     'feTile', 'feTurbulence', 'filter',
                                     'font', 'font-face',
                                     'font-face-name', 'font-face-src',
                                     'font-face-uri', 'foreignObject',
                                     'g', 'glyph', 'glyphRef', 'hkern',
                                     'image', 'line', 'linearGradient',
                                     'marker', 'mask', 'metadata',
                                     'missing-glyph', 'mpath', 'path',
                                     'pattern', 'polygon', 'polyline',
                                     'radialGradient', 'rect',
                                     'set', 'stop', 'svg',
                                     'switch', 'symbol', 'text',
                                     'textPath', 'tref',
                                     'tspan', 'use', 'view',
                                     'vkern']) !== -1;

I get the list of elements from Mozilla MDN.

I added this method to main jQuery.extend({ that add methods to jQuery object. Next thing is to replace createElement with createElementNS, there are only 2 places with this in parseHTML (in the same jQuery.extend) and createSafeFragment function, only one is responsible for inserting elements – parseHTML. Below is the code for that function.

    parseHTML: function( data, context, keepScripts ) {
        if ( !data || typeof data !== "string" ) {
            return null;
        if ( typeof context === "boolean" ) {
            keepScripts = context;
            context = false;
        context = context || document;

        var parsed = rsingleTag.exec( data ),
            scripts = !keepScripts && [];

        // Single tag
        if ( parsed ) {
            if ( jQuery.isSVGElement( parsed[1] ) ) {
                return [ context.createElementNS(
                    parsed[1]) ];
            } else {
                return [ context.createElement( parsed[1] ) ];

        parsed = jQuery.buildFragment( [ data ], context, scripts );
        if ( scripts ) {
            jQuery( scripts ).remove();
        return jQuery.merge( [], parsed.childNodes );

Unfortunetnly not all manupulation methods that create new elements will work but it’s better then refresh hack. The stuff that don’t work is when you create more then one element from a string or if you put some attributes. In this case jQuery use document fragments and innerHTML to create the DOM. If we will want to fix that too we will need to write a parser that will call createElementNS.

You can also include those function after you load jQuery so original code will not be changed.

The other way is to use document.createElementNS and element.setAttributeNS that need to be used if you want to add xlink:href attribute (it use namespace).


Leave a comment

Image color picker in javascript that work without canvas in every browser

If you want to create color picker from image in JavaScript you probably will use canvas, but what if you need it work in IE8 as I needed. You can use some server side help to fetch pixels data from the server. I use php for that and GD library. The code is below.

cross-browser color picker

Server side code that return json pixel data


function rgb($color) {
    $result[] = ($color >> 16) & 0xFF;
    $result[] = ($color >> 8) & 0xFF;
    $result[] = $color & 0xFF;
    return $result;

function imagecreatefrom($filename) {
    $path = pathinfo($filename);
    switch($path['extension']) {
        case 'png':
            return imagecreatefrompng($filename);
        case 'jpg':
            return imagecreatefromjpeg($filename);
        case 'gif':
            return imagecreatefromgif($filename);
            return null;

function getImageData($filename) {
    list($width, $height) = getimagesize($filename);
    $image = imagecreatefrom($filename);
    $image_data = array();
    for ($y = 0; $y < $height; ++$y) {
        $row = array();
        for ($x = 0; $x < $width; ++$x) {
            $row[] = rgb(imagecolorat($image, $x, $y));
        $image_data[] = $row;
    return $image_data;

if (isset($_GET['filename'])) {
    $filename = $_GET['filename'];
    if (file_exists($filename)) {
        $result = array(
            'error' => null,
            'result' => getImageData($filename)
    } else {
        $result = array(
            'error' => "The file '$filename' don't exist",
            'result' => null
} else {
    $result = array(
        'error' => "You need to put filename",
        'result' => null
header('Content-Type: application/json');
echo json_encode($result);


Then you can fetch the data using ajax and add get invidial pixels on mousemove

    var img = $('img');
    $.getJSON('image_data.php', {filename: img.attr('src')}, function(data) {
        if (data.error) {
        } else {
            $('.eyedropper').click(function() {
                picker = true;
                return false;
            img.mousemove(function(e) {
                if (picker) {
                    var x = Math.round(e.pageX - offset.left);
                    var y = Math.round(e.pageY -;
                    if (x >= 0) {
            }).click(function(e) {
                picker = false;

Now all you need to have is element with class eyedropper like a link in your html:

<a href="#" class="eyedropper">pick the color</a>

, ,


Mastering Emacs – Search inside minibufer

This is by far best feature in Emacs that you can use interactive search inside Minibuffer. If you want to find a function that you don’t know exact name, I do this a lot – who will remember function names in Emacs, you can use:

C-h f C-s

And search for functions. How cool is that? Almost all functions that have completion inside minibuffer work with search like

  • C-x b C-s – search for buffer to switch
  • C-h <KEY> C-s – search for help

When I switch to Emacs 24, I found that M-x C-s don’t work. So you can’t search for interactive function to execute. But you can get it back. As describe by this Emacs Bug Ticket by putting this to you .emacs file.

(defun read-extended-command ()
  "Read command name to invoke in `execute-extended-command'."
      (lambda ()
        (set (make-local-variable 'minibuffer-default-add-function)
             (lambda ()
               ;; Get a command name at point in the original buffer
               ;; to propose it after M-n.
               (let ((def (with-current-buffer
                              (window-buffer (minibuffer-selected-window))
                            (and (commandp (function-called-at-point))
                                 (format "%S" (function-called-at-point)))))
                     (all (sort (minibuffer-default-add-completions)
                                (lambda (a b) (string< a b)))))
                 (if def
                     (cons def (delete def all))
    ;; Read a string, completing from and restricting to the set of
    ;; all defined commands.  Don't provide any initial input.
    ;; Save the command read on the extended-command history list.
     (concat (cond
	      ((eq current-prefix-arg '-) "- ")
	      ((and (consp current-prefix-arg)
		    (eq (car current-prefix-arg) 4)) "C-u ")
	      ((and (consp current-prefix-arg)
		    (integerp (car current-prefix-arg)))
	       (format "%d " (car current-prefix-arg)))
	      ((integerp current-prefix-arg)
	       (format "%d " current-prefix-arg)))
	     ;; This isn't strictly correct if `execute-extended-command'
	     ;; is bound to anything else (e.g. [menu]).
	     ;; It could use (key-description (this-single-command-keys)),
	     ;; but actually a prompt other than "M-x" would be confusing,
	     ;; because "M-x" is a well-known prompt to read a command
	     ;; and it serves as a shorthand for "Extended command: ".
	     "M-x ")
     obarray 'commandp t nil 'extended-command-history)))

And tell me, why you will ever need Ido-mode

One note about that, when you use one interactive function inside another one you will need to call C-g twice to exit minibuffer

1 Comment

Faster Emacs Window Switching Inside a Frame

I needed something to speed up my work when I have few windows, I was looking at switch-window.el, but it kind of silly that the content of the window disappear. So I’ve written this simple 2 functions which act similar to my other sollution Faster buffer bookmarking in Emacs. But you mark window to which you want to jump.

It set Keyboard C-c C-0 for mark and C-C C-<1-9> to jump.

(setq window-alist '())

(defun mark-window (number)
  "Give current window a number to select it later using `switch-to-number-window'"
  (interactive "nNumber Your Window: ")
  (let ((pair (assoc number window-alist))
        (window (get-buffer-window)))
    (if pair
        (setf (cdr pair) window)
      (add-to-list 'window-alist (cons number window)))))

(defun switch-to-number-window (number)
  "Jump to the window  marked with a `mark-window' function"
  (interactive "nNumber Your Window: ")
  (let ((pair (assoc number window-alist)))
    (if pair
        (select-window (cdr pair))
      (message "Invalid Number"))))

;; create 9 keyboard shortcuts using closure (forced using lexical-let)
(dolist (i (map #'1+ (range 9)))
    (global-set-key (read-kbd-macro (concat "C-c C-"
                                            (number-to-string i)))
                    (lexical-let ((i i))
                      (lambda ()
                        (switch-to-number-window i)))))

(global-set-key (kbd "C-c C-0") 'mark-window)

It use range helper function (same as in python)

(defun range (n &optional list)
  "function return list of numbers from 1 to n"
  (if (eq n 0)
    (let ((n (- n 1)))
      (range n (cons n list)))))

EDIT: I just found windmove commands which are much better.

, ,

Leave a comment

%d bloggers like this: