Development at paulcarvill.com, the home of Paul Carvill on the web

link: paulcarvill on twitter

link: paulcarvill at flickr

paulcarvill.com

Hi, I'm Paul Carvill, I'm a web developer. I'm currently working as Technical Lead at LBi, Europe's largest digital agency.

I also like walking, cooking, Bollywood and rock 'n' roll.

Archive for the ‘Development’ Category

A collection of links about JavaScript and the MVC development pattern

Wednesday, August 19th, 2009

I’ve been doing most of my JavaScript development over the past 12 months in a Model-View-Controller pattern. This separation of function has a multitude of benefits for a programmer — it enables writing of clearer, more readable code, can result in more modular, reusable code and, perhaps most importantly, it enables a Test-driven approach to development.

Now A List Apart has published their own JavaScript MVC tutorial. It’s very useful, and very easy to follow. I recommend it as a starting point if you’re interested in this method of development.

Developing JavaScript in an MVC pattern is an especially good fit for web developers familiar with the now widespread, if not quite ubiquitous, separation of form and content in their programming methodology. Something like the paradigm which you apply to delivery of web content using HTML to provide data and CSS to provide the look and feel can be applied to JavaScript in an MVC pattern, where the Model provides the data, as might a server-side database, the Controller plays the part of the web server delivering that data and the View is the browser itself, rendering content to the page.

I’d also encourage developers to roll their own MVC framework rather than dive right in and using one of the existing ones, at least initially. It will give you a much better understanding of the pattern, it’s benefits and probably pitfalls, and will let you comfortably learn to extend your programming prowess at your own pace.

Note: you could also implement JavaScript on the server side using MVC, with the caveat that your View function is possibly less conceptually decoupled, but once you’re happy with the pattern this shouldn’t hamper you too much.

A List Apart: JavaScript MVC by Jonathan Snook
“At this stage of JavaScript and Ajax development and adoption, we need to consider separating our code’s components—MVC-style. Such separation may not be necessary in every situation—in some cases, it may even make things needlessly verbose. As our applications get more complex and require JavaScript interactions across multiple parts of our sites, however, separating JavaScript into Model, View, and Controller parts can produce more modular, reusable code.”

Ben Godfrey: Client-side MVC is maturing
“MVC is the dominant model for UI development in the desktop world. A modified form is very popular in web development. It makes a lot of sense to stick to MVC for RIA development. RIAs running in the browser (or Flash or AIR) bear close resemblance to desktop applications.”

A list of JavaScript MVC libraries and associated stuff:

I don’t use any of these, but they might be helpful for developers in a hurry. To be honest, the vast majority of my stuff is really VCS (View, Controller, Service), where the Service layer may be a remote API or HTML fragment which you needn’t model, but adding a Service layer isn’t a great leap once you’ve made the conceptual jump from more basic procedural JavaScript programming to MVC.

IE6 state of play linkage

Friday, August 14th, 2009

A few links to recent IE6 coverage.

Digg the Blog: Much Ado About IE6
“This goes directly to why most folks use IE6: they don’t have a choice. Three out of four IE6 users on Digg said they can’t upgrade due to some technical or workplace reason…Giving them a message saying, ‘Hey! Upgrade!’ in this case is not only pointless; it’s sadistic.”

BBC News | Technology: Microsoft backs long life for IE6 [support will continue until 2014, four years beyond their original deadline]
“‘Friends do not let friends use IE6,’ said Amy Barzdukas, Microsoft’s general manager for Internet Explorer.
‘If you are in my social set and I have been to your house for dinner, you are not using IE6,’ she said. ‘But it is much more complicated when you move into a business setting.’”

blogs.msdn.com | IEBlog: The Engineering Point Of View
“The engineering point of view on IE6 starts as an operating systems supplier. Dropping support for IE6 is not an option because we committed to supporting the IE included with Windows for the lifespan of the product. We keep our commitments.”

MoD to stick with IE6 despite security concerns
“According to parliamentary written answers received by Labour MP Tom Watson, the majority of [government] departments still require staff to use IE6. Most have plans to upgrade to the more secure IE7, and some to IE8, but the Ministry of Defence (MoD) has no plans to change.”

Conclusion: IE6, at least in corporate environments, isn’t going away for a few years yet.

HTML5 HTML text semantics granularity

Thursday, June 18th, 2009

Wow, HTML5 HTML semantic text description options are so granular I had to spend several minutes pondering whether my previous code snippet about relaxing Apache permissions warranted <code>, <kbd>, or <samp> elements, or a combination of all three.

In the end I settled on a <kbd> element for the bit I want you to type in (opening a file in vi from the command line), as that’s the bit you’re going to type. For the contents of the file I chose a <samp> element, as the text shown in the file is a sample of the output of my file, rather than a chunk of code you need to enter.

The difference between <code> and <samp> is very small, but it’s great that we actually now have this level of specificity, which should help ensure that HTML5 HTML is robust enough to last well into the future.

Geocoding location data in a Google spreadsheet

Wednesday, June 3rd, 2009

The problem: I have a spreadsheet full of locations, addresses and place names that I want to publish, along with a map, for at least tens of thousands of people to view.

A solution: Easy — I can put it in a Google spreadsheet, publish it, add a Google map to a page, download the data, geocode the locations and display them on the map.

Another problem: While this is ok in most cases, with a large spreadsheet the geocoding can take a very long time, making my page appear unresponsive and slow. In addition, I have no way of checking that the location data is good enough to map with.

Another solution: Download the data, geocode it using Yahoo!’s Placemaker service, generate a new spreadsheet containing accurate latitude/longitde data and use that in place of the original. The client then does no geocoding their side, it’s all supplied along with the data. Everybody’s happy!

— Go straight to the spreadsheet geocoder! —

I’ve done just that with this PHP script. It takes a Google spreadsheet key, and you must tell it what columns your location data is in. It will download the spreadsheet data, concatenate those location columns, make a request to Placemaker to geocode each location, and return a new CSV file with the geodata columns appended on the end.

I’ve detailed here the various bits that make up the script. The workflow is as follows:

Capture spreadsheet data from user > Load in spreadsheet from Google > For each line in spreadsheet make a Placemaker request > Append geolocation data columns to spreadsheet > Output all results into a CSV file

The script is set to not autodisambiguate, meaning that if it’s not sure what location you’ve supplied, it will return all likely candidates, in order of likelihood. I should mention that Yahoo!’s Placemaker is utterly awesome in find out the ‘whereness‘ of things.

To build your own version of my script will need a Placemaker API key. Other than that, please feel free to copy and paste the code, fix it, amend it and let me know if it’s useful, or if it needs more commenting, or how I could improve it. I wrote this code to fix a particular problem I was encountering, but I’m sure it could work in a few more cases too.

Something to note before I start: the script doesn’t much like having commas in the location data in your spreadsheet. Because Google only output CSV with a comma delimiter, this upsets my CSV parsing. Any suggestions welcome.

This function gets some CSV data from a published Google spreadsheet using a supplied key:


<?php

function getCsvDataFromGoogle($spreadsheetKey) {
	$key = $spreadsheetKey;
	$output = 'csv';
	$apiendpoint = 'http://spreadsheets.google.com/pub?key='.$key.'&output='.$output;
	$ch = curl_init();
	$options = array(CURLOPT_URL => $apiendpoint,
	                 CURLOPT_HEADER => false,
	                 CURLOPT_RETURNTRANSFER => true
	                );
	curl_setopt_array($ch, $options);
	$r = curl_exec($ch);
       curl_close($ch);
	return $r;
}

This function makes a Placemaker geocode request:


function getPlacemakerGeodata ($location) {
	$key = 'MY_PLACEMAKER_API_KEY';
	$apiendpoint = 'http://wherein.yahooapis.com/v1/document';
	$inputType = 'text/plain';
	$outputType = 'xml';
	$focus = '28298150'; // sets focus to Great Britain, not sure how effective this is yet
	$autoDisambiguate = 'false'; // returns the 1 most-likely place, else returns many likely places
	$post = 'appid='.$key.'&documentContent='.$location.'&documentType='.$inputType.'&outputType='.$outputType.'&focusWoeid='.$focus.'&autoDisambiguate='.$autoDisambiguate;
	$ch = curl_init($apiendpoint);
	curl_setopt($ch, CURLOPT_POST, 1);
	curl_setopt($ch, CURLOPT_POSTFIELDS, $post);
	curl_setopt($ch, CURLOPT_RETURNTRANSFER, 1);
	$results = curl_exec($ch);
	return $results;
}

This function does the bulk of the work, and makes calls to all the other functions:


function parseCsvData($googleSpreadsheetKey) {
	$lines=split( "\n", getCsvDataFromGoogle($googleSpreadsheetKey) );
	if($_POST['format'] == 'csv') {
		if($_POST['locationColumns'] == '' || $_POST['key'] == '') {
			echo "please go back and specify both your google spreadsheet key and which columns contain your location data (in comma separated format, zero-indexed e.g. 0,1,9)";
			exit();
		}
		else {
			// get location columns from url
			$locations = $_POST['locationColumns'];
			$splitLocations = split(',', $locations);
			// set headers to 'csv'
			header("Content-type: application/csv;");
			header("Content-Disposition: attachment; filename=yourgeodata.csv");
			$out = fopen('php://output', 'w');
			for($i=1;$i

This function parses the XML which gets returned from Yahoo! Placemaker:


function parsePlacemakerXML($results, $delineator) {
	if($delineator == 'comma') { $delStart = ''; $delEnd = ','; }
	else { $delStart = '<td>'; $delEnd = '</td>'; }

	$places = simplexml_load_string($results, 'SimpleXMLElement', LIBXML_NOCDATA);
	$locarr = array();
	if($places->document->placeDetails) {
		foreach($places->document->placeDetails as $p) {
			if($delineator == 'comma') {
				$locarr[] = $p->place->name;
				$locarr[] = $p->place->centroid->latitude;
				$locarr[] = $p->place->centroid->longitude;
				return $locarr;
			}
			else {
				echo $delStart.$p->place->name.$delEnd;
				echo $delStart.$p->place->centroid->latitude.$delEnd;
				echo $delStart.$p->place->centroid->longitude.$delEnd;
			}
		}
	}
}

This bit runs when you load the page and works out if you're submitting some data or just viewing the page. If you've submitted data, it runs the main function:

if(ISSET($_POST['submit'])) {
	parseCsvData($_POST['key']);
}

Or if you're viewing the page for 1st time, you get a form to fill out:

else {
	echo "<html><head><title></title></head>";
	echo "<body>";
	echo "<p>Please enter your spreadsheet key and specify which columns contain your location data (use comma separated list e.g. 9,10,11):</p>";
	echo "<form method=POST><p><label>Key:<input type='text' name='key' /></label></p><p><label>Location columns: <input type='text' name='locationColumns' /></label></p><p><label>Format: <select name='format'><option value='csv'>csv</option><option value='table'>table</option></select></label></p><p><input type='submit' name='submit' /></p></form>";
	echo "</body></html>";
}
?>

Why people should build their own URL shorteners

Sunday, April 5th, 2009

Lots of blogposts regarding the evil that URL shortening services do appeared this weekend, from Jason Kottke (“URL shorteners suck”), Joshua Schachter, creator of Delicious (“on url shorteners”) and Dave Winer (“Josh is right, URL shorteners are risky”).

None of them are happy, with concerns, variously, about spam, speed, efficiency, transparency, longevity and what Joshua calls “the great linkrot apocalypse”.

I think the one idea that we should take from all of these arguments is that if you are using a URL shortening service you have no control over your links, now or in the future. One such service may get bought up by a nasty commercial entity who redirects all your existing short URLs to its own ends. Your URLs pointing to all your lovingly curated content will, effectively, become spam.

One solution to this quandry, which no one has mentioned, and one which large media organisations should pay attention to, is that building a URL shortener is really, really easy. I spoke to Simon Willison about it and he thinks a URL shortener will soon be the example app that someone builds to learn their way around a new language or web framework, like they currently do with a creating a blog. That way you get to solve many problems at once: control over your own links’ destiny and complete consumer confidence that your own-brand short URLs (gu.com/abcde, nyt.com/vwxyz) won’t take them someone nasty.

And the ability to shorten your own URLs isn’t necessarily restricted to large companies with lots of resources. Many people who want to use this sort of service already have all the tools they need — their blogging software. All that Movable Type or Wordpress, among others, need to do is add an extra database lookup table and pretty soon all their users can take care of their own URL shortening needs.

UPDATE: A useful comparison of existing URL shortening services (even the method of redirection matters: 301 is a permanent redirect, 302 a temporary one, with implications for SEO and link credit).

Using Python with MySQL on Mac OSX 10.5 Leopard

Wednesday, March 18th, 2009

If you try to build the MySQL Python library on Mac OS X 10.5 (Leopard) you’ll get an error similar to this:

/usr/include/sys/types.h:92: error: duplicate ‘unsigned’
/usr/include/sys/types.h:92: error: two or more data types in declaration specifiers
error: command 'gcc' failed with exit status 1

I found the fix for this error here: http://www.keningle.com/?p=11, via a comment on this site: http://dotnet.org.za/ncode/archive/2007/01/31/setting-up-mysql-for-python-mysqldb-on-mac-os-x-2.aspx. It’s just a couple of lines in Terminal, adding a symlink so the library knows where to look to find the files.

Some background: Unless you’re using SQLite, you need to install a Python library to interface with your chosen database (PostgreSQL, MySQL or Oracle).

The MySQL library can be downloaded here: http://www.djangoproject.com/r/python-mysql/

But it seems there are a few problems running this library on Mac OS X 10.5 (Leopard), hence the above fix.

You’ll miss me when I’m gone: IE6, cross-browser consistency and device independence

Saturday, February 14th, 2009

A flurry of IE6 related activity on the web this week coincided with a discussion we are having at The Guardian on the same subject. We have been talking about the relative benefits of keeping website performance in IE6 consistent with that of other browsers, and the disproportionate amount of work this requires on the parts of developers and the QA team. We’ve been trying to figure out better processes to reduce the number of styling bugs in IE6, while not compromising the user experience or the hard work put in by our design team.

It turns out people have surprisingly strong views on cross-browser consistency. For some, IE6 represents much more than just ‘a browser’. It also represents, variously: a large market share; an important group of corporate users; a user’s freedom to choose whichever device she wishes to browse the web. Once you start dropping a browser for technological reasons, the argument goes, you might as well arbitrarily drop support for anything which you consider below par – mobile browsers, text browsers, people with small monitors.

The opposing view says that IE6 is many years old and two versions out of date, a huge security risk and a drain on resources. We shouldn’t be pandering to slow or paranoid IT departments who refuse to upgrade their systems. Anyway no one chooses to use IE6, it is forced upon them by said IT departments.

I’m loath to branch the code to produce a separate version of the site for any reason, be it a device or a browser. But I also see the amount of pain IE6 causes developers, especially when they’re trying to do something fancy with JavaScript, and even more especially trying to do so without using a standard library which might easily provide you with cross-browser methods for doing stuff.

I support IE because I have to. But I do also believe strongly in wide accessibility, through as many devices as possible. We should assume nothing — nothing — about how our users access the web. But I don’t think this is the point here. The point here is adhering web standards, which apply to both code and content. Remember, the content itself — the information — usually isn’t broken. It’s what you’re trying to do with it that’s broken. The CSS and the JavaScript. Go back to Tim Berners-Lee’s 2002 document on universality and device independence for a lesson in what putting stuff on the web is all about. Work with the web, not against it. It’s really good at presenting and sharing text and pictures. But it’s not a magazine layout. Berners-Lee once said,

“Anyone who slaps a ‘this page is best viewed with Browser X’ label on a Web page appears to be yearning for the bad old days, before the Web, when you had very little chance of reading a document written on another computer, another word processor, or another network.”

We can infer from this that a site isn’t ‘best viewed in’ anything: it’s just ‘viewed’, however it might end up. So, yes, your site might look lovely, but if getting it there is so complex that it breaks browsers, or takes up 50% of your development time, then you’re plainly doing something wrong.

Try taking your page back to basics, get rid of the awful advertisement JavaScript and the three different kinds of page tracking, and start paying more than just lip-service to web standards and accessibility. That XHTML doctype declaration you’re using, trying adhering to it. There, it probably works a lot better now, yes?

But ultimately, and as usual, I think the whole issue comes down to a business decision: how much time/money are we spending on development versus how much money that development brings in. It’s a brave person who decides to cut off 25% of their users.

Some points that came up as part of our ongoing discussion:

  1. Should the design be 100% consistent across all browsers, or would our designers be happy to sacrifice certain style elements? We currently stop a code release if something looks bad in IE6, although we have already made one or two decisions to remove an element from IE6 in order to expedite a code release. In both cases we ran things past the Guardian’s Creative Editor, Mark Porter, before doing so.
  2. If you want to drop suport for IE6, you have to completely and utterly drop support for it. And in all likelihood never look at it again. Because the next time you do, it will be horrifically broken. Stopping development on that browser doesn’t just mean it won’t get cool new features. It still gets the features, but they won’t be tailored to it, and will break it. That smart Javascript widget you just wrote? That breaks the page in IE6. Some new element you put in with a fixed width and margins? That breaks the page in IE6. You have to cut the cord. Be strong, give it a firm handshake and say goodbye.
  3. Turns out Microsoft haven’t quite cut the cord yet, though. Microsoft support Windows XP Service Pack 3 as a current product (it shipped in April 2008), and will retire support for it 2 years after the next service pack is released, or at the end of the Windows XP product lifecycle, whichever comes first. IE6, which shipped as a component of XPSP3, continues to have Mainstream Support as part of that product:
  4. Our current browser usage figures look like this:
    • IE 7: 35%
    • IE 6: 25%
    • Firefox 3: 25%
    • Safari: 7%
    • Firefox 2: 3%
    • Google Chrome: 1.5%
    • Opera: 0.5%
  5. We currently have a problem even testing in IE6, because the corporate build on the PCs we use doesn’t contain it, it has IE7 as standard. And you can’t run IE7 and IE6 concurrently. Ironically, our technical infrastructure is sufficiently advanced that we have difficulty supporting old technology.

That flurry of activity in full: