Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [Wtp-wst-dev] XPath 2.0 User Doc Plugin
Xerces is the biggest known user at this time, and we are required to be API stable with a point release (i.e. 1.0 to 1.1). Now that isn't to say that we can't change our minds and move the PsychoPath processor from 1.x to 2.0 and still keep the wst.xsl component at 1.1.
We'll just have to make sure we update the New Help for Old Friends V page if we go breaking the API and document the reasons why. 

http://wiki.eclipse.org/New_Help_for_Old_Friends_V
As Xerces is planning a release in December for XML Schema 1.1, if we do break API for the XPath 2.0 processor, we need to do it soon so that we give Xerces enough time to adapt.
Also, please make sure that you have your PDE API Tooling Target Platform setup so that we can catch any minor api breakage, and where we have to add a compatibility setting. I've taken care of the ones for the latest check in.
Latest code has been tagged as well and released to the build, with your changes and my fn:resolve-uri implementation, we are down to 290 failures. 

Dave

Jesper Steen Møller wrote:
I would usually totally agree with your comments about API stability, but are you sure anybody except Mukul are actually using PsychoPath at this point? And are we under obligation to be API stable between milestones? 

-Jesper

On 02/09/2009, at 16.39, David Carver wrote:


I agree that the API needs refactoring.
Some of these classes are currently in an API layer, so if we go this route, we need to make sure that their is an API compatibility layer.
Where the classes/interfaces are internal, we have a bit more freedom on the refactoring.
The method names with the underscores are legacy, and came with the initial code contribution. That type of refactoring especially in classes that are in API packages now, will require a bump of the major version since we are breaking the API contract.
I might suggest for the DefaultDynamicContext, marking it Deprecated, and then creating a new Interface with your suggested refactorings. Please also remember to update the User Interface wiki page with any proposed redesign. 

http://wiki.eclipse.org/PsychoPathXPathProcessor/UserInterface

As well as the Design Document:

http://wiki.eclipse.org/PsychoPathXPathProcessor/Design
For now, the biggest pressing issues is to make sure that we get to 100% compliance on the test suite. These refactorings and re-designs can be done after that work is completed. 

Dave


Jesper Steen Møller wrote:

Nice done, tying the Wiki part to the distributed docs.
However, it reminds me that I would really like to change some parts of API. The overall changes I'll propose are these: 

A) ResultSequence
A1) Turn ResultSequence into an interface, and make it immutable, i.e. no more clear(), add(), and concat(). A2) Provide simple implementations for the the empty sequence, singleton sequences, and fixed length sequences. A3) Introduce a class ResultBuilder which has the methods required for building a ResultSequence. A4) Introduce a ResultSequence.modify() which returns a ResultBuilder and provide copy-on-use semantics, w.r.t. the underlying List. 

B) DefaultDynamicContext
We are mixing the elements of the dynamic and static contexts - I can't see how that's good for the adopters. Example when using XPath for Schema 1.1: - On loading the grammar, a static context is produced, one for each XPath expression (since e.g. namespace prefixes are going to be different). It may make sende for Xerces to provide their own StaticContext, so the shared parts like function libraries are shared, but e.g. the namespaces can differ. Then, for each assertion in a parse/validation run, a dynamic context is in use, and has to "adhere to" the static context of that expression. 
The proposed changes are then:
- Change DefaultDynamicContext to delegate to the DefaultStaticContext instead of extending it. Or just provide a "getStaticContext" getter instead.
C) Functions coupled to the static and even the dynamic contexts. This prevents sharing of function libraries. Rather, we should pass the dynamic context in when evaluating (especiallly if we add the getStaticContext to the DynamicContext interface?)
D) Cosmetic: why are the function names not camelCase, but underscoreRidden ? shouldn't we at least mimic the style of the DOM standards, e.g. http://www.w3.org/TR/DOM-Level-3-XPath/java-binding.html which is obviously for XPath 1.0, but still worth looking at?
Comments, anyone? Unless someone objects, I'll open a bug for each of A-D and submit suggested patches and explanations. 

-Jesper
On Wed, Sep 2, 2009 at 3:07 AM, David Carver <d_a_carver@xxxxxxxxx <mailto:d_a_carver@xxxxxxxxx>> wrote:
With the complete of bug 286070 [1], the PsychoPath user documentation 
   and SDK documentation plugins have been created.   The user
   documentation is sourced from the Eclipse Wiki [2] and then the
appropriate help content is generated using mylyn-wikitext and docbook 
   xsl stylesheets.

   [1] https://bugs.eclipse.org/bugs/show_bug.cgi?id=286070

   [2] http://wiki.eclipse.org/PsychoPathXPathProcessor/UserManual



   _______________________________________________
   wtp-wst-dev mailing list
   wtp-wst-dev@xxxxxxxxxxx <mailto:wtp-wst-dev@xxxxxxxxxxx>
   https://dev.eclipse.org/mailman/listinfo/wtp-wst-dev
------------------------------------------------------------------------ 

_______________________________________________
wtp-wst-dev mailing list
wtp-wst-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/wtp-wst-dev




_______________________________________________
wtp-wst-dev mailing list
wtp-wst-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/wtp-wst-dev


_______________________________________________
wtp-wst-dev mailing list
wtp-wst-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/wtp-wst-dev

Back to the top