Weird characters transmitted to and from server through jQuery AJAX call

A simplified example of my app:
I have a HTML form as the output of a PHP script that gets a text from a database and fills an input of that form with it. There I can edit the text that on form submit is sent to a PHP script via a jQuery AJAX call. Through PHP the text is saved in the database and then the saved value is retrieved in PHP and sent in the JSON result of the AJAX call.
The character encoding of the HTML page is ISO-8859-1:

Let’s say the he HTML form looks like this:

<form id="my_form">
<input type="text" id="txtId" name="txt" value="" />
<input type="submit" name="btn" value="save">
</form>

On form submit this AJAX call is made:

$.ajax({
	type: "POST",
	url: "my_script.php",
	data: $("#my_form").serialize(),
	success: function (jsonObj) {
		if(!jsonObj) {
			return;
		}
		if("txt" in jsonObj) {
			$("#txtId").val(jsonObj[txt]);
		}
		return false;
	},
	error: showError,
	dataType: "json"
});

In PHP, after saving the text in the database and retrieving the saved text, I add it to an associative array which I convert into a JSON object displayed as the response of the AJAX call:

$item['txt'] = $value; //$value is the text saved in the database
header("Content-type: application/json");
echo json_encode($item);

At first I submited the form with exactly the text that came from PHP at page load. It looked fine, bun when reloading the page, a weird text filled my form input. I had the text a×b and now I got a×b.
I submitted the form again with text a×b and I studied the AJAX call response with Firebug. In both Console and Net tabs of Firebug, under Post tab of the call everything looked fine, but under Response tab I got “txt”:”a\u00d7b” instead of “txt”:”a×b”.
It looked like the text somwhere on the way back to the form got encoded in a weird manner. The × character is a Windows-1252 encoded character, not an UTF-8 encoded character and I should discover where the encoding of the text changed.
I submitted the correct text again and outputted the text saved in the database without json_encode-ing it:

echo $item['txt'];

In Firebug, in the Response tab of the AJAX call, even though under the Console tab of Firebug the text looked fine, under the Net tab it appeared like this: a×b.

Again I submitted the correct text and outputted the value that came via AJAX:

echo $_POST['txt'];

And again I obtained the correct text under the Console tab and the incorrect one under the Net tab in Firebug, which meant that the encoding broke before the text got to the server.

Then, under Headers tab of the call, I noticed among the Request Headers: Content-Type application/x-www-form-urlencoded; charset=UTF-8 and I thought maybe setting the character encoding of the jQuery.ajax call to ISO-8859-1 instead of UTF-8 would solve my problem:

$.ajax({
	type: "POST",
	url: "my_script.php",
	data: $("#my_form").serialize(),
	contentType: "application/x-www-form-urlencoded;charset=ISO-8859-1",
	success: function (jsonObj) {
		if(!jsonObj) {
			return;
		}
		if("txt" in jsonObj) {
			$("#txtId").val(jsonObj[txt]);
		}
		return false;
	},
	error: showError,
	dataType: "json"
});

But the result remained the same and even more, the Content-Type header did not change either. After lots of thinking and testing, I came to these conclusions:
1. If the data parameter of the jQuery.ajax call is not empty and the type parameter is set to “POST”, the character encoding of the request remains UTF-8 no matter what, so (if I want my encoding to take effect) what I would normaly put in the data parameter I should add to the query string of the url of the AJAX call and not specify or leave the data parameter empty (setting the value of empty string to the data parameter).
2. Explicitly setting the character encoding of the AJAX request to ISO-8859-1 didn’t help at all with my problem.
3. jQuery serialize function `messes up` special characters that are not UTF-8 encoded, because it uses JavaScript function encodeURIComponent which UTF-8-encodes special characters, so make sure to UTF-8-decode the texts in the server script when using jQuery serialize or JavaScript encodeURIComponent function in an AJAX call.

So I left the JavaScript code as it initially was (without specifying the contentType parameter to the jQuery.ajax call) and, in the PHP code, decoded the string before saving it in the database:

$txt = utf8_decode($_POST['txt']);

By now I have the correct text saved in the database, but another problem arises: the text in the response of the AJAX call is null. But why?
PHP function json_encode only works with UTF-8 encoded characters, that’s why. So I should have may own JSON-maker function:

function make_json($item) {
	foreach($item as $key => $value) {
		if(is_array($value)) {
			$arr[] = '"'.$key.'":'.make_json($item[$key]);
		} else {
			$arr[] = '"'.$key.'":"'.str_replace(array("\\", "\""), array("\\\\", "\\\""), $value).'"';
		}
	}
	return '{'.implode(",",$arr)."}";
}

And the code in the PHP script becomes:

header("Content-type: application/json");
echo make_json($item);

But now I get this weird result: a�b (diamond shaped character with question mark inside instead of special characters). This time in the Net tab of Firebug everything looks fine, while the diamond shaped characters appear in the Console tab and in the HTML page.

I solved it by explicitly setting (in PHP) the character encoding of the response of the AJAX call to ISO-8859-1 (thanks to this post):

header("Content-type: application/json; charset=ISO-8859-1");
echo make_json($item);

base html tag does not apply correctly to all javascript and style files

Problem: The

<base href="http://example.com/path/to/res/">

  did not work at all on Internet Explorer – no javascript or style files were loaded and worked partially in Mozilla Firefox – a few javascript files were not pointing to the path that should be transformed by base href attribute.

Example:

Array(
[0] => 1
)
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
<base href="http://example.com/path/to/res/">
<link rel="stylesheet" href="./css/styles.css" type="text/css" />
<scrupt src="./js/jquery.js" type="text/javascript"></script>
<scrupt src="./js/j1.js" type="text/javascript"></script>
<scrupt src="./js/j2.js" type="text/javascript"></script>

This code produced a page with no style files and  no javascript files included in Interenet Explorer 8 and a few javascript files (j1.js and j2.js) were not included in Firefox 3.5 because base tag was not applied. After several tests and all kind of different arrangements we discovered that the problem was the output before the page source the

Array ...

thing.

Solution: no output before the page’s headers otherwise you can get all kind of strange results including a html base tag not working correctly.

Problem: When submitting a form, seems like another form gets submitted

Actually, I had two forms on the same page, one with the GET method (let’s call it get_form), the other with the POST method (we’ll call it post_form). When I pressed the submit button of get_form it seemed that the post_form got submitted. In fact, all the variables in get_form and post_form were passed with the POST method.

So the natural thing that came into my mind was that the two forms were intersecting. But how? I knew for sure they were separate forms because get_form was in a file that I was including in the file that contained post_form, after post_form.

After verifying the HTML source over and over again I finally solved the mistery. There was a stupid mistake I made: I did not close post_form properly. Instead of using the form closing tag </form> at the end of post_form, i put the form opening tag <form> and every time I checked for the problem, I missed an important detail: the lack of the slash sign.

Conclusion: When opening a HTML element, make sure to close it accordingly.

How to check/uncheck a bunch of checkboxes without using ids for the checkbox inputs

My solution of checking/unchecking a group of HTML checkboxes using Javascript implies using an array of checkboxes, which means naming all the inputs of type ‘checkbox’ like array_name[]. Example:

<form name="cb_form">
<input type="checkbox" name="cb[]" value="0" />Zero
<input type="checkbox" name="cb[]" value="1" />One
<input type="checkbox" name="cb[]" value="2" />Two
<input type="checkbox" name="cb[]" value="3" />Three
</form>

In this example, the name of the checkboxes array is cb.

Next, we place two links for checking/unchecking all checkboxes in our array. If someone clicks one of these links, the checkAll() JavaScript function is called:

<a href="" onclick="checkAll('cb_form', 'cb[]', true); return false;">Check all<a>
<a href="" onclick="checkAll('cb_form', 'cb[]', false); return false;">Uncheck all<a>

The checking/unchecking all checkboxes function in JavaScipt looks like this:

function checkAll(form_name, cb_name, value)
{
var cb_arr = document.forms[form_name].elements[cb_name];
// if the checkboxes exist
if(cb_arr)
{
// if the number of checkboxes is at least 2
if(cb_arr.length > 1)
{
// for each checkbox
for(i = 0; i < cb_arr.length; i++)
{
// check (value == true) or uncheck (value == false) it
cb_arr[i].checked = value;
}
}
else // cb_arr.length is undefined which means there is a single checkbox element that is not considered an array of one element
{
cb_arr.checked = value;
}
}
}

Note that if we only have one checkbox, the variable cb it is not considered an array, but a normal variable. This is useful in the situation of dinamically generated HTML pages (using PHP, for example) and the number of checkboxes varies from page to page.

You can test the example here:

Zero One Two Three

Check all Uncheck all