Hi! I am working on it, but progress is slow. There are many other things in my pipeline. I hope I have something by end of the week.
Best ... Michael -----Ursprüngliche Nachricht----- Von: Evert Hoff [mailto:[EMAIL PROTECTED]] Gesendet: Dienstag, 28. Januar 2003 10:09 An: [EMAIL PROTECTED] Betreff: Re: Schema annotations Hi, Is anyone busy working on implementing the annotations at the moment? If so, I would appreciate a rough guess as to when it might be available in CVS. Thanks in advance, Evert On Thu, 2003-01-16 at 23:53, [EMAIL PROTECTED] wrote: > Many thanks to Elena for the detailed message. Here I want to provide my > view on one of the details. > > As Elena mentioned, currently the schema document is first parsed into a > DOM representation (a DTM in our case) using a DOM parser, then the DTM > tree is processed by various traversers (for various schema components): > > schema document --> (DOM parser) --> DTM --> (traversers) --> schema > components > > If we following the same pattern, annotations would be process like the > following: > > <annotation> elements --> (DOM parser) --> DTM element nodes --> > (annotation traverser) --> XSAnnotation implementation > > Also as Elena mentioned, it might be a good idea to store annotations as > Strings. (Storing the DOM/DTM would require more space, and might require > storing the whole tree; while storing the string saves memory, but takes > longer to retrieve information from it. There is always a trade-off.) So > it'd be like: > > <annotation> elements --> DTM nodes --> Strings > > Wouldn't it be more efficient if, in the DOM parser, we directly serialize > the parser events to a String, to save the time building (and GC'ing) the > DTM nodes? > > <annotation> elements --> Strings > > Of course, to go this route, more things need to be taken care of. In the > current XSDAbstractTraverser#traverseAnnotationDecl() method, some checking > needs to be performed: > - The attributes on <annotation> need to be valid; > - The names of child-elements need to be one of "appinfo" and > "documentation"; > - The attributes on the above 2 elements need to be valid. > But they can (easily) be checked in the DOM parser, instead of in a > traverser. > > Advantage: > - Performs better. > - Solves problems regarding storing texts, comments and PI's. > > Disadvantage: > - Might be (a little bit) harder to implement. > - Need some code duplication (to check the validity of > sub-elements/attributes mentioned above). > > So IMO, as a start, we can still use DTM (ignoring texts / comments / > PI's). But in the long run, we might need to consider skipping it, and > generate the strings directly. > > Thanks, > Sandy Gao > Software Developer, IBM Canada > (1-905) 413-3255 > [EMAIL PROTECTED] > > > > > Elena > Litani/Toronto/IB To: [EMAIL PROTECTED] > M@IBMCA cc: > Subject: Re: Schema annotations > 01/16/2003 03:49 > PM > Please respond to > xerces-j-dev > > > > > > Hi Michael, > > "Dr. Michael Treichel" wrote: > > We would like help implementing access to schema annotations. How do we > get > > into business? > > In general, we thought that annotation should be stored as an XML string > in each XML schema component (i.e > org.apache.xerces.impl.xs.XSElementDecl), following the XML syntax > defined in the XML Schema specification [1]. > The writeAnnotation() will invoke SAX or DOM parser that will parse this > annotation string, issuing either SAX events or producing a DOM tree. > > Here some additional information: > 1. Traversing of annotation from the schema is done in the > org.apache.xerces.impl.xs.traversers XSDAbstractTraverser > traverseAnnotationDecl. The schemas are traversed into dom-like > structure (see implementation in org.apache.xerces.impl.xs.opti) but > traversal is implementation independent (see DOMUtil class). You will > need to walk the DTM tree (using DOMUtil) and create annotation string. > At this time you will also need to save all the namespace declarations > in scope (since you need those declarations to parse the string, etc...) > -- XSDocumentInfo has a pointer to the namespaces in scope. You need to > get all prefix declarations and add those as namespace declarations > (xmlns:prefix="http://...";) to the xs:annotation. > Btw, for serialization of the annotations you might look at > XMLSerializer (in org.apache.xml.serialize) -- there is a serializeNode > method, however I am not sure if it will work since it assumes DOM > implementation. > > 2. The DTM (org.apache.xerces.impl.xs.opti) implementation currently > does not create any text nodes, meaning that any text in <documentation> > or <appinfo> won't be available... To fix it you need to modify the > SchemaDOMParser (org.apache.xerces.impl.xs.opti) parser, to create > string when in context of annotation element. This could be done by > having a global StringBuffer, appending the data to it in characters() > method and retrieving / resetting data at start/endElement calls > (assuming there are no comments, PIs, etc in the annotation context). > The string could be later stored either as a Text node or just a string > field on the DefaultElement node (org.apache.xerces.impl.xs.opti). > > 2. In the writeAnnotation you will need to get a parser. For performance > reasons you might need to implement some kind of synchronized parser > pool to avoid creating a new parser for each writeAnnotation method > invocation. As far as I know, Xerces grammar implementation is > thread/safe (grammars could be added to different grammar pools used by > parsers in the different threads). So you must synchronize access to > your parser. > > 3. writeAnnotation allows you to serialize to DOM or SAX. Sax > implementation is trivial -- you only need to set correct > DocumentHandler on the SAXParser. For the DOM, you will need to make > some changes to be able to pass Document object to the DOMParser. As it > works now, AbstractDOMParser (org.apache.xerces.parsers) queries > document-class property in startDocument(..) method and creates an > appropriate document instance. You will need to modify this method to > allow using pre-set Document class -- the class could be either set via > a new public method or via new INTERNAL xerces property. > > Let us know if you have more questions. > > [1] http://www.w3.org/TR/xmlschema-1/#declare-annotation > > > Hope it helps, > -- > Elena Litani / IBM Toronto > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
