If you’re interested it was created using Code Swarm and MEncoder.

I want tests to be:

• Easy to write (and hence read)
• Low maintenance, i.e. I don’t want to have to update tests whenever I make changes that aren’t important to the test
• Good at picking up unexpected behaviour rather than changes to HTML structure

Running your GSP from TestApp

Grails ships with a handy class called GroovyPagesTestCase which allows your test to get the output of a given GSP file given a defined input, like this:

def file = new File("grails-app/views/myview.gsp")
def model = [someVariable:12345]
def htmlString = applyTemplate( file.text, model )

We’re passing in the text from the GSP file as a template, along with a model comprised of whatever variables and mock objects the view should need.

Now I’ve got a string containing a bunch of HTML, okay, that’s the right direction but my lazy gene is not satisfied yet.

Note 1: If your template calls other templates then it makes life easier to use absolute URLs, e.g. templateUrl"/>

Note 2: This method assumed you’ll specify an explicit model for each sub-view, e.g. myObj:myObj]}"/>

From a sticky HTML mess to something useful

The easiest way to get from unparsed HTML to a useful searchable structure seems to be Groovy’s XMLSlurper. It parses XML rather than HTML by default, but you can instantiate it with a more HTML-friendly parser like TagSoup like so:

def slurper = new XmlSlurper( new org.ccil.cowan.tagsoup.Parser() )
def parsedHtml = slurper.parseText(text)

Easy.

Pulling it all together

import grails.test.GroovyPagesTestCase
import org.ccil.cowan.tagsoup.Parser

class GspXyzTests extends GroovyPagesTestCase
{
	boolean transactional = false
	def slurper = new XmlSlurper( new Parser() )

	// This test looks for a specific thing in the resulting parsed HTML
	void testSomeImportantOutput() {

		//Open file containing GSP under test
		def file = new File("grails-app/views/myfolder/template.gsp")
		assertNotNull(file)

		//Obtain result of GSP page
		def text = applyTemplate(file.text, [
			pagenumber:123,
			pagesize:10,
			someMockObject:[
				foo:"bar",
				nestedMockObject:[
					[id:12345],
					[id:67890]
				]
			]
		])

		def html = slurper.parseText( text )

		// test some aspect of the parsed structure, the trick would be to make the test resilient to a degree of cosmetic change
		assertEquals 1, html.head.meta.list().findAll{ it.@name?.text().toLowerCase() == "robots" }.size()

	}

}

Being a big fan of data visualisation I decided to have a go at representing the data in a way that can be more easily absorbed. So, armed with the source data, a list of ISO country codes, and the docs for the Google Chart API, I started playing.

The biggest question when visualising data, just like with statistics, is deciding what you’re looking for. This data is complex enough to be difficult to show in its entirity, involving maybe a dozen or so possible pieces of information for each location.

Here I’ve opted to look at the difference between the age of consent for straight couples and gay couples.

The Elgin, via flickr/Ewan-M

There’s a theory that claims expertise and associated salaries increase more rapidly in cities than they do in the country because physical proximity decreases the cost of sharing ideas (I’m desperately trying to dig up the source amongst the noise; which is either ironic or poetic depending on how you look at it).

The interwebs are a different beast. Proximity doesn’t exist in the same way, perhaps instead becoming a cultural rather than a geographic measure of separation. The cost of spreading an idea on the internet is more related to trust than physical distance.

On the internet, that trust must be earned by expertise and clear communication. In the physical world people behave differently, people tend to trust each other more quickly and be more open when they can look each other in the eye, or buy each other a beer.

[Only on a technology blog would you find a justification this obscure for going to the pub.]

On Tuesday we held this month‘s London Open Source Search Social at The Elgin in Notting Hill. This was the first time we’d used our shiny new Meetup account to organise the event, so it was nice not to have to send out reminders manually (laziness #ftw).

A few notes from the evening for those whose memories are as bad as mine

There’s plenty missing, and some of this may be fictitious.

Bruno from Jobomix talked about his use of Hadoop to detect duplicate job data, leading to a conversation about Pig and Cascade, then other distributed systems like Scala. Ben from OneIS brought up the subject of Duby, a Ruby-like-but-tidier language targeting the JVM, and when prompted gave us an outline of his company’s free-text graph store.

We talked about duplicate detection in various fields, thresholds, and the cost of false positives. We touched on human relevance testing; Richard told us he’d found people generally need to be paid to do it and not for more than 30 minutes at a time.

Joao from the Royal Library of the Netherlands told us how they digitise and index millions of pre-digital documents per month. Ben told us about a method of querying Xapian from Postgres using an SQL JOIN.

While the first reaction when a potential competitor folds is to celebrate, that reaction is both immature and short-sighted. Not only is it a sad state of affairs that one company can destroy another using just the costs of the patent infringement process (regardless of whether the patent is valid or the infringment is demonstrable), but since competition is the fire that drives progress and innovation; seeing competition fold for any reason other than poor products is always disappointing.

Daniel Tunkelang has a eulogy of Modista over on the Noisy Channel.

Respect to AJ Shankar, Arlo Faria, and any others on the team; you guys did some impressive work.

Google Chrome advertising (via flickr/iainpurdie)

As any self-respecting nerd will have noticed, and others have already noted, Google recently started advertising its Chrome web browser on billboards and in newspapers around the UK. This represents an escalation of the second phase of the browser wars, and one of the few occasions Google has resorted to billboards to advertise a product.

Why bother advertising a free product?

The answer to why Google are advertising Chrome (which is a free download) is unsurprisingly similar to the answer to the bigger question; why bother building and supporting a free product?

Google make money by monetising user’s searches. People are great at optimising finding and using short-cuts, and modern browsers have built-in search bars. In short, more people using your search bar means more money, and Chrome (like Firefox) defaults to searching on Google.

Billboards – dated but still relevant

Let’s face it, it’s not Google’s style to put up great big billboards. It’s not smart, it’s not targeted, it’s not high-tech. However, ironically those attributes are exactly why they work in this situation.

Google’s main competitor in the search space is Microsoft (who have incidentally been advertising their search engine Bing heavily) and Microsoft’s largest user-base is the slow-moving majority who get Internet Explorer bundled with their PC. Via its default status in Internet Explorer Bing is by default used by that same slow-moving majority.

Since the majority is too big to be worth the extra cost of targeting; the common or garden billboard is a suitable way to get through to them (at the same time as reinforcing the brand with nerds who already know about it).

What does it do?

This site has a very similar model to our own consumer-facing MAST app; Empora. It makes money by sending consumers to retailer sites, who for obvious reasons are willing to pay for suitable traffic. The main forces that influence the design of a site like this are retention, and the clickthrough and conversion rates of your traffic:

Retention – you need to impress people, then ideally remind them to come back to you

Clickthrough – you need to send a good proportion of visitors to retailers in order to make money

Conversion – if the visitors you send aren’t interested in buying the clicked product then the retailers won’t want to pay for that traffic on a per-click basis (although they might be interested in the CPA model, which doesn’t pay until someone buys)

First Impressions

People’s first impressions are usually determined by a combination of design and how well a site conforms to their expectations. I’ve probably got distorted expectations considering my experience working with this type of application, but in that respect I was pleasantly surprised; Shopachu has some good features and makes them known. In terms of design I was less impressed, the icons and gel effects don’t seem to fit and I think there are whitespace and emphasis issues (sorry guys, trying to be constructive).

Finding stuff

It’s fairly easy to find things on Shopachu. The filters are easy to use (although I could get the brand filter to work, could be a glitch). The navigation is pretty easy, although it doesn’t currently provide second generation retail search features like facet counts (i.e. showing the number of products in a category before you click on it).

The biggest interesting technological problem I’ve noticed with their navigation is the colour definitions. There’s a big difference between a colour being present in an image, and the eye interpreting that colour as being present in an image. I think there are some improvements to be made in the way colours are attributed to images (e.g. here I’ve applied a pink filter but am seeing products with no pink returned). Similarly there’ll be another marked improvement with better background removal (e.g. here I’d applied a light blue filter and am seeing products with blue backgrounds).

Similarity search

Shopachu’s similarity search is quite different to Empora’s.  They’ve chosen to opt for maximum simplicity in the interface rather than user control, resulting in a single set of similarity search results. In contrast, Empora allows users to determine whether they’re interested in colour similarity, or shape similarity, or both. Simplicity often wins over functionality (iPod example #yawn) so it’ll be interesting to see how they do.

Another issue is the quality of the input data. This challenge is the same for Empora, or anyone else aggregating data from third parties, in that category information is inconsistent. One effect of this is that when looking at the similarity results for an often poorly-classified item like a belt you may also see jewellery or other items that have been classified as “accessories” or “miscellaneous” in the retailer’s data, another effect is that you often see duplicate items.

Keeping the traffic quality high

An interesting design decision for me is that the default image action on Shopachu is a similarity search, i.e when you click on the image it takes you to an internal page featuring more information and similar products. This is in contrast to the default action on Empora or Like.com, which is to send the visitor to the retailer’s product page.

The design trade-off here is between clickthrough and conversion rates. If you make it easy to get to the retailer your clickthrough rate goes up, but run the risk of a smaller proportion converting from a visit to a purchase. Here Shopachu are reducing that risk (and also the potential up-side) by keeping visitors on their site until they explicitly signal the intent to buy (the user has to click “buy” before they’re allowed through to the retailer).

Getting people hooked

There are a few features on Shopachu aimed at retention, namely Price Alerts and the ability to save outfits (Polyvore style). These features seem pretty usable, although I think they’re still lacking that level of polish that inspires passionate users. I’d be interested to know what the uptake statistics look like.

In summary

I think this implementation shows that Incogna have thought about all the right problems, and I think have clearly got the capability to solve the technological issues. On the down-side; cleaning up retailer’s data is a tough business which will be time-consuming, and I think they need to find a little inspiration on the visual design side.

Sign up on the London Search Social page to get notifications of events.