I need some guidance and clarification about how artifacts are declared in gradle tasks, I would like to upload files to maven/artifactory reading these files from any kind of output container instead of having to hardcode the path for each artifact generated.
This is simpler if you have a JAR, ZIP, File, and a few other types, because you can make use of project.artifacts.add(...), but this is not trivial when you have some random file types.
Im going to provide an example to specify and clarify what I need exactly:
if I have:
task generateMyFile {
doLast {
buildDir.mkdirs()
['touch', new File(buildDir, 'myfile.random').absolutePath].execute().waitFor()
}
}
What's the right way to declare myfile.random as the output for generateMyFile ?
How do I declare myfile.random to be a valid artifact that will be used by maven/artifactory later by the 'publish' task ? currently I assume the file was generated in build/myfile.random, but Im looking for a smarter solution, and gradle documentation (or any other source) is not clear about this.
Thanks in advance
Related
I'm packaging my java application using Gradle's Distribution plugin. I wanted to make 2 distributions, one which doesn't include a JRE and another one that bundles a JRE with the app.
I've set up a copyJre task and wanted to only make Distributions plugin include a folder (jre-8 in the example below) only when copyJre task is in the tasks graph. Here's my attempt which doesn't work.
distributions {
main {
contents {
from('/') {
include 'tools/**'
}
// my attempt to conditionally copy
// jre-8 directory only when tasks graph contains
// a task named 'copyJre'
if (tasks.findByName('copyJre') != null) {
from('../../jre-dist/') {
include 'jre-8/**'
}
}
}
}
}
There probably should be a better approach in general. This looks like kludges.
From a Gradle perspective, you are better expressing what you need the other way around:
Create a different distribution that will include the JRE, possibly extracting the common part of the copy spec.
And if you really only want a single output, make it replace the default distribution after building it.
I'm trying to understand gradle distribution. In the gradle documentation, section 7.3.4, there's the following code example:
task dist(type: Zip) {
dependsOn spiJar
from 'src/dist'
into('libs') {
from spiJar.archivePath
from configurations.runtime
}
}
I was trying to find method dist() to understand what exact it does. I was searching for in org.gradle.api.tasks.bundling.Zip but there is no such method. So where is it declared?
Could You please provide a link to the example You mentioned?
It seems that dist() method is defined nowhere. The code sample You provided is just a task definition so in this particular case dist is just a name of the defined task of type Zip. from and into methods are taken from AbstractCopyTask.
EDIT
So, as stated above dist() is just an ordinary task definition, while for instance the next piece of code in the example:
artifacts {
archives dist
}
has a dedicated method defined in AbstractProject class:
public void artifacts(Closure configureClosure) {
ConfigureUtil.configure(configureClosure, getArtifacts());
}
Now, why this error:
Could not find method dist() for arguments [{type=class org.gradle.api.tasks.bundling.Zip}, txt, build_275gv6pdo8dsig251h253koq9t$_run_closure2#a81512] on proj ect ':MP'.
occurs for this input:
task dist(type: Zip, 'txt')
?
During script parsing (which is a dynamic and quite complicated process) the declaration above should be turned into invocation of one of create methods on TaskContainer instance, where dist is a task name of type String and type: Zip, 'txt' should be passed as arguments. As You can see in the docs for TaskContainer there's no create method that takes String (dist - task name), then Map (type: Zip - task config), and again a String (txt - redundant/invalid argument). That's why it's failing.
If You're interested how it works, it's good idea to put the following piece of code in build.gradle:
task someTask {
throw new RuntimeException()
}
and investigate the stacktrace. It will tell You a lot about how it works step by step.
There is no dist method. In this example, you are using the gradle dsl to create a new task called "dist" whose type is "Zip".
This is accomplished through the use of "method missing". See http://groovy.codehaus.org/Using+methodMissing+and+propertyMissing for more info.
Note: if you change "dist" to "foo", this is still a valid example, but with a less self-explanatory task name.
I have a Gradle build script that successfully builds my project and compiles all the artifacts I need.
However, in a couple of cases I'd like to give other developers the option to patch some of the files. For example, in one of the archives there's an xml file with information about database hooks - some of the devs use other versions (or even engines) and need to change these before they can use the build output.
Instead of having them make changes to a version-controlled file, which they might commit by mistake, I'd like to give them the option to have a local, individual patch file which the build script applies.
In an old ant script, we did something like this
<target name="appcontext-patch" if="applicationContext.patch.file">
<patch patchfile="${applicationContext.patch.file}" originalfile="${dist.dir}/applicationContext.xml"/>
</target>
but I can't figure out how to do the equivalent in Gradle. Is there a better (i.e. more idiomatic) way of doing this than trying to directly convert this into a call to ant.patch?
Some context
This is how the file ends up in the archive in the first place:
into('META-INF') {
from 'deployment', {
include 'applicationContext.xml'
rename { fn -> "jboss-spring.xml" }
}
}
It would be fantabulous if I could just do something like
into('META-INF') {
from 'deployment', {
include 'applicationContext.xml'
rename { fn -> "jboss-spring.xml' }
patch 'local/applicationContext.xml.patch'
}
}
and have the patch file applied before the file is put in the archive. I don't mind writing some code to make this possible, but I'm quite new to Gradle and I have no idea where to begin.
You should be able to translate your ant call into gradle pretty directly.
The gradle doc on how to do this generically. Basically attributes become named arguments and child tags become closures. The documentation has a bunch of good examples.
Once you have your translated ant task you can put in in a doFirst or doLast block on an appropriate task.
My first guess would be something like this:
apply plugin: 'java'
assemble.doFirst {
ant.patch(patchfile: applicationContext.patch.file,
originalFile: "${dist.dir}/applicationContext.xml")
}
That's untested, so but I'm pretty sure it will get you started on the right path. The intent is that just before the java plugin assembles your archive you want gradle to call a closure. In this case the closure will perform an ant action that patches your xml.
Alternately you could use the task you have above that performs a copy and tag onto that.
task myCopyTask(type: Copy) {
...
} << {
ant.patch(patchfile: applicationContext.patch.file,
originalFile: "${dist.dir}/applicationContext.xml")
}
In this case you are writing the task yourself and the left-shift operator (<<) is equivalent to .doLast but a whole lot cooler. I'm not sure which method you prefer, but if you already have a copy task that gets the file there in the first place, I think doLast keeps the relevant code blocks as close to each other as possible.
RFC 5621 defines an XML patching language that uses XPath to target the location in the document to patch. It's great for tweaking config files.
There is an open source implementation in Java (Disclaimer: I am the author). It includes a filter that can be used from Gradle to patch XML files during any task that implements CopySpec. For example:
buildscript {
repositories { jcenter() }
dependencies { classpath "com.github.dnault:xml-patch:0.3.0" }
}
import com.github.dnault.xmlpatch.filter.XmlPatch
task copyAndPatch(type: Copy) {
// Patch file in RFC 5621 format
def patchPath = 'local/applicationContext-patch.xml'
inputs.file patchPath
into('META-INF') {
from 'deployment', {
include 'applicationContext.xml'
rename { 'jboss-spring.xml' }
filter(XmlPatch, patch: patchPath)
}
}
}
If you'd like to do this more on the fly I can think of two main techniques. Both involve writing some code, but they may be more appealing to you and I'm pretty confident gradle doesn't have this behavior built-in anywhere.
Personally I think #1 is the better solution, since you don't need to muck around with the internals of the Copy task. A custom filter feels cleaner and more reusable.
1) Write a custom filter that you specify in your copy task. I can't help with the details of how to write a custom filter, but I'd start here. You should be able to put the custom filter in buildSrc (lots of info about that at gradle.org) and then you simply need to import it at the top of your gradle file. If you write it in groovy I think you can even just use ant.patch() again.
task copyAndPatch() {
into('META-INF') {
from 'deployment', {
include 'applicationContext.xml'
rename { fn -> "jboss-spring.xml' }
filter(MyCustomFilterThatDoesAPatch, patchFile: 'local/applicationContext.xml.patch')
}
}
2) Write a custom task. Again, I'll leave the details to the experts but you can probably get away with subclassing the Copy task, adding a 'patch' property, and then jumping in during execution to do the dirty work.
Following on from this question.
If I have a build with two instances of the Test task, what is the best (cleanest, least code, most robust) way to completely separate those two tasks so that their outputs don't overlap?
I've tried setting their testResultsDir and testReportsDir properties, but that didn't seem to work as expected. (That is, the output got written to separate directories, but still the two tasks re-ran their respective tests with each run.)
Update for the current situation as of gradle 1.8: The testReportDir and reportsDir properties in dty's answer are deprecated since gradle 1.3. Test results are now separated automatically in the "test-results" directory and to set different destination directories for the HTML reports, call
tasks.withType(Test) {
reports.html.destination = file("${reporting.baseDir}/${name}")
}
Yet again, Rene has pointed me in the right direction. Thank you, Rene.
It turns out that this approach does work, but I must have been doing something wrong.
For reference, I added the following to my build after all the Test tasks had been defined:
tasks.withType(Test) {
testReportDir = new File("${reportsDir}/${testReportDirName}/${name}")
testResultsDir = new File("${buildDir}/${testResultsDirName}/${name}")
}
This will cause all instances of the Test task to be isolated from each other by having their task name as part of their directory hierarchy.
However, I still feel that this is a bit evil and there must be a cleaner way of achieving this that I haven't yet found!
Ingo Kegel's answer doesn't address the results directory, only the reports directory. Which means that a test report for a particular test type could be built that includes more test results than just that type. This can be addressed by setting the results directory as well.
tasks.withType(Test) {
reports.html.destination = file("${reporting.baseDir}/${name}")
reports.junitXml.destination = file("${testResultsDir}/${name}")
}
Just an update. The reports.html.destination way is deprecated.
This is the "new" way (Gradle > 4.x):
tasks.withType(Test) {
reports.html.setDestination file("${reporting.baseDir}/${name}")
}
I have a multi-project C++ Gradle build, which produces a number of libraries and executables. I'm trying to get the executables (but not the libraries) subprojects to get compiled in with a 'fingerprint' object. This works fine if I sprinkle smth like this in individual subprojects' build.gradle:
compileMain.doFirst {
// code to generate a 'BuildInfo.cpp' from from a template.
// embeds name of executable in so has to be generated anew for each exe
}
Following DRY principles, I'd much rather do this once and for all in a top level build.gradle. This is my attempt, to apply it to just the subprojects that use the cpp-exe plugin, following these instructions:
configure(subprojects.findAll { it.plugins.hasPlugin('cpp-exe') }) {
compileMain.doFirst {
// same code as above
}
}
Alas, this doesn't get triggered. However, if I put smth like this in a less restrictive configure, block, this demonstrates that the idea of querying the plugin should work:
configure(subprojects.findAll { true }) {
task mydebug << {
if ( project.plugins.hasPlugin( 'cpp-exe' ) ) {
println ">>> $project.name has it!"
}
}
}
Could it be that the plugins don't get applied to the subprojects at the time the configure closure is evaluated (in the top-level build.gradle)? There may well be a much simpler way of achieving this altogether?
You probably apply the cpp-exe plugin in the child projects' build scripts. By default, a parent build script gets evaluated before its children, which explains why it's not finding any projects that have cpp-exe applied.
There are several ways to solve this problem. One way is to move all configuration that's specific to a cpp-exe project (like applying the plugin and adding the action) to the same spot. Either you do all such configuration from the parent build script (for example by enumerating the cpp-exe subprojects and configuring them with a single configure(cppExeProjects) { ... }), or you move the cpp-exe specific configuration into its own build script (say gradle/cpp-exe.gradle) and apply it from selected subprojects like so: apply from: "$rootDir/gradle/cpp-exe.gradle".
Another solution is to change the evaluation order of build scripts. But I would only use this as a last resort, and it is certainly not necessary here.
Gradle 1.5 is recently out, I am not sure if this is a new feature but as it looks, you can solve the issue by using afterEvaluate.
Take a look at section 53.6.1 in http://www.gradle.org/docs/current/userguide/build_lifecycle.html
Something like:
subprojects {subProject ->
afterEvaluate {
if ( subProject.plugins.hasPlugin('cpp-exe')){
println "Project $subProject.name has plugin cpp-exe"
}
}
}
would give you a start.