To install JSR-331, simply download, unzip, and import the Eclipse project “org.jcp.jsr331.tck” “to your Eclipse workspace. JSR-331 v.1.0.0 comes with 3 working implementations based on open source CP solvers: Choco, Constrainer, and JaCoP. One more implementation is on the way. All included software comes with sources and available under the terms of open source licenses. You may look at the latest JSR-331 User Manual. There is a Google support group for all your technical questions. Your comments and suggestions are very welcome.

A paper that describes this approach in details has been published at RuleML 2011 – America, LNCS 7018, pp. 208-221, 2011, Springer-Verlag Berlin Heidelberg 2011.

1)      Developing JSR-331, a Java CP API.  JSR-331 has successfully passed the Public Review Vote – see http://jcp.org/en/jsr/results?id=5118.  Now it has 3 working implementations based on open source CP solvers (Choco, Constrainer, and JaCoP) and many examples that explain how to use the standard to represent and solve different CSPs. These JSR-331 implementations currently support integer, boolean, and set constrained variables, and many popular binary and global constraints. The implementation of real constrained variables is under way.  The latest User Manual can be found at http://4c110.ucc.ie/cpstandards/files/JSR331.UserManual.v072.pdf. To download and try JSR-331 with all examples and sources please send a request to jacobfeldman@openrules.com. The next step is to produce the JSR-331 Final Draft that will include an enhanced TCK (Technology Compatibility Kit) and more practical examples. We expect to publish the Final Draft in August.

2)      Developing a common CP XML format. A small group of ACP members (Julien Fischer, Peter Stuckey, Radoslaw Szymanek, Guido Tack) have modified FlatZinc XML, the XML version of FlatZinc to create a draft standard for CP solver interchange. When this standard is finalized, an ability to save/load constraint satisfaction problems in the CP XML format will be added to the common level of the JSR-331 implementation. Read more about CP XML in Peter Stuckey’s post at the CP Standardization Discussion Forum and provide your comments and suggestions.

- FlightCrews - configuring flight crews under different constraints such as cardinality, knowledge of foreign languages, and more.

- Manners - this is a set of examples that deal with a notorious benchmark “Miss Manners” frequently used to compare performance of different rule engines.  The problem is to find an acceptable seating arrangement for all guests at a dinner party.  An acceptable arrangement should match people with the same hobbies, and to seat everyone next to a member of the opposite sex.

First we demonstrate how to represent and solve this problem with well-defined data for parties of 16, 64, and 128 guests.  Then we demonstrate how to find an “acceptable” solution even when not all constraints can be satisfied – a situation when rule engines or a simple Java implementation usually stop short.  Our CP-based solution explains how to minimize the total constraint violation. Read more about this problem and comparison of different approaches at my presentation at RulesFest-2009 “Rule Violations and Over-Constrained Problems“.

The implementation of Miss Manners with set variables required an Element constraint to be defined on an array of sets with a constrained set variable served as an index. To implement this constraint on a common level (in a solver independent way), I needed an ability to add new custom constraints. These capabilities currently are not included in the standard JSR-331 interface, but it is a part of the JSR-331 implementation package called “javax.constraints.extra”. This package allows a user to define custom propagators and associate them to certain events occurred on constrained variables.  To implement “ConstraintElementAtOnSets” I used this package and Constrainer as an underlying CP solver.

You may download the latest JSR-331 and try to use constrained set variables for your own problems – simply send me a request to jacobfeldman@openrules.com and I will provide you with an access to all jars and sources.

As you may notice, this interface has its own, unique naming convention like the most of CP solvers do. However, conceptually there should be no problem to build a JSR-331 implementation using a new Google CP solver.  At the same time, the fact that the actual solver’s implementation is done in C++, may cause some configuration issues.  Everything is open-sourced, and if somebody wants to do it now, please contact me at jacobfeldman@openrules.com and I would provide you with an access to the JSR-331 SVN repository. You also may count on my personal support as well. If this development goes well, I hope soon we will have one more (fourth!) JSR-331 implementation.

Rule Solver^™ includes JSR-331 based templates that allow business analysts to define their CSPs directly in Excel. For example, here is a rules table that defines a Sudoku problem:

An Excel user with simple copy-pastes defines rows, columns, and blocks. The rules table “postSudokuConstraints”with one AlDiff constraint does the entire work.

While JSR-331 does not include (yet!) a scheduling component, OpenRules added a simple package “javax.constraints.scheduler” that extends JSR-331 to handle scheduling and resource allocation problems.  It allows a user to express a scheduling problem in terms of activities and resources.  Activities may have unknown start, duration and end. Activities may require or produce different resources. Resources may have different types (e.g. recoverable like humans or consumable like money) and their limited capacities may vary over time. The package includes major temporal and capacity constraints specified in accordance with the JSR-331 requirements.

Rule Solver^™ includes OpenRules-based templates for scheduling problems. For example, let’s consider 5 different orders (activities) that fire batches of bricks in the oven (a resource with limited capacities). Each order requires certain oven’s capacity as described in the following picture:

With Rules Solver^™ a user can define the problem in the following Excel rules tables:

To solve the problem a user specifies the following main method (also in Excel):

Here are the results:

Solution:
A[0 -- 1 --> 1)    requires Oven[2]
B[3 -- 4 --> 7)    requires Oven[1]
C[5 -- 4 --> 9)    requires Oven[1]
D[5 -- 2 --> 7)    requires Oven[1]
E[11 -- 4 --> 15) requires Oven[2]

You may read the proper user manual with a lot of examples at http://openrules.com/pdf/RulesSolver.UserManual.pdf and download and try the software from http://openrules.com/insideTrack.htm#RuleSolver.

Hopefully, other BRMS vendors will follow OpenRules lead and will integrate JSR-331 in their Business Rules products.

However, from a pure Java perspective I think there were more deserving nominees, see for example JSR-299, Contexts and Dependency Injection for the Java EE platform.  Here what JCP says about it: “Critics have foreshadowed the death of the Java language, claiming that there is little more than can be done to evolve it and therefore needs to be replaced. JSR 299 silenced these viewpoints.”

It is interesting that we did nothing to nominate or somehow promote our JSR – I simply had no idea about this competition. Of course, I understand this award is more related to the recognition of the CP technology itself. But it also shows that CP is still not within main stream of real-world application development where it certainly deserves to be. Hopefully this award and the standard API will make CP more visible and accessible to Java developers.  So, now it becomes even more important to expedite creation of a JSR-331 User Manual and making initial implementations available for free downloads to all application developers.

See http://jcp.org/en/press/news/awards/2010award_nominees

JCP Awards:  http://blogs.sun.com/jcp/
JSR-331 Specification (Early Draft): http://4c110.ucc.ie/cpstandards/files/JSR331Specification.v061.pdf
http://www.cpstandards.org
http://cp2010.cs.st-andrews.ac.uk/events.php