Java 8 modules, Jigsaw and OSGi - Neil Bartlett

Post on 13-May-2015

12765 Views

Category:

Technology

0 Downloads

Preview:

Click to see full reader

DESCRIPTION

Presentation by Neil Bartlett (Paremus) from OSGi DevCon 2012 BOF (22 March, 2012) Video recording of the presentation is available at http://youtu.be/QJaSW5dW5c0 Abstract: The idea of integrating a module system into the core Java runtime has been proposed and discussed for many years, and in Java SE 8 it may finally be happening. How will this affect OSGi developers and users? This talk will describe the main differences between OSGi and Jigsaw, the prototype OpenJDK module system. Pros and cons of each in different environments will be discussed. Finally, opportunities and challenges for interoperability: from the perspective of both application developers (who may need to integrate modules from both kinds) and from library module developers (who may need to target both module systems).

Transcript

JAVA 8 MODULES, JIGSAW AND OSGi

Neil Bartlett with Tim Ellison

MOTIVATION

WHY MODULARISE THE JDK?

• The JRE is monolithic

• Download time and start-up time are directly affected by number of types available at runtime

• Start-up time includes a linear search through the class path to find system and application code

• Oracle 1.7 Windows boot classpath nearly 20k classes

• rt.jar index alone is approx 1Mb

OSGi

• If only Java had a module system that could help with that!

• As we all know, OSGi will not be used for JDK modularity

JIGSAW MOTIVATION

• JRE libraries evolved organically and haphazardly over 13+ years

• Many cyclic dependencies

• Many weird/unexpected dependencies

• E.g.: java.util is an incoherent mess

• Splitting packages is unavoidable

Good news, everyone!

OSGi supports split packages, using Require-

Bundle.

SINGLE CLASSLOADER

• Yeah but there’s the single classloader assumption...

• Many parts of the JRE assume a single boot classloader

• Package-private (“default”) accessibility requires whole packages in single classloader

• Need to split packages across modules but not across classloaders

• Break the one-to-one mapping of modules to classloaders

OSGi supports shared classloaders, using

Fragments.

DEPENDENCY DIRECTION

• Yeah, but the dependency goes in the “wrong” direction (fragment depends on host).

• Jigsaw wants the “host” to depend on the “fragment”: host shouldn’t resolve if fragment unavailable.

OSGi R4.3 supports that too using generic

requirements/capabilities.

REUSE

• Yeah but the fragment still depends on the host and cannot be used by other hosts.

...

Pretty sure we could have supported this, if only you

had talked to us.

REFACTORING

• Usual and best solution: move classes to the correct, most logical place

• Obviously impossible in the JRE.

• 9 million Java developers (Sun estimate, 2009) and billions of apps depend on this library.

JIGSAW

MODULES

• Module declared in module-info.java (module-info.class)

• All new keywords are “scoped”

module B @ 1.0 { ...}

REQUIRES

• Modules require other modules by name (and optionally, version)

module B @ 1.0 { require A @ [2.0,3.0);}

LOCAL

• Requirements can be marked “local”

• Target module is loaded in same classloader

module B @ 1.0 { require local A @ [2.0,3.0);}

EXPORTS

• Modules list their exports, at package and type level

• May include re-exported contents of required modules

module B @ 1.0 { require A @ [2.0,3.0); export org.foo.ClassFoo; export org.bar.*;}

FRIENDS

• Modules can control which other modules require them

• Compare with “friend” classes in C++

• N.B. permit clause is not versioned.

module A @ 2.0 { permit B;}

PROVIDES

• Modules can logically “provide” other modules names

• Compare with “virtual packages” in Debian

• Supports substitution, but not refactoring (splits or joins)

module com.ibm.stax @ 1.0 { provide jdk.stax @ 2.0;}

ENTRY POINT

• Modules can have a single entry-point class

• Compare with Main-Class header in Jar manifests.

module A @ 2.0 { permit B; class org.foo.Main;}

VERSIONS

• Modules will be versioned

• Requirements use exact version or a range

• No version semantics beyond ordering

COMPARISONS

LIFECYCLE

• Resolver solves dependencies in the face of multiple valid alternatives

• OSGi resolver finds best fit for currently installed bundles at runtime

• Jigsaw resolves during build and installation

• Jigsaw has no dynamics, no module lifecycle

METADATA

• Tools required to inspect module-info.class

• OSGi uses kind-of readable MANIFEST.MF

COMPARISON

• Whole-module dependencies and split packages are the biggest differences

• Jigsaw will suffer all the same problems seen in OSGi with Require-Bundle (e.g. Eclipse Core Runtime refactoring mess)

• While Jigsaw will technically achieve JDK modularity it will increase developer maintenance burden

INTEROP

INTEROP

• Both are here to stay... try to get the best of both worlds!

• OSGi is established and will continue to be used widely

• Jigsaw is underway and a key component of Java 8

• It need not be a zero-sum game

JAVA 8 REQUIREMENTS

• “It must be demonstrated by prototype to be feasible to modify an OSGi micro-kernel such that OSGi bundles running in that kernel can depend upon Java modules. The kernel must be able to load Java modules directly and resolve them using its own resolver, except for core system modules. Core system modules can only be loaded using the module system’s reification API.”

• http://openjdk.java.net/projects/jigsaw/doc/draft-java-module-system-requirements-12

PROJECT PENROSE

• OpenJDK-hosted project to work on Jigsaw/OSGi interop

• Project lead is Tim Ellison

LEVELS

0

1

2

3+

Tolerate

Understand

Exploit

Cooperate

LEVEL 0: TOLERATE

• Ensure that OSGi frameworks continue to run unmodified on Jigsaw-enabled runtime.

• Creating modules/bundles that have both Jigsaw and OSGi metadata on the same JAR.

LEVEL 1: UNDERSTAND

• Teach OSGi to read Jigsaw module info

• Mapping Jigsaw metadata into OSGi concepts e.g. requires = Require-Bundle, exports = Export-Package.

• Resolve Jigsaw modules using the OSGi resolver.

LEVEL 2: EXPLOIT

• OSGi implementation exploits Jigsaw modularity

• E.g. use Jigsaw publication repositories.

LEVEL 3+: COOPERATE

• A blend of OSGi and Jigsaw cross-delegation on module phases.

ACHIEVEMENT 1

• Passed the OSGi tests on a Jigsaw-enabled runtime

• Equinox 3.7, the OSGi Reference Implementation.

• OSGi R4.3 Compliance Tests

ACHIEVEMENT 2

• Run a Java application as either OSGi or Jigsaw modules

• Took a Java 2D demo

• Broke into multiple functional units, run demo as either OSGi bundles or Jigsaw modules

GOAL 1

• Map Jigsaw module metadata to OSGi equivalents

• Discussion: how to interpret Jigsaw directives

GOAL 2

• Modify OSGi (Equinox) to use Jigsaw reification APIs

• Load a 3rd-party module from the Jigsaw repository

• Resolve it using the OSGi resolver

TIMELINE

• Java 8: summer 2013? Code freeze early 2013.

• Jigsaw seems to be basically “done”, with little recent activity

• No JSR yet, or perhaps ever.

• If we need changes in Jigsaw, we need to push them soon.

GET INVOLVED

• Penrose home & mailing list:

• http://openjdk.java.net/projects/penrose/

• Hg repository:

• http://hg.openjdk.java.net/penrose/jigsaw/

top related