Consider using JavaExec when calling DataNucleus

[thc202]

The DataNucleusEnhancer and SchemaTool use System#exit in case of errors which ends up terminating the Gradle daemon, with JavaExec it is created a different process and the exit calls wouldn't affect the Gradle daemon.

Is there are specific reason why they are being called programmatically instead of as command line?

Happy to provide a pull request if this is acceptable.

[thc202]

While not directly related to this issue, have you considered using lazy configuration and also task configuration avoidance?

[rm3l]

Hi @thc202 ,

Thanks for your feedback.

The DataNucleusEnhancer and SchemaTool use System#exit in case of errors which ends up terminating the Gradle daemon, with JavaExec it is created a different process and the exit calls wouldn't affect the Gradle daemon.

Is there are specific reason why they are being called programmatically instead of as command line?

Happy to provide a pull request if this is acceptable.

No specific reason for this, except that I tried to use the methods already available in datanucleus-core.
I think only the schema tool tasks explicitly call the datanucleus-core SchemaTool main method, which indeed might end up calling System#exit:

datanucleus-gradle-plugin/buildSrc/src/main/java/org/rm3l/datanucleus/gradle/tasks/schematool/AbstractSchemaToolTask.java

Line 273 in 9c4bc33

SchemaTool.main(args.toArray(new String[0]));

On the other hand, I don't think the enhancer tasks have a similar issue. They instantiate a new DataNucleusEnhancer object, which contains methods to perform the operations needed, without calling the main method.

datanucleus-gradle-plugin/buildSrc/src/main/java/org/rm3l/datanucleus/gradle/tasks/enhance/AbstractEnhanceTask.java

Lines 284 to 302 in 9c4bc33

	final DataNucleusEnhancer enhancer = new DataNucleusEnhancer(api.name(), null)
	.setVerbose(verbose != null && verbose)
	.setClassLoader(new URLClassLoader(classloaderUrls, Thread.currentThread().getContextClassLoader()))
	.addPersistenceUnit(persistenceUnitName)
	.setDetachListener(detachListener != null && detachListener)
	.setGenerateConstructor(generateConstructor != null && generateConstructor)
	.setGeneratePK(generatePK != null && generatePK)
	.setSystemOut(quiet == null || !quiet);
	if (targetDirectory != null) {
	enhancer.setOutputDirectory(targetDirectory.getAbsolutePath());
	}
	final int result;
	if (this.checkOnly) {
	result = enhancer.validate();
	getProject().getLogger().info("Enhancement validation succeeded for {} class(es)", result);
	} else {
	result = enhancer.enhance();
	projectLogger.info("Enhanced {} class using DataNucleus Enhancer", result);
	}

Anyways, thanks for pointing this out. This definitely makes sense. So if you get a chance to contribute a PR, that would be greatly appreciated.
Thanks.

[rm3l]

While not directly related to this issue, have you considered using lazy configuration and also task configuration avoidance?

Not yet, but any contribution is more than welcome :-)
I've created #331 and #332 to help keep track of these.
Thanks.