Peter Ledbrook

Pub language discussion in London

For those of you that aren’t members of the London Java Community there has been a pretty active discussion going on through the mailing list about learning a new language (and which one to learn).

The conversation will no doubt be continuing at the monthly social event (which goes out to members of the Clojure, Scala & Groovy communities in London) tomorrow night. Feel free to come down and fight for Groovy’s corner: http://www.meetup.com/Londonjavacommunity/calendar/13588662/

London GGUG, 17th May – DSLs in Groovy & Grails 1.3

I’m pleased to announce that the next London GGUG on the 17th May will have a guest from the US: Peter Bell. He’ll be giving a talk on the practical development of a Groovy DSL with emphasis on the practical side. As he says on his blog, this is for both people that want to go further with DSL development and those that simply would like to see in what ways a DSL can help them.

I’ll be speaking on the new features that come with Grails 1.3, focusing in particular on plugins deployed to Maven-compatible repositories.

You can register for the event on the SkillsMatter website, where you can also see synopses of the talks and the speaker bios. Please register soon and I hope to see you there!

Testing Groovy code – an undesirable burden?

Many of the complaints I’ve heard directed at Groovy and Grails derive from the same issue: the compiler doesn’t pick up type errors. People worry that simple typos will make it into production and that they’ll be less productive due to MissingMethod and MissingProperty exceptions popping up when they run the application.

The standard answer to this is a simple one: write tests for your code. In fact, prefer test-driven (or test-first) development. Yet this doesn’t seem to satisfy some people. “What projects do you know that have 100% test coverage?” One or two that have close to 100%. But are the rest reliable? No, not even the Java-based ones. Any code that isn’t tested has a significantly high chance of being buggy. How can this be a serious argument against dynamic languages?

That’s not to say I think dynamic languages should take over from static languages. Both have their place and deciding which to use for any given task will often come down to personal preference. They both have their strengths and weaknesses, but a discussion of those will have to come another time.

Too much testing!

Another argument I’ve heard is that dynamic languages require more testing because you have to explicitly test for type errors that would otherwise be picked up by the compiler. I think this is frankly rubbish. What do we do when we’re testing? We’re checking that code behaves as expected given certain inputs. If there are any type errors in the code under test, you won’t see the expected behaviour.

Let me demonstrate. Say we have a Grails controller {{MyController}} with a corresponding unit test. The test starts simply:

package org.example

class MyControllerUnitTests extends grails.test.ControllerUnitTestCase {
    void testIndex() {
        controller.index()
    }
}

All we do is invoke the index action on the test controller. Now see what happens when we run the test against this controller code:

package org.example

class MyController {
    def index = {
        def id = parms.id
        render "You have reached ID ${ids}"
    }
}

If you run the test in SpringSource Tool Suite, you’ll see something like this:

As you can probably tell, the test is failing because the variable parms does not exist. Not surprising since this was a deliberate typo. Once you correct that typo, the test fails on the next one, ids. Change that to id and the test is now passing.

So, even though the test simply executes the action and doesn’t even bother checking the results, we get feedback on missing properties and methods. Granted, the feedback isn’t as instantaneous as you get in an IDE with a statically typed language, but in my experience it’s not a huge hit on productivity. And while we’re on IDEs, I know that both STS and IntelliJ IDEA will underline properties and methods it can’t resolve, so you would probably notice that parm and ids were mistyped before running the test.

This doesn’t really do much for my argument yet because I’ve just shown that you need an “extra test” to get the same checks you’d get with a static language. But remember that the aim of a test is to ensure that code behaves as it should, so we should check the action renders the string we expect! That’s easily done:

package org.example

class MyControllerUnitTests extends grails.test.ControllerUnitTestCase {
    void testIndex() {
        controller.params.id = 10
        controller.index()

        assertEquals "You have reached ID 10", mockResponse.contentAsString
    }
}

The test still passes, but it now checks that the action is rendering the appropriate string to the response. The key point I want to make here is that the test is no different than it would be if we were testing Java code, yet it will still pick up typos.

What about type-checking?

I may have justified my position when it comes to missing methods and properties, but what about real type errors, such as using a string where an integer is expected or vice versa? This is an interesting question because in dynamic languages, it’s typically the wrong one to ask. How can you have a type error in a dynamically typed language? You might have some code like this:

def n = 100
n = n.substring(1)

which you may see as a type error (a number is being treated like a string), but as far as the language is concerned all we have is a missing method on the value of n. This is important because you could very well add the method substring() to integers, in which case the code would work:

Integer.metaClass.substring = { start ->
    return delegate.toString().substring(start)
}

def n = 100
n = n.substring(1)

In other words, “type errors” in a dynamic language are simple missing method or property exceptions, which I’ve already demonstrated will be covered by your normal tests (without any extra shenanigans).

I would stop there, but that isn’t the whole story when it comes to Groovy. After all, it has static types! It also frequently interfaces with (statically typed) Java code. How does this affect our testing? Ah, I wish there was a simple answer to this. In some ways, passing the incorrect type to a method will be picked up. Try running this script:

def someMethod(String str) {
    println "Some method: $str"
}

someMethod(100)

Boom! You’ll get a missing method exception because Groovy can’t find an instance of someMethod() that takes an integer. Unit tests can handle cases like this, but in practice such problems raise their head once you start wiring your objects together and they start interacting with each other. At that point, you have to start thinking about integration tests – a topic for another time.

Have I convinced you?

The main thrust of this blog post has been to highlight that your bog standard unit tests, which you can run easily from your IDE, will pick up those typos and “type errors” you’re worried about without you having to do any extra work. All you need to do is focus on writing tests that check the behaviour of your code. It’s not trivial to write thorough tests, but that’s independent of the language you use.

Something else you should bear in mind is that It’s a lot easier to test code written in a dynamic language because you don’t have to always create objects of the correct type – they just have to have the appropriate methods and properties. So dynamic languages both strongly encourage you to adopt best practice and make that best practice easier.

Despite all this, I understand that it can be difficult in many environments to write tests for code, let alone write tests before writing any code. But I think this is a problem of culture and habit. Once you get used to writing tests, and then writing them first, the whole process becomes easier. You then find you have more confidence in your code. I can only strongly recommend that everyone takes any opportunity they can to try this approach out and get used to it. You never know, the managers may even come round to realising it’s a good idea!

GitHub Pages: publish your Grails plugin docs

Currently, the primary site for Grails plugin documentation is the plugin portal on the Grails web site. The trouble with the portal is that it doesn’t readily support multi-page documentation. On top of that, you can use gdoc to write a user guide for your plugin, but where will the generated manual go?

Thanks to Luke Daley, I discovered GitHub Pages as a potential solution for those plugins I have hosted on GitHub. This feature allows you to create simple websites on a per-project basis.

Getting started

Before you can start publishing websites to GitHub, you first have to create a ‘personal’ page. The process is straightforward: create a new GitHub repository, clone it locally, add an index page, and then push your additions to GitHub.

For example, let’s say your GitHub username is ‘franklin’. First, create a new GitHub repository with the name franklin.github.com. This repository will store your personal website. Once that’s ready, clone it and add an index page:

$ git clone git@github.com:franklin/franklin.github.com.git franklin.github.com
$ cd franklin.github.com
$ touch index.html
$ git add .
$ git push origin master

After the above steps, you will have a blank home page for http://franklin.github.com/. Note that it can take a while for your GitHub Pages site to activate. Also, you can access the URL http://<username>.github.com/ before you perform the above steps and you will see some instructions for setting up the various pages.

Now that you have a personal Pages site, you can create project sites with URLs of the form http://<username>.github.com/<projectname>/. The process for this is trickier, so I refer you to the instructions on the GitHub Pages. You can either do it via your local git, or you can got to the project page on GitHub, click on the “admin” button, and select the “Generate project page” link. I prefer the latter approach because it’s easier and involves less typing!

Once you complete the steps from either set of instructions, you will end up with an extra local tracking branch called ‘gh-pages’. This branch contains the website for your project and it’s where we will be storing the plugin documentation.

The plugin documentation

When the grails doc command first appeared, it allowed you to automatically generate API documentation for your projects. With Grails 1.2, its scope was expanded and it can now generate a manual that looks and feels just like the Grails user guide. You don’t have to do anything to get the API documentation, but the manual requires a bit of work.

The basic idea is that you create a src/docs/guide directory and populate it with gdoc files. By prefixing the names of the gdoc files with appropriate numbers, you get a hierarchy of numbered sections. To get some idea of what you need to do, check out the source files that Luke Daley created for the Grails JMS plugin user guide.

Publishing the docs

Generating the documentation leaves you with a docs directory full of HTML files. What do you do with it? The next step is to get it onto your project-specific GitHub Pages site. Doing this manually is a bit of a bind, but I’ve created a Grails script and some template files that make it almost trivially easy (once you have the ‘gh-pages’ branch locally!).

First, copy this script to your $HOME/.grails/scripts directory. Next, switch to your ‘gh-pages’ branch and copy this HTML template to the root directory of the project. Next, create an index page for your site: index.html, index.textile, and index.markdown are all acceptable. Choose whichever format you prefer, but it should start with this section:

---
layout: main
title: My Grails Plugin
---

The three dashes are important! I’ll explain those and the layout/title later. Then,

$ git add main.html.tmpl
$ git add index.textile
$ git commit -m "Adding template layout and index page for GitHub Pages site."

In the above set of commands, replace ‘index.textile’ with the name of the index file you created.

You’re now ready to publish your plugin docs for the first time! Simply switch to the main branch and execute the new Grails command:

$ git checkout master
$ grails publish-github

Once the command successfully completes, you will find yourself back on the ‘gh-pages’ branch with all the documentation staged for a commit. You can then make any changes you want, commit everything, and then push the branch to GitHub. You can even do this in one step with the grails publish-github --push command, or if you want to automatically commit the documentation but not push, use grails publish-github --commit.

A few details

The publish-github command does a few things. First, it generates the plugin’s documentation via Grails’ doccommand. It then copies the generated documentation to the ‘gh-pages’ branch (using an rather ugly technique). Once it’s on the ‘gh-pages’ branch, it generates a new ‘main’ layout based on the main.html.tmpl file. It basically reads the information from the XML plugin descriptor, plugin.xml, and populates the template with things like the plugin version and author. The template is then generated to _layouts/main.html, which you can reference from your individual pages.

It’s up to you to write the site’s pages, but at least you have the option of three different syntax types. I prefer Textile for the moment, because it’s very similar to the gdoc format. The key part of each page is the header block, which you’ve seen already:

---
layout: main
title: My Grails Plugin
---

The ‘layout’ value is used to determine which file under _layouts to inject the page into. Ideally, the index page should use the _layouts/main.html template that the publish-github command generates, hence the value ‘main’. The ‘title’ option is specific to the generated template, which inserts the corresponding value into the title of the page. This is because it’s tough to get a suitable plugin name with correct capitalisation from the plugin descriptor.
Update Somebody has pointed out that the publish-github command doesn’t work if plugin.xml is committed to your main branch (typically master). You could try modifying the script to handle this, but the XML plugin descriptor doesn’t need to be in GitHub at all – it’s generated after all – so I would simply git rm it.

That’s pretty much it. You also have the full power of Jekyll (as described on the GitHub site), so you can build more elaborate sites as you see fit. If you want to see the results of all this jiggery-pokery, then take a look at the GWT plugin pages.

Until next time.

Java’s new keyword

Creating objects in Java is easy with the new keyword. In fact, it’s one of those things that you don’t think about. Need to access a file? Just create a new File instance: new File("build.properties"). For most Java developers, that’s all they need to know. Life becomes more interesting, though, when you start working with multiple class loaders.

Class loaders? Argh! Run away, run away!

That was pretty much my reaction for many a year. I just didn’t want to know about them. They were some kind of black magic and always Somebody Else’s Problem. It’s strange, because class loaders are actually pretty straightforward. Most Java developers know that you compile Java files to these *.class files and that those compiled classes have to be loaded by the JVM somehow. That’s basically what the class loader does. But like threads, the problem is not understanding what they do, but getting them to work together.

How many times have you heard the phrase “it’s a class loader issue?” I’ve certainly heard (and said) it more times than I’d care to admit. As soon as you have more than one class loader in an application, you have to start worrying about which classes can “see” which others. It can easily become a nightmare. But class loader behaviour is perhaps a post for another time. Let’s get back to new.

So, the first time that you create a new object, the JVM has to first load the class. This happens transparently when you use new. The question is, what class loader is used? And why does it matter?

Consider a scenario from Grails. We have a build system based on Gant that loads build scripts and executes them. In one of them, we instantiate a Jetty server and start it. The sequence of object creation goes like this:

In fact, the above is a simplification of what actually happens, but it suits the purpose of this post.

The JARs for the first three classes are all on the classpath of what we will call the build class loader. This loads all the classes used directly by the build. So what about Jetty’s Server class? The most important thing to understand is that the Server class must be loaded by the same class loader that loads the Grails web application. Although you can pass your own class loader to the embedded server, if it’s different to the one that loads Server you’ll run into those dreaded class loader issues.

Bearing that in mind, let’s look at what happens if the RunApp script uses new to create the server instance:

def server = new org.mortbay.jetty.Server()
...
server.start()

Right about now, you should be asking yourself “what class loader was used to load the Server class?” It’s a critical question because it determines what class loader is used to load the entire web application and hence what classpath the application’s runtime dependencies should be on. In this case, the class loader used is whichever one loaded the RunApp script. The new operator effectively delegates to this.getClass().getClassLoader().

What does that mean for our example? It means that the build class loader is used to load the Server class and therefore must also be used to load the web application classes. In other words, all the application’s runtime dependencies must be included in the build class loader! What’s the problem with that, you may ask. There is one potential problem and one actual.

The potential problem is class conflicts. What if the web application depends on a different version of a library that’s already on the build system? It’s a particular problem if any of the Apache XML API libraries are on the classpath. These cause absolute havoc.

The other problem is that the more JARs you have on the classpath, the longer it takes for the JVM to find the class it’s after. That means longer start up times. It’s one of the problems OSGi was designed to solve (he was told by a man in the pub). Why put JARs on the build classpath that the build itself doesn’t need?

The solution is to work out where you want a class loader boundary and use reflection to instantiate your object:

def runtimeClassLoader = new URLClassLoader(...)
def server = runtimeClassLoader.loadClass("org.mortbay.jetty.Server").newInstance()
...
server.start()

This is pretty easy in Groovy because the start() method is evaluated at runtime, but Java needs to know the type at compile-time. You can’t do this:

ClassLoader runtimeClassLoader = new URLClassLoader(...)
Server server = (Server) runtimeClassLoader.loadClass("org.mortbay.jetty.Server").newInstance()
...
server.start()

because you’ll get a ClassCastException on line 2. The declared type of server is loaded by this.getClass().getClassLoader(), whereas the new instance is loaded in a different class loader. Different class loader means different classes. So you have to use reflection to invoke the methods and access the fields you need. Fortunately, you only have to jump through these hoops at class loader boundaries.

As you’ve seen, the new operator is normally something you don’t have to think about, but as soon as you start dealing with multiple class loaders, you have to be aware of and understand its behaviour. The trick is to work out suitable class loader boundaries and then use reflection to load and instantiate classes at those boundaries. It may sound like unnecessary extra work, but you can gain real improvements in application/framework reliability. If you’re lucky, things may even run a bit faster