Hello Qi Wang I think this is a really good idea! I am one of the developers of the XPath2 engine at the Eclipse WTP project (in fact I don't know much Xalan-J, actually I just contributed two Xalan-J bug fixes which were never picked up)
For the release this summer (Indigo), a new API has been introduced which allows pluggable type models, and allows much better access to the XPath2 engine without using internal packages. Also, the library uses a better structure for separating the parsing and execution of the expression. Also, we will now be JRE 1.4 compatible (sigh). Lots of improvements you can use there. However, "PsychoPath" (I hate that name) is still bound to using a Level 3 DOM, and no, we do not have a streamy front end. The internal tree implementation for Xalan would thus have to be wrapped for use with the Eclipse XPath2 engine. But for type-sensitive template handling, this might not be sufficient anyway. So you will face quite a number of challenges! Also, I hope you can do all this without having to fork the XPath2 engine, since it is also used elsewhere, but we might be able to "semi fork" it and work your changes back into the mainline for the next major release. As for how to work with the Xalan codebase, I cannot provide guidance on your overall strategy, I think the Xalan-J people can guide you much better there. Kind regards, Jesper Steen Møller On 17/04/2011, at 16.43, Qi Wang wrote: > Hi all, > > All Xalan developer and users, I wan to do something to push XSLT 2.0 > implemention job for Xalan, i know this topic was discussed in mail list for > many many times in recent years, so many guys are expecting it come true. But > due to low activity in Apache Xalan community, no committer to push this > thing, it is really very regretful. > > These days, i submited "Implement XSLT 2.0 and XPath 2.0 for Xalan" project > proposal as an GSoC[1] project. Now, i need your feedbacks to improve the > proposal, all the advises, suggestions are welcome :-) > > I am not a Xalan committer(alrough i dream to be), but if there is a chance, > i want to try my best to so something for XSLT 2.0 implemention job for > Xalan, many i can not finish all the job in the summer, but i will make it is > available, once wheels start turn, i thing there may be some other guys can > work together with me to finish this big project. Hoping all the comunity can > work together for the better future of Xalan. > > [1] http://socghop.appspot.com > > =========Proposal details of "Implement XSLT 2.0 and XPath 2.0 for > Xalan"================ > > Apache Xalan-Java is a powful XSLT processor for XML transforming based on > W3C recommendation XSL Transformations (XSLT) Version 1.0 and XML Path > Language (XPath) Version 1.0, XSL Transformations (XSLT) Version 2.0 > specification was published on 23 January 2007,this new version give us more > powful functions, new dat model and serialization etc. > > > > In my daily work, i use XSLT and XPath to parse XML and change XML document > to another format, i did this with Xalan.As we know,Xalan support XSLT > version 1.0 and XPath version 1.0 only.But my XML parse job is so complex, > and usually, i must use some XPath v2.0 functions to finish my parse job. So, > i encountered trouble. > > > > Then I Google keywords with Xalan and XSLT 2.0,i found many developers hope > that Xalan based XSLT 2.0 could become reality besides me. At the same time, > i discuss this topic with open source developers in mail list, so many guys > are interested in implementing complete XSTL 2.0 specification for Xalan. We > discuss this project deeply, and i got many good ideas and advises from the > discussion. And also they told me Xalan lost its most active contributors a > few years ago due to time demands from Real Jobs, and development slowed down > as a result,but they love to see Xalan tackle it.Due to XSLT 2.0 and XPath > 2.0 are large specifications that I imagine could take the developers years > to implement correctly,anyway it need some guy get the ball rolling, i want > to be this guy. > > > > Compared with XSLT 1.0, XSLT 2.0 and XPath 2.0 have some important new > characteristic. > > > > New features in XPath 2.0: > > 1. Sequences, in XPath2.0 everything is sequences > 2. For expression,in XPath 2.0,we can use for expression to iterator every > item in the sequence, compute expression value for each item, at last return > the sequence result by connect all the expression values. > 3. Condition expression, XPath 2.0's condition expression can be used to > compute different value on the basis of the condition's true or false value. > 4. Limited expression,limited expression in XPath 2.0 can be used to judge > that whether every item in the sequence satisfy appointed condition, its > value is always true of false. There are two kinds of limited expression: > some expression and every expression. > 5. Datatype,XPath 2.0's datetype is based on XML Schema, it supports all > the basic built-in datatype,such as xs:integer,xs:string and xs:date ect. > 6. Date and time, XSLT 1.0 has not date and time datatype, it must use > string to represent date and time. Due to XPath 2.0 system is based on XML > schema, so there is date and time data type in XPath 2.0. > 7. More functions support,XSLT 2.0 specification has more powful functions > definition.Functions in XPath 2.0 is defined in a special recommedation > "XQuery 1.0 and XPath 2.0 Functions and Operations", functions in this > recommendation specification belong to namespace > "http://www.w3.org/2005/xpath-functions";, this namespace is bound with prefix > "fn". > > It includes these serveral types: > > * Constructor Functions > * Functions and Operators on Numerics > * Functions on Strings > * Functions on anyURI > * Functions and Operators on Boolean Values > * Functions and Operators on Durations, Dates and Times > * Functions Related to QNames > * Operators on base64Binary and hexBinary > * Operators on NOTATION > * Functions and Operators on Nodes > * Functions and Operators on Sequences > * Context Functions > > There are also many new features in XSLT 2.0: > > 1. Group, basic group syntax,group sort,group-adjacent and > group-starting-with etc. > 2. Connotive document node, or we can call it temporary tree > 3. Element result-document, we can use to generate multi output file in > XPath 2.0 > 4. Improvement of element value-of > 5. Char mapping,XSLT 2.0 supply more flexible solution for us to handler > specific characters in XML such as '<' and '&' > 6. Custom stylesheet function,in XSLT 2.0, we can use to create our own > functions in stylesheet > 7. Other new features. > > In order to implement all the XSLT 2.0 specification,it requires some > moderately basic data structure changes to track Schema types and to handle > Sequences which aren't simply nodesets,and also,i must implements all the new > functions defined in XSLT 2.0,fortunately,there is a open source XPath 2.0 > processor (http://wiki.eclipse.org/PsychoPathXPathProcessor) which is at > Eclipse.It is already being used by Xerces-J for XSD 1.1 Assertion > support.Why re-invent the wheel if we can leverage something that already > exists ? PsychoPath already is XML Schema Aware, leverages > Xerces-J(http://xerces.apache.org/xerces2-j) and passes about 99.9% of the > Xpath 2.0 test suite. It is currently undergoing testing against the the > official W3C Test Suite. > > > > So, in order to implement all the XSLT 2.0 specification for Xalan, there are > two steps: > > 1. Merge Xalan codes with PsychoPath codes,this will help us a lot,once > this job is done, Xalan has implemented all the XPath 2.0 features. There are > some issues to resolve during this merging process,for example: > * PsychoPath is not "streamy" yet, nor is Xalan, it is a big > challenge > * In PsychoPath,type informations is not part of current > interfaces,i will have to extend or replace the DTM interfaces to carry the > additional data > * XSLTC fit in problem. XSLTC runs against the same data model > interfaces (DTM), and In Theory implementing compilation should just be a > matter of generating code equivalent to what the interpreter is doing. In > practice, it's a lot of detail work. > 2. Implement all the other XSLT 2.0 features which were described before > for Xalan. > > In my plan, i will hold the whole XSLT 2.0 implemention project as a long > term open source project under Apache, do it step by step, may be as a GSoC > project,i can only implement a subset of this capability i was thinking of > working on,at lease, i will finish the following tasks: > > 1. Merge Xalan codes with PsychoPath codes, i will do this first, but this > is not a simple task,may be it will cost me one month or even more to merge > it correctly. > 2. Resolve merge problems between Xalan and PsychoPath > 3. Implement XSLT 2.0 Group feature for Xalan > 4. Implement XSLT 2.0 temporary tree feature for Xalan > 5. Implement Element result-document feature for Xalan > > If everything goes well and i have enough time,i will do my best to do more > job. After the summer, i will continue with this project,try to submit my > codes to Xalan as proposed changes and try to become one of Apache > committers,this is my dream.And i know this is a huge project,but where there > is a will there is a success, I love open source, i will do my best to > improve it through my attemp and ability. > > > Regards > > Victor Wang
