Pages

Wednesday, 20 August 2014

5 Things I've learnt being a scrum master

I've been a scrum master now for about 6 months. Having been involved in scrum previously as a product owner, as well as a developer, moving into this role has really opened my eyes to some of the more political and arguably awkward elements of trying to get rid of impediments.

Stay calm when others aren't 


Something that I think is really key about being a scrum master, you have to be thick skinned. You have to not only push back, but when people whine at you and bring office politics into play, it's vital that you remember exactly what the end game is, to meet the sprint goal!  

I should also point out, this doesn't mean ruling with an iron fist, as a scrum master, you still succeed with the team and fail with the team.  You can tell when something isn't working and using an agile methodology shouldn't be a painful process, it should be motivating.

You are neither a manager, nor just a developer

It's quite an interesting position to be in as from my experience, it's not unrealistic to get your hands dirty in some code while being a scrum master.  You have to be strong enough to fend people off from poaching your team members even if they're higher up the food chain to yourself.  You aren't "in charge" of the team, but you do have a responsibility to push back on poachers.

Don't be afraid to say "no"

If your product owner is telling you to put 60 points into a sprint, when you know the velocity you've been hitting for the past 4 sprints has consistently been 40 points, don't be afraid to say "what you're asking is unattainable".  It's much better to be honest early on and push back.  Blindly saying yes, promising to the customer and then having to deal with the consequences later on isn't where anyone wants to be.

Make sure stand ups are to the point

This might be like telling grandma to suck eggs, but it's vital that stand ups really are short, sharp and to the point.  There's nothing worse than listening to a stand up where everyone in the back of their mind is thinking "I wish s/he'd just get to the point!".  This is a situation where you have to stick to your guns and if people get offended, they get offended.  You have to tell the person, "we don't need to know any of the extra detail, we just need to know from a high level; what it is you're going to achieve today, did you achieve what you set out to do yesterday and importantly, do you have any impediments".

Keep things moving

Sometimes things in a sprint can get stale, tasks can get stuck in one condition, you need to keep them moving!  As an example, If there's a task that's been coded for half a day but not released to testing, find out why.  You never know, the CI server might be down, there might be a problem releasing, you need to get it out before it becomes an impediment.  Keeping the task board fresh and a true representation of what's actually happening in your sprint can really boost morale if you know there's only a few tasks left and you're literally "sprinting" to the finish :)

Wednesday, 14 May 2014

Jax-RS custom exception handling

One of the nice things about working with JEE is the components available really are pretty standard.  While using JAX-RS, sometimes you need to control how exceptions get processed and fed back to the user.  If an exception is thrown, by default, you'll get some horrible HTTP 500 internal server exception, exposing the internal failings of your web service.

Consider the following gist, this endpoint would be used to view a user based on the Id.

Now, the implementation of this interface would look something like the following, to do the donkey work of actually retrieving the user.

This looks fine, but consider if the userDao was doing some entity business logic, using Query.getSingleResult and the user with this ID didn't exist?

According to the JEE6 API documentation, you'd receive a NoResultException, which would cause an HTTP 500 error exposing your internal server exception, which is definitely something end users shouldn't see. We need to leverage the exception handling of Jax-RS!

First, we need a dumb exception object, appropriately named, which will be what we'll actually throw, consider the code below..

Next, we need to modify our original code to take into account this exception, I've modified the original UserWebService and associated implementation appropriately below.

This will now throw an appropriate exception when a user cannot be found. However, we still need to create a Handler object to convert this exception into an actual JSON response so we get a nice friendly error message. The class below handles this exception and will convert the error message in the exception into a JSON response. The important annotation you'll see on this class is the @Provider annotation.

You'll notice we create an ErrorMessage object to respond to from the web service. This is just a simple dumb object to hold the details of the actual error that'll be marshalled into JSON.

The final step in mapping our exception handler provider to the web-app is to add the following into our web.xml for our webapp.

Now when we call this REST endpoint with a user ID that doesn't exist (let's say "DAG"), we'll happily receive the following JSON response rather than a stacktrace.

Friday, 17 January 2014

Stripping Java string SQL statements automatically

Do you ever get sick of typing out an SQL statement in Java with all the String syntax?

Then, when the statement doesn't work, you have to copy it back into pure SQL to run it to check it works?

This was a pretty common problem for me, so I made a simple syntax stripper http://idavid.co.uk/sqlstripper/ so I could quickly covert from SQL into a Java string and back again.  Consider the following Java String...

String query = "SELECT * " +
"   FROM USERS u" +

"INNER JOIN CODE cd ON cd.userid = u.id " +
"WHERE u.id in ('1','2') ";

This would parse fine if it was being sent into stmt.executeQuery(query)

But it I wanted to run this on SQL server just to see what it produced, I'd have to manually remove all the syntax etc to make it run and generate the following SQL.

SELECT *
FROM USERS u
INNER JOIN CODE cd ON cd.userid = u.id
WHERE u.id in ('1','2')

This is exactly what my SQL stripper does.  It also gives you the ability to paste in an SQL statement into the right hand textbox and convert it into a nicely formatted Java String, so converting the SQL statement above would give...

"SELECT *"
+ " FROM USERS u"
+ " INNER JOIN CODE cd ON cd.userid = u.id"
+ " WHERE u.id in ('1','2')";

Monday, 2 December 2013

Importing a libgdx project into Intellij

I've started getting into Intellij.  I must admit, I'm much more comfortable with Eclipse so it was a bit of a learning curve.  On top of that, I'd never used the libGDX gaming framework for Android.  I found documentation a little sparse online so I thought I'd record a YouTube video just to give a simple step-by-step tutorial on importing the demo projects into IntelliJ and playing them on a device.  This was the first time I'd ever made a YouTube video, so apologies for the audio quality.

Friday, 15 November 2013

Strategy Pattern ain't meant for Spring!

Right, so lets say you're writing a Spring MVC app and you decide, "I want to do seperate encapsulated algorithms that can be swapped to carry out a specific behaviour".

The classic response to this would be "you need the Strategy Pattern ma' boy!".  So, that's what I did, consider the code below...

Interface
public interface MealStrategy {
    cook(Meat meat);
}
First strategy
@Component
public class BurgerStrategy  implements
MealStrategy {
  @Autowired CookerDao cookeryDao;

  @Override
  public void cook(Meat meat) {
      cookeryDao.getBurger(meat);
  }
}
Next strategy...
@Component
public class SausageStrategy  implements
MealStrategy {
  @Autowired CookerDao cookeryDao;

  @Override
  public cook(Meat meat) {
      return cookeryDao.getSausage(meat);
  }
}
Context...
@Component
@Scope("prototype")
public class MealContext {
    private MealStrategy mealStrategy;

    public void setMealStrategy(MealStrategy strategy) {
        this.strategy = strategy;
    }

    public void cookMeal(Meat meat) {
        mealStrategy.cook;
    }
}
Now say this context was being accessed through an mvc controller, like...
@Autowired
private MealContext mealContext;

@RequestMapping(method = RequestMethod.POST)
public @ResponseBody Something makeMeal(Meat meat) {
    mealContext.setMealStrategy(new BurgerStrategy());
    mealContext.cookMeal(meat);
}

Normally, this would make perfect sense for use with the strategy pattern.  However, the one line that makes it all come crashing down is the line in the controller...

mealContext.setMealStrategy(new BurgerStrategy())

This will cause the new instance of the bean to be created outside Spring's application context.  This is curious, so how should you go about implementing a pattern within Spring if you can't use new to create a new instance?  After asking this question on StackOverflow, the conclusion I came to was to use the Dependency Injection pattern (as per the standard autowiring in Spring) and make individual injections of strategies.  Firstly, I'd need to amend the strategies so they have named components.

@Component("burger")
public class BurgerStrategy implements MealStrategy { ... }

@Component("sausage")
public class SausageStrategy implements SausageStrategy { ... }
Now in the controller, I have individual strategies to use, so rather than setting a strategy, I just pick the appropriate one that's injected.
@Resource(name = "burger")
MealStrategy burger;

@Resource(name = "sausage")
MealStrategy sausage;

@RequestMapping(method = RequestMethod.POST)
public @ResponseBody Something makeMeal(Meat meat) {
    burger.cookMeal(meat);
}

Wednesday, 12 June 2013

Java testing with Selenium and dynamic ids in html

One of the cool aspects of Selenium, is that not only can you record yourself using a site, you can actually run this as a junit test.
Firstly, I'll install Selenium in Firefox (as this is the official version) and record a quick test.  It's important to note that Selenium will give you a number of different ways to remember which html tag you invoked.  For instance, it can just invoke a specific id on a page.
However, when using a portal system like say JSF under Liferay, the id values are generated on the fly, so you'd record one test, then never be able to run it successfully again.
One really nice feature of Selenium is you can invoke a HTML xpath so in the Liferay example, your code would still find the tag it needed to click.  Lets say I record myself logging into the page below...


Now because this page is generated with liferay, I can see the input text id for the form is...
As JSF under Liferay will create a new id quite regularly for this textbox (each time the server is restarted I believe, although it may be even more frequent), this means we can't just get the id and hook into that, as the tests will only ever run once. 



























What we can do however is hook into liferay by using the html tag directly as this won't be different each time Liferay loads the JSF.  I noticed I had to use this same technique for every page in Liferay as the id for nearly all the html rendered through JSF had a different id each time the page was accessed.


We can then export this to a junit class from the file menu File | Export Test Case As... | Java / JUnit 4 / Web Driver which would give us the following class to run and test.

Sunday, 12 May 2013

Tomcat deployment script

So this weekend I deployed http://www.playerviewer.com (shameless attempt at self promotion!).  I've spent a lot of time with EC2 on Sonar / Jenkins boxes, but never deployed over Tomcat on it in production other than small tests.

One thing I found during the inital deployment that really started to be a constant theme...

stop tomcat, download war, delete old war folder, move war, start tomcat.

This started to become a pain, so I ended up just using a shell script.


So the maven process in my pom is to move my war to a directory, that then gets uploaded to http://www.mysite.com/deploy/ROOT.war and my shell script does the rest.
There are probably much nicer ways to do this such as using ant to ftp the script to the location for the shell script to download it, but it's not really that important.
Also - all the code from player viewer is available open source here