In my gradle scripts, I've built a task that runs a java process. This process depends on a target property. The task is defined by:
task('bulk', type: JavaExec, dependsOn: 'classes', description : 'Bulk data import on a target (defined by -Ptarget=[event|member|...]]') {
//available imports
ext{
event = relativePath('src/main/scripts/events.csv')
member = relativePath('src/main/scripts/member.csv')
membership = relativePath('src/main/scripts/membership.csv')
}
//check the target is set
doFirst {
if(!project.hasProperty('target')){
println "\nUsage:"
println "\tgradle bulk -Ptarget=[event|member|...]"
println "where target is : "
bulk.ext.each{ println it }
throw new GradleException('Target argument required')
} else {
println "\nBulk import of $target\n"
}
}
main = 'org.yajug.users.bulkimport.BulkImport'
classpath = sourceSets.main.runtimeClasspath
if(project.hasProperty('target')){
bulk{
args target
args bulk.ext[target]
debug false
}
}
}
And to run it:
gradle bulk -Ptarget=event
It's working fine, but know I have to run this process for different targets:
gradle bulk -Ptarget=event
gradle bulk -Ptarget=member
gradle bulk -Ptarget=membership
...
How can I group all these calls into an other single task with the gradle's dependency model ? (I know the list of targets)
SOLUTION
task bulk;
['event','member','membership'].each {target ->
task("bulk${target}", type: JavaExec, dependsOn: 'classes', description : "Bulk data import of ${target}s") {
//available imports
ext{
event = relativePath('src/main/scripts/events.csv')
member = relativePath('src/main/scripts/member.csv')
membership = relativePath('src/main/scripts/membership.csv')
}
//check the target is set
doFirst {
println "\nBulk import of $target\n"
}
main = 'org.yajug.users.bulkimport.BulkImport'
classpath = sourceSets.main.runtimeClasspath
args target
args ext[target]
debug false
}
bulk.dependsOn("bulk${target}")
}
How can I group all these calls into an other single task with the gradle's dependency model ?
You can't, because a task (instance) will be executed at most once per build. Instead, the way to go is to declare multiple task instances. You could do this by putting the task declaration in a loop, putting it in a method and calling it multiple times, or by writing a task class and instantiating it multiple times. Then you'll add one other task that depends on all bulk tasks. Assuming the execution order between bulk tasks is irrelevant, that's it.
By adding a helper method or two, you can create a nice little API around this, to improve readability and allow reuse in other places.
Another way to tackle this is with a task rule. You can read more on task rules in the Gradle User Guide.
What about creating a -Ptarget=all, and deal with that specific case in the build.gradle file. This could be done as:
['event','member','membership'].each { t ->
task("bulk${t}", ...) {
onlyIf project.hasProperty("target") && (project.getProperty("target").equals(t) || project.getProperty("target").equals("all"))
args target
args bulk.ext[target]
debug false
}
}
Related
I am working with a build.gradle file that has multiple ways to specify executions for a task - setup. To call a task from another gradle file - runtests.gradle, I created a task - testTask and added task dependency using dependsOn, but this implementation does not seem to work and giving out an error like :
Could not find property 'testTask' on root project 'GradleFile
My build file looks like this :
build.gradle
task setup(dependsOn: testTask) <<
{
println "In main execution"
}
// new task
task testTask(type: GradleBuild) {
if (getEnvironmentVariable('RUN_TEST').equalsIgnoreCase("true")) {
buildFile = "../Behave/runtests.gradle"
tasks = ['mainTask']
}
else {
println "Exiting runTests Task"
}
}
setup.doFirst {
println "In first execution"
}
setup.doLast {
println "In last execution"
}
D:\>gradle -q GradleFile/build.gradle setup
I am not looking to make much changes to existing tasks, so is there any other workaround I should try?
I have been through many links but could not find anything that suits this scenario. Looking for suggestions please.
Gradle is sensitive to the ordering of tasks in the build script if a task instance is given in the dependsOn. The task setup depends on task (instance) testTask which, at the moment the build script is compiled, doesn't exist yet. The most common options to solve the issue are:
Define task setup below testTask:
task testTask(type: GradleBuild) {
}
task setup(dependsOn: testTask) {
}
Use a relative path to the task, i.e. the task's name, in the dependsOn
task setup(dependsOn: 'testTask') {
}
task testTask(type: GradleBuild) {
}
Please find more details in Javadoc of Task.
Our build has some generic logic to create tasks from configuration data. There are also several Gradle files on whose execution order I do not want to depend.
Now I need to add a task dependency to a task without knowing if that task will be defined before my part of the build script runs (at configuration time) or after my part.
[Edit: as #canillas writes I can say myTask.dependsOn "otherTask" and "lazily" depend on yet undefined "otherTask", but I need it the other way around.]
Since I can't write tasks["otherTask"].dependsOn myNewTask before "otherTask" is defined, I made the following helper function:
void safelyDoWithTask(String taskName, Closure func) {
def task = tasks.findByName(taskName)
if (task != null) {
func(task)
} else {
project.tasks.whenTaskAdded { task_ ->
if (taskName.equals(task_.getName())) {
func(task_)
}
}
}
}
safelyDoWithTask('otherTask', { it.dependsOn myNewTask })
Now I wonder if there is more canonical way to achieve this?
Consider the following:
// section 1
def dynamicDependencies = [:]
dynamicDependencies['otherTask'] = 'myNewTask'
// section 2
afterEvaluate { project ->
// taskA dependsOn taskB
dynamicDependencies.each { taskA, taskB ->
def task = project.tasks.findByName(taskA)
if (task) { task.dependsOn "${taskB}" }
}
}
// section 3
task myNewTask() {
doLast {
println 'myNewTask !'
}
}
task otherTask() {
doLast {
println 'otherTask !'
}
}
The gist is:
section 1 defines our dependency info in a custom map
section 2 uses the afterEvaluate hook to process the map
because the above is decoupled from the task definitions, we can simply put them in section 3 (or wherever)
I need to create an initialize task that will run before all other task when I execute it.
task A {
println "Task A"
}
task initializer {
println "initialized"
}
If I execute gradle -q A, the output will be:
>initialized
>Task A
Now if i'll add:
task B {
println "Task B"
}
Execute gradle -q B, and I get:
>initialized
>Task B
So it doesn't matter which task I execute, it always get "initialized" first.
You can make every Task who's name is NOT 'initializer' depend on the 'initializer' task. Eg:
task initializer {
doLast { println "initializer" }
}
task task1() {
doLast { println "task1" }
}
// make every other task depend on 'initializer'
// matching() and all() are "live" so any tasks declared after this line will also depend on 'initializer'
tasks.matching { it.name != 'initializer' }.all { Task task ->
task.dependsOn initializer
}
task task2() {
doLast { println "task2" }
}
Or you could add a BuildListener (or use one of the convenience methods eg: Gradle.buildStarted(...))
Seems like you aim execution phase, and you want a task precursing each task or just run as a first task in the execution phase?
If you want a task to always execute in every project before each other task after its being evaluated you can add a closure to he main build.gradle:
allprojects {
afterEvaluate {
for(def task in it.tasks)
if(task != rootProject.tasks.YourTask)
task.dependsOn rootProject.tasks.YourTask
}
}
or
tasks.matching {it != YourTask}.all {it.dependsOn YourTask}
You can also use the Task Execution Graph to define the lifecycle. There are few ways of achieving your goal, depending on your needs and a project structure.
The previously suggested solution with dependsOn works fine, but I don't like about it that it changes and clutters the task dependencies. The first solution coming to my mind is using Gradle Initialization Scripts. They are really cool. But, the usage is a bit tedious: currently there is no way to have a default project-local Gradle init script. You have to either explicitly specify the script(s) on command line, or place them in USER_HOME/GRADLE_HOME.
So another solution (already briefliy mentioned by #lance-java) which can be used to run some initialization, like a init task/script, is "build listeners". Depending on how early/late the initialization code should run, I use one of these two:
gradle.afterProject {
println '=== initialized in afterProject'
}
or
gradle.taskGraph.whenReady {
println '=== initialized in taskGraph.whenReady'
}
Here the docs of the Gradle interface and of BuildListener.
Note that some of the events occur very early, and you probably can't use them because of that, like e.g. beforeProject and buildStarted (explanations here and there).
I'm defining a task in gradle:
task releaseCandidate(type: Exec) {
commandLine 'git', 'checkout', 'develop'
// Increment version code in Manifest
String manifest = new File('AndroidManifest.xml').getText('UTF-8')
Pattern pattern = Pattern.compile('android:versionCode="([0-9]+)"')
Matcher matcher = pattern.matcher(manifest)
matcher.find()
int newVersionCode = Integer.parseInt(matcher.group(1)) + 1
manifest = manifest.replaceAll(
"android:versionCode=\"([0-9]+)\"", "android:versionCode=\"$newVersionCode\""
)
new File('AndroidManifest.xml').write(manifest, 'UTF-8')
commandLine 'git', 'diff'
}
Which I want to execute only when I explicitly call it as gradle releaseCandidate. However, when I run any other task, such as gradle assembleDebug, it also runs task releaseCandidate. I don't want that behaviour to happen. There is no task depending on releaseCandidate or vice-versa.
My project is an Android app, so I am using android gradle plugin.
A common pitfall. Add an action to the task otherwise code will run at configuration phase. Sample task with action:
task sample << {
}
As I see You'd rather need to write a custom task than using Exec type. I suppose it's not valid to define commandLine twice.
EDIT
You can read this post to get the general idea how it all works.
You are mixing Task configuration and groovy code. Everything that is part of the main body of a task definition will be executed in the configuration phase. The task task1 << { code } is a shorthand for
task task1 {
doLast {
code
}
}
commandLine is part of the Exec Task but your other code is not and should be wrapped into a doLast this will execute the commandline first and then execute your additional code. If you need another exec commandLine then you'll need another task.
task releaseCandidate(type: Exec) {
commandLine 'git', 'checkout', 'develop'
doLast {
// Increment version code in Manifest
String manifest = new File('AndroidManifest.xml').getText('UTF-8')
Pattern pattern = Pattern.compile('android:versionCode="([0-9]+)"')
Matcher matcher = pattern.matcher(manifest)
matcher.find()
int newVersionCode = Integer.parseInt(matcher.group(1)) + 1
manifest = manifest.replaceAll(
"android:versionCode=\"([0-9]+)\"", "android:versionCode=\"$newVersionCode\""
)
new File('AndroidManifest.xml').write(manifest, 'UTF-8')
}
}
Just to complete #Opal answer for cases when Exec is really used (for example CommandLine reference) :
task task1 << {
exec {
List<String> arguments = new ArrayList<String>()
//..
commandLine arguments
}
}
I have a task that runs a simple JavaExec.
What I cant seem to get working is the ability to run the JavaExec multiple times while iterating a Filetree object (containing the files) each of while I want to pass into the main JavaExec class one by one. Unfortunately the compiler or code generation tool as it is doesnot accept a directory as an arg so I need to pass the file as an arg per loop. Here's what I have:
task generateClasses(type: JavaExec) {
description = 'Generates Json Classes...'
classpath configurations.all
main = "org.apache.gora.compiler.Compiler"
FileTree tree = fileTree(dir: 'src/main')
tree.include '**/*.json'
tree.each {File file ->
println file
args = [ "src/main/json/$file.name", "$buildDir/generated-src/src/main/java" ]
}
}
compileJava.source generateClasses.outputs.files, sourceSets.main.java
From the above it works and I get all files listed but the JavaExec is called just the once on the very last file read.
How do I address the above? Please help.
How about using the project.javaexec method? See the API Documentation or the
DSL ref.
task generateClasses {
description = 'Generate Json Classes'
fileTree(dir: 'src/main', include:'**/*.json').each { file ->
doLast {
javaexec {
classpath configurations.all
main = 'org.apache.gora.compiler.Compiler'
args = ["src/main/json/$file.name", "$buildDir/generated-src/src/main/java"]
}
}
}
}