Artifacts

Specifying Artifacts
Specifying Repositories
Downloading Artifacts
Install and Upload

In Buildr, almost everything is a file or a file task. You compile source files that come from the file system using dependencies found on the file system, generating even more files. But how do you get these dependencies to start with, and how do you share them with others?

Artifacts. We designed Buildr to work as a drop-in replacement for Maven 2.0, and share artifacts through the same local and remote repositories. Artifact tasks know how to download a file from one of the remote repositories, and install it in the local repository, where Buildr can find it. Packages know how to create files and upload them to remote repositories.

We’ll get into all of that in a second, but first, let’s introduce the artifact specification. It’s a simple string that takes one of two forms:

group:id:type:version
group:id:type:classifier:version

For example, 'org.apache.axis2:axis2:jar:1.2' refers to an artifact with group identifier org.apache.axis2, artifact identifier axis2, a JAR file with version 1.2. Classifiers are typically used to distinguish between similar file types, for example, a source distribution and a binary distribution that otherwise have the same identifier and are both ZIP files.

Specifying Artifacts

If your Buildfile spells out 'org.apache.axis2:axis2:jar:1.2' more than once, you’re doing something wrong. Repeating the same string over and over will make your code harder to maintain. You’ll know that when you upgrade to a new version in one place, forget to do it in another, and end up with a mismatch.

You can use Ruby’s syntax to do simple string substitution, for example:

AXIS_VERSION = '1.2'

compile.with "org.apache.axis2:axis2:jar:#{AXIS_VERSION}"

Better yet, you can define all your artifacts at the top of the Buildfile and use constants to reference them in your project definition. For example:

AXIS2 = 'org.apache.axis2:axis2:jar:1.2'

compile.with AXIS2

Note that we’re not using a separate constant for the version number. In our experience, it’s unnecessary. The version number intentionally appears at the end of the string, where it stands out easily.

If you have a set of artifacts that belong to the same group and version, and that’s quite common, you can use the group shortcut:

AXIOM  = group('axiom-api', 'axiom-impl', 'axiom-dom',
  :under=>'org.apache.ws.commons.axiom', :version=>'1.2.4')

Buildr projects also define a group attribute which can lead to some confusion. If you want to define an artifact group within a project definition, you should use the explicit qualifier Buildr::group.

If you have several artifacts you always use together, consider placing them in an array. Methods that accept lists of artifacts also accept arrays. For example:

OPENJPA = ['org.apache.openjpa:openjpa:jar:1.2.1',
  'net.sourceforge.serp:serp:jar:1.12.0']
AXIS_OF_WS = [AXIS2, AXIOM]

compile.with OPENJPA, AXIS_OF_WS

Another way to group related artifacts together and access them individually is using the struct shortcut. For example:

JAVAX = struct(
  :activation  =>'javax.activation:activation:jar:1.1',
  :persistence =>'javax.persistence:persistence-api:jar:1.0',
  :stream      =>'stax:stax-api:jar:1.0.1',
)

compile.with JAVAX.persistence, OPENJPA

In our experience, using constants in this manner makes your Buildfile much easier to write and maintain.

And, of course, you can always place your artifact specifications in a separate file and require it into your Buildfile. For example, if you’re working on several different projects that all share the same artifacts:

require '../shared/artifacts'

When you use require, Ruby always looks for a filename with the .rb extension, so in this case it expects to find artifacts.rb in the shared directory.

One last thing. You can also treat artifact specifications as hashes. For example:

AXIS = { :group=>'org.apache.axis2', :id=>'axis2', :version=>'1.2' }
compile.with AXIS
puts compile.dependencies.first.to_hash
=> { :group=>'org.apache.axis2', :id=>'axis2',
     :version=>'1.2', :type=>:jar }

Specifying Repositories

Buildr can download artifacts for you, but only if you tell it where to find them. You need to specify at least one remote repository, from which to download these artifacts.

When you call repositories.remote, you get an array of URLs for the various remote repositories. Initially, it’s an empty array, to which you can add new repositories. For example:

repositories.remote << 'http://www.ibiblio.org/maven2/'

If your repository requires HTTP authentication, you can write,

repositories.remote << URI.parse("http://user:password@repository.example.com")

repositories.remote << { :url => "http://repository.example.com", :user => "user", :pass => "password" }

If you need to use a proxy server to access remote repositories, you can set the environment variable HTTP_PROXY to the proxy server URL (use HTTPS_PROXY for proxying HTTPS connections). You can also work without a proxy for certain hosts by specifying the NO_PROXY environment variable. For example:

$ export HTTP_PROXY = 'http://myproxy:8080'
$ export NO_PROXY = '*.mycompany.com,localhost,special:800'

Alternatively you can use the Buildr options proxy.http and proxy.exclude:

options.proxy.http = 'http://myproxy:8080'
options.proxy.exclude << '*.mycompany.com'
options.proxy.exclude << 'localhost'

All the artifacts download into the local repository. Since all your projects share the same local repository, you only need to download each artifact once. Buildr was designed to be used alongside Maven 2.0, for example, when migrating projects from Maven 2.0 over to Buildr. By default it will share the same local repository, expecting the repository to be the .m2/repository directory inside your home directory.

You can choose to relocate the local repository by giving it a different path, for example:

repositories.local = '/usr/local/maven/repository'

That’s one change you don’t want to commit into the Buildfile, so the best place to do it is in the buildr.rb file in the .buildr directory under your home directory.

Buildr downloads artifacts when it needs to use them, for example, to compile a project. You don’t need to download artifacts directly. Except when you do, for example, if you want to download all the latest artifacts and then go off-line. It’s as simple as:

$ buildr artifacts

Mirrors

You can specify mirrors to override remote repositories. This is useful when you use a Nexus proxy or Artifactory, for example.

You can use the same syntax as repositories.remote, for example:

repositories.mirrors << 'http://corporateserver001.com/repository'

This is even more useful when you place this in your user settings.

See the Settings/Profiles section.

Downloading Artifacts

Within your buildfile you can download artifacts directly by invoking them, for example:

artifact('org.apache.openjpa:openjpa:jar:1.2.1').invoke
artifacts(OPENJPA).each(&:invoke)

When you let Buildr download artifacts for you, or by invoking the artifact task yourself, it scans through the remote repositories assuming each repository follows the Maven 2 structure. Starting from the root repository URL, it will look for each artifact using the path group/id/version/id-version.type (or …/id-version-classifier.type). The group identifier becomes a path by turning periods (.) into slashes (/). So to find org.apache.axis2:axis2:jar:1.2, we’re going to look for org/apache/axis2/axis2/1.2/axis2-1.2.jar.

You’ll find a lot of open source Java libraries in public repositories that support this structure (for example, the Ibiblio Maven repository). And, of course, every remote repository you setup for your projects.

But there are exceptions to the rule. Say we want to download the Dojo widget library and use it in our project. It’s available from the Dojo Web site, but that site doesn’t follow the Maven repository conventions, so our feeble attempt to use existing remote repositories will fail.

We can still treat Dojo as an artifact, by telling Buildr where to download it from:

DOJO = '0.2.2'

url = "http://download.dojotoolkit.org/release-#{DOJO}/dojo-#{DOJO}-widget.zip"
download(artifact("dojo:dojo:zip:widget:#{DOJO}")=>url)

Explaining how it works is tricky, skip if you don’t care for the details. On the other hand, it will give you a better understanding of Buildr/Rake, so if not now, come back and read it later.

We use the artifact method to create an Artifact task that references the Dojo widget in our local repository. The Artifact task is a file task with some additional behavior added by Buildr. When you call compile.with, that’s exactly what it does internally, turning each of your artifact specifications into an Artifact task.

But the Artifact task doesn’t know how to download the Dojo widget, only how to handle conventional repositories. So we’re going to create a download task as well. We use the download method to create a file task that downloads the file from a remote URL. (Of course, it will only download the file if it doesn’t already exist.)

But which task gets used when? We could have defined these tasks separately and used some glue code to make one use the other. Instead, we call download with the results of artifact. Essentially, we’re telling download to use the same file path as artifact. So now we have two file tasks that point to the very same file. We wired them together.

You can’t have more than one task pointing to the same file. Rake’s rule of the road. What Rake does is merge the tasks together, creating a single file task for artifact, and then enhancing it with another action from download. One task, two actions. Statistically, we’ve doubled the odds that at least one of these actions will manage to download the Dojo widget and install it in the local repository.

Since we ordered the calls to artifact first and download second, we know the actions will execute in that order. But artifact is slightly devilish: when its action runs, it adds another action to the end of the list. So the artifact action runs first, adds an action at the end, the download action runs second, and downloads the Dojo widget for us. The second artifact action runs last, but checks that the file already exist and doesn’t try to download it again.

Magic.

SSL and Self-signed certificates

There’s always that Maven repository you learnt to hate, because it’s using a faulty SSL certificate, or a self-signed one.

On top of installing that certificate everywhere, it’s messing with your build!

To get out of there, you can use the environment variable SSL_CA_CERTS to point at a folder containing your certificates.
For example:

export SSL_CA_CERTS=/Users/john/certs
buildr package

You can also change the OpenSSL verify mode so it won’t barf on your certificate. Use the environment variable SSL_VERIFY_MODE to specify one of the following: VERIFY_NONE, VERIFY_PEER, VERIFY_CLIENT_ONCE, VERIFY_FAIL_IF_NO_PEER_CERT. See OpenSSL::SSL for more info.

For example:

# Don't verify certificates
export SSL_VERIFY_MODE=VERIFY_NONE
buildr package

Install and Upload

Generally you use artifacts that download from remote repositories into the local repository, or artifacts packaged by the project itself (see Packaging), which are then installed into the local repository and uploaded to the release server.

Some artifacts do not fall into either category. In this example we’re going to download a ZIP file, extract a JAR file from it, and use that JAR file as an artifact. We would then expect to install this JAR in the local repository and upload it to the release server, where it can be shared with other projects.

So let’s start by creating a task that downloads the ZIP, and another one to extract it and create the JAR file:

app_zip = download('target/app.zip'=>url)
bean_jar = file('target/app/bean.jar'=>unzip('target/app'=>app_zip))

When you call artifact, it returns an Artifact task that points to the artifact file in the local repository, downloading the file if it doesn’t already exist. You can override this behavior by enhancing the task and creating the file yourself (you may also want to create a POM file). Or much simpler, call the from method on the artifact and tell it where to find the source file.

So the next step is to specify the artifact and tell it to use the extracted JAR file:

bean = artifact('example.com:beans:jar:1.0').from(bean_jar)

The artifact still points to the local repository, but when we invoke the task it copies the source file over to the local repository, instead of attempting a download.

Use the install method if you want the artifact and its POM installed in the local repository when you run the install task. Likewise, use the upload method if you want the artifact uploaded to the release server when you run the upload task. You do not need to do this on artifacts downloaded from a remote server, or created with the package method, the later are automatically added to the list of installed/uploaded artifacts.

Our example ends by including the artifact in the install and upload tasks:

install bean
upload bean

Calling the install (and likewise upload) method on an artifact run buildr install. If you need to download and install an artifact, invoke the task directly with install(<artifact>).invoke.

We’ll talk more about installing and uploading in the next chapter, but right now we’re going to package some artifacts.