[ https://issues.jboss.org/browse/SHRINKWRAP-523?page=com.atlassian.jira.pl... ] Ian Brandt updated SHRINKWRAP-523: ---------------------------------- Description: AWS Lambda supports deployments of Java functions in [a simple ZIP format|https://docs.aws.amazon.com/lambda/latest/dg/create-deployment-pkg...]. As with standard JARs, the class files, resources, manifest resources, and service provider configurations are all included from the root of the archive. Unlike standard JARs, dependent JARs are included under a 'lib/' folder. Currently, to create ad-hoc Lambda deployments for testing I'm starting with a {{JavaArchive}}. I add the standard JAR assets, and then use the [{{Archive.add(Archive, ArchivePath, StreamExporter)}}|https://repository.jboss.org/nexus/content/repositories...] API (as used by [{{ContainerBase.addAsLibraries(Archive...)}}|https://repository.jboss.org/nexus/content/repositories/unzip/org/jboss/shrinkwrap/shrinkwrap-impl-base/1.2.6/shrinkwrap-impl-base-1.2.6-javadoc.jar-unzip/index.html] to add the dependencies. Finally, I export it to a temp file with a '.zip' extension for upload: {code:java} final JavaArchive lambdaZip = ShrinkWrap.create(JavaArchive.class); lambdaZip.addClasses(handlerClass); final ArchivePath archiveLibraryPath = ArchivePaths.create("/lib"); Maven.resolver() .loadPomFromFile("pom.xml") .importCompileAndRuntimeDependencies() .resolve() .withTransitivity() .asList(JavaArchive.class) .forEach(javaArchive -> lambdaZip.add(javaArchive, archiveLibraryPath, ZipExporter.class)); lambdaZip.as(ZipExporter.class).exportTo(tempZipFile, overwrite); {code} It would be nice to be able to use the [{{LibraryContainer}}|https://repository.jboss.org/nexus/content/repositories/unzip/org/jboss/shrinkwrap/shrinkwrap-api/1.2.6/shrinkwrap-api-1.2.6-javadoc.jar-unzip/index.html] API directly instead. I can think of a few ways to accomplish this (and there are likely others): # Make {{JavaArchive}} a {{LibraryContainer}}. The library path would be defaulted to 'lib/'. #* Easy to do since {{JavaArchiveImpl}} extends {{ContainerBase}}, which already has the all the necessary implementation. #* The API would now extend the [JAR File Specification|https://docs.oracle.com/javase/8/docs/technotes/guides/jar/...], so there's the question of utility vs. conformance. #* There's also the issue of the default archive name having a '.jar' extension instead of '.zip' (per the service descriptor properties file for {{JavaArchive}}). # Add {{LibraryContainer}} and {{ServiceProviderContainer}} to the list of interfaces extended by {{GenericArchive}}. #* Easy again with the underlying implementation being {{ContainerBase}}. #* There's no presumption of a spec with {{GenericArchive}}, so presumably there's no harm in the additional API. Perhaps that makes defaulting the library path to 'lib/' more assuming, though. #* I was surprised to find the default extension for {{GenericArchive}} is '.jar', which implies the JAR spec. I would have guessed '.zip'. I'd propose changing it as part of this approach, but that would be a breaking change for anyone who's come to depend on it being '.jar'. #* Another issue here is if you further say the {{WebContainer}}, {{EnterpriseContainer}}, and {{ResourceAdapterContainer}} API might as well be added too. Their respective implementation classes extend {{ContainerBase}}, creating a multiple inheritance situation on the implementation side. There's also conflict between the standard paths, e.g. 'lib/' vs. 'WEB-INF/lib/'. # Add a new {{LambdaArchive}} interface and implementation to ShrinkWrap proper. #* It would be the equivalent of approach #1, but with a default extension of '.zip'. #* AWS Lambda seems here to stay. #* I would think it better to take this as an opportunity to make the base API more generic, rather than to specialize with a new subtype. # Create the same {{LambdaArchive}}, but in a ShrinkWrap extension project. I'm happy to create a pull request given guidance on the preferred approach. For myself, I'd lean towards #2 if changing the default extension is deemed acceptable. was: AWS Lambda supports deployments of Java functions in [a simple ZIP format|https://docs.aws.amazon.com/lambda/latest/dg/create-deployment-pkg...]. As with standard JARs, the class files, resources, manifest resources, and service provider configurations are all included from the root of the archive. Unlike standard JARs, dependent JARs are included under a 'lib/' folder. Currently, to create ad-hoc Lambda deployments for testing I'm starting with a {{JavaArchive}}. I add the standard JAR assets, and then use the [{{Archive.add(Archive, ArchivePath, StreamExporter)}}|https://repository.jboss.org/nexus/content/repositories...] API (as used by [{{ContainerBase.addAsLibraries(Archive...)}}|https://repository.jboss.org/nexus/content/repositories/unzip/org/jboss/shrinkwrap/shrinkwrap-impl-base/1.2.6/shrinkwrap-impl-base-1.2.6-javadoc.jar-unzip/index.html] to add the dependencies. Finally, I export it to a temp file with a '.zip' extension for upload: {code:java} final JavaArchive lambdaZip = ShrinkWrap.create(JavaArchive.class); lambdaZip.addClasses(handlerClass); final ArchivePath archiveLibraryPath = ArchivePaths.create("/lib"); Maven.resolver() .loadPomFromFile("pom.xml") .importCompileAndRuntimeDependencies() .resolve() .withTransitivity() .asList(JavaArchive.class) .forEach(javaArchive -> lambdaZip.add(javaArchive, archiveLibraryPath, ZipExporter.class)); lambdaZip.as(ZipExporter.class).exportTo(tempZipFile, overwrite); {code} It would be nice to be able to use the [{{LibraryContainer}}|https://repository.jboss.org/nexus/content/repositories/unzip/org/jboss/shrinkwrap/shrinkwrap-api/1.2.6/shrinkwrap-api-1.2.6-javadoc.jar-unzip/index.html] API directly instead. I can think of a few ways to accomplish this (and there are likely others): # Make {{JavaArchive}} a {{LibraryContainer}}. The library path would be defaulted to 'lib/'. #* Easy to do since {{JavaArchiveImpl}} extends {{ContainerBase}}, which already has the all the necessary implementation. #* The API would now extend the [JAR File Specification|https://docs.oracle.com/javase/8/docs/technotes/guides/jar/...], so there's the question of utility vs. conformance. #* There's also the issue that the default archive name will have a '.jar' extension instead of '.zip', per the service descriptor properties file for {{JavaArchive}}. # Add {{LibraryContainer}} and {{ServiceProviderContainer}} to the list of interfaces extended by {{GenericArchive}}. #* Easy again with the underlying implementation being {{ContainerBase}}. #* There's no presumption of a spec with {{GenericArchive}}, so presumably there's no harm in the additional API, though that also makes defaulting the library path to 'lib/' a bit more assuming. #* I was a bit surprised to find the default extension for {{GenericArchive}} is '.jar' as well, as that implies the JAR spec. I would have guessed '.zip'. I'd propose changing that as part of this approach, but that would be a breaking change for anyone that's come to depend on it being '.jar'. #* Another issue here is if you further say the {{WebContainer}}, {{EnterpriseContainer}}, and {{ResourceAdapterContainer}} API might as well be added too. Their respective implementation classes extend {{ContainerBase}}, creating a multiple inheritance situation on the implementation side. There's also conflict between the standard paths, e.g. 'lib/' vs. 'WEB-INF/lib/'. # Add a new {{LambdaArchive}} interface and implementation to ShrinkWrap proper. #* It would be the equivalent of approach #1, but with a default extension of '.zip'. #* On one hand, AWS Lambda seems here to stay. #* I would think it better to take this as an opportunity to make the base API more generic, rather than to specialize with a new subtype. # Create the same {{LambdaArchive}}, but in a ShrinkWrap extension project. I'm happy to create a pull request given guidance on the preferred approach. For myself I'd lean towards #2 if changing the default extension is deemed acceptable. -- This message was sent by Atlassian JIRA (v7.5.0#75005)