1. Disclaimer

LotusXSL is not at this time a product. For the moment the term 'Developer Preview' is the best description of it. It implements a draft standard that is still in design. APIs will continue to change, with no promise of backwards compatibility. This software contains known and unknown bugs (well... I think all software does...). I would recommend against using this software for mission critical applications. Lotus Development Corporation will not take responsibility for any disaster that may ensue from using this software.

2. Introduction

XSL is a language for expressing stylesheets and other types of transformations. It consists of two parts:

XSLT is a language for transforming XML documents into HTML, XML, Text, or other type of document. It is composed of the XML XSLT vocabulary, and XPath, a language for addressing parts of an XML document, designed to be used by both XSLT and XPointer.
An XML vocabulary for specifying formatting semantics (called Formatting Objects).

LotusXSL implements at this time only the first part of XSL, the W3C Working Draft 13 August 1999 XSL Transformations (XSLT).

An XSLT stylesheet specifies the transformation of a class of XML documents by describing how an instance of the class is transformed into another XML tree of nodes.

LotusXSL uses Version 2.0.15 (preferred) or 1.1.16 of IBM's XML for Java (called "XML4J") to parse an input XML document, or it can be adapted to other DOM-producing mechanisms. LotusXSL produces SAX events, an output DOM, or XML result document based on the transformations specified in the XSL stylesheet.

The processor can be used from the command line, from a servlet, or from a wrapper applet.

Version 0.18.0 is a radical rearchitecture of LotusXSL. In addition to support for the new draft, the old QueryEngine has been re-written and broken into an XPath package, the liaisons and formatters have been moved to a com.lotus.xml package, the Stylesheet classes have been separated from the main processor and are reusable and serializable, and the XSL instruction handlers have been moved into separate classes. In addition, we are working on an XML Query API, and a driver interface for data providers. In general, this release should perform much faster.

LotusXSL is now thread-safe with an instance per thread. You can not run the same instance in multiple threads.

Note that stylesheets can now be compiled (into an internal format) and serialized to disk. The resulting disk files are *big* (I mean really big), but this feature might be useful to someone.

See XQuery Overview for information about the XML query architecture.

In the SQL directory is an implementation of a XLocator object that interfaces with JDBC databases to return a canonical row-set XML. This will allow transformation of SQL data to any XML or HTML. The rows try to be lazily evaluated, though I have some more work to do for making this work well.

Code provides base servlet support to apply XSL stylesheets (retrieved from various sources) to XML (also retrieved from various sources). See the DefaultApplyXSL class. (Many thanks to the WebSphere folks for this.)

Stylesheets that worked with the previous version of LotusXSL will have to be upgraded to match the new draft. Watch for stylesheets that declared <xsl:stylesheet> with the xmlns:xsl="http://www.w3.org/TR/WD-xsl" namespace declaration. The namespace should now be xmlns:xsl="http://www.w3.org/XSL/Transform/1.0".

LotusXSL contains a DTD for XSL when used for HTML. (Thanks be to Henrique M. Holschuh for this). This DTD should not be used for run-time production when performance is a concern, but is useful during the development stages. It is used in the readme XSL.

A local copy of the IBM LotusXSL Comercial License Agreement is included.

Special thanks goes to Sanjiva Weerawarana of IBM Research, who has implemented the extension mechanism of LotusXSL, and has helped me in countless other ways with this software.

Thanks goes to Thomas Rowe, Don Day, David Marston, David A Epstein, Noah Mendelsohn, Dean Burson, Joseph Kesselman, Pat O'Connor, Mike Pogue, Chip Faulkner, Paul Dick, Adam Peller, David Bertoni, the XSL WG, and others, for the support, help, and feedback they have given! Also, many thanks for all the folks who have downloaded LotusXSL so far, and given me feedback, whether good or bad, and bug reports! Special thanks to Dan, Henrique, Adam, and Craig!

3. Installation Instructions

Need to run the following command to extract the files
tar xvf lotusxsl_0_18_1.tar
These commands create a lotusxsl_0_18_1 sub-directory in the current directory containing all the files.
Make sure to have XML4J on the system class path

To test, you can type run -in test.xml -xsl test.xsl. You should see a bunch of HTML output.

In the LotusXSL root directory there is a 'run' batch file, which you pass command line parameters. These instructions use version 2.0.15 of XML4J with the DOM classes. To use XML4J 1.1.16 you'll need to change the batch files to call com.lotus.xsl.Process or, to use the 2.0.15 version with the TX classes, call com.lotus.xsl.xml4j2tx.XML4JLiaison.

The command line utility is com.lotus.xsl.Process. It can take the following command line switches:

        -IN inputXMLURL
        -XSL XSLTransformationURL
        -OUT outputFileName
        -LXCIN compiledStylesheetFileNameIn
        -LXCOUT compiledStylesheetFileNameOutOut (it will be big!)
        -PARSER fully qualified class name of parser liaison
        -E (Do not expand entity refs)
        -V (Version info)
        -QC (Quiet Pattern Conflicts Warnings)
        -Q  (Quiet Mode)
        -LF (Use linefeeds only on output {default is CR/LF})
        -CR (Use carriage returns only on output {default is CR/LF})
        -ESCAPE (Which characters to escape {default is <>&"'\r\n}
        -INDENT (Control how many spaces to indent {default is 0})
        -TT (Trace the templates as they are being called)
        -TG (Trace each result tree generation event.)
        -TS (Trace each selection event.)
        -TTC (Trace the template children as they are being processed)
        -VALIDATE (Set whether validation occurs.  Validation is off by default)
        -EDUMP (Do stackdump on error)
        -XML (Use XML formatter and add XML header)
        -TEXT (Use simple Text formatter)
        -HTML (Use HTML formatter)
        -PARAM name expression (Set a stylesheet parameter)

4. Samples

See the Demonstration Page for some fairly simple examples of producing XML-styled HTML. [NOTE: This requires Java 1.1 and so will not work on earlier versions of browsers that only support Java 1.0.] Most of the XML and XSL files that are used, are in the 'samples' directory. However, probably the best sample is the XSL and XML that produced this file. It is located in the 'readme_production' directory.

I hope to be getting some more work done on better samples over time. If you come up with something cool that you wouldn't mind sharing, please let me know!

5. Test Suite

Our test suite is being moved into a separate download on AlphaWorks, and will no longer be contained in the LotusXSL zip file.

6. Performance Notes

We did a great deal to improve performance in version 0.18.0. However, as these things often go, we didn't quite make our goal for 10x performance increase. This list of what we understand the bottlenecks to be is provided in order that users may understand the issues, and work around the issues in some cases. The bottlenecks seem to be:

Class load and JIT (Just In Time compiler): As we've added architecture to improve raw transform performance, we've also increased the number classes and instances that need loading, and the amount of code that needs JITing. Although we can shrink the code size a bit, the time to load the processor will continue to be an issue. The one thing the caller of the API can do is to pre-warm the JIT by transforming a dummy document. An example of this technique can be found in the com.lotus.xsl.client.LotusXSLControl class, in the init() function, where I try to pre-warm the JIT so that click response time is reasonable.
Stylesheet Compilation: LotusXSL now reads the stylesheet from the parser's SAX events (unless you give it a DOM, in which case it produces SAX events from the DOM), and builds an internal tree structure. This structure process the attributes, precompiles the XPath expressions, and does other things that can be done before it knows about the input tree. The construction of large stylesheets is not as fast as we would like. We have been focusing on the transformation itself, and will have to concentrate on optimizing the stylesheet compilation down the road a bit.
The DOM: DOM nodes are big. Every node, including attribute nodes and whitespace nodes, have a next, previous, parent, first child, last child, ownerDocument, name, value, userData, and a few internal flags. So building the input DOM can take a lot of memory management. We are working on ways to get around this when LotusXSL builds the input tree, while still maintaining our DOM input architecture.
Union Sorting in Document Order: XSLT has to return nodes in document order. Determining document order from a pure DOM is difficult. We have developed a method to determine the order that is pretty good, but it is not as fast as simply determining document order based on a integer. Also, the collation method of combining two node lists is not as effecient as it needs to be, simply because we haven't had time to address it yet (the issues are more complex than is first apparent). The net of this is that union expressions aren't scalable to large node sets. For instance, we have one user who has a large document with 3800 elements right under the root node, and they apply a standard select="*|@*|comment()|processing-instruction()|text()" expression to these nodes. The 3800 elements happen to have 3800 whitespace nodes interleafed between them, and the union selection has to create a nodelist of 3800 elements and 3800 whitespace nodes, and combine them, testing for document order as it goes. The result is the processing takes forever. We're working on ways to fix this, but, the answer for this user was to not do the union selection at the top level (they didn't care about the whitespace nodes anyway).
Selects in Large Documents: LotusXSL currently builds no inverted indexes for the input document, and does not take advantage of DTD structures to optimize, and otherwise does no special magic to locate nodes other than try to implement efficient algorithms. While inverted indexes and taking advantage of DTD structures are on our list for options, users should be aware that patterns such as '//foo' at the top of the tree can be expensive.
xsl:number and xsl:sort: Both sorting and numbering are not as effecient as they could be.

7. API

Please see the API Overview document for information about the LotusXSL Application Programmer's Interface and the source code.

The API Documentation is in the 'apidocs' directory.

8. Tips

The rule for whitespace that can be stripped is: If a text node contains whitespace followed by a non-whitespace character followed by whitespace, none of the whitespace is ever stripped. If a text node is all whitespace, and does not have siblings that are character data, it should be stripped. A CDATA section is interpreted as character data.
Be aware that select arguments for variables, or arguments passed through the command line via the -PARAM switch, are expressions, not strings. So to set a variable to a string you need to do:
<xsl:variable name="ExpressionTest" select="'A test for xsl:constant and xsl:value-of'"/>
Notice the extra quotes.
If you're using an XSL stylesheet with a DTD, attributes typed as ID's will have a problem when you try and put an attribute template into them. Just something to watch out for. I don't really have a solution for this, except that you probably can't type literal result element attributes as ID's for the XSL stylesheet.

9. Extension Mechanism

See the LotusXSL Extension Mechanism Doc for instructions on how you can extend XSLT with Java and JavaScript.

10. Proprietary Extensions

10.1. Redirect Extension (Multiple Output Docs)

LotusXSL has a built in extension to let you produce multiple output documents from a single stylesheet. The extension is limited right now to producing files, when the output method is "xml", "html", or "text". See Redirect class for details.

11. Version Notes

11.1. Not Yet Implemented

This section will give you an idea of what I haven't implemented yet.

For number formatting, I don't yet handle the letter-value, digit-group-sep, and n-digits-per-group, and sequence-src attributes. Also, I suspect my internationalization support for number is not all that hot.
xsl:locale is not yet implemented.

11.2. Known Bugs

Bugs I haven't been able to fix yet for one reason or the other.

11.3. Features and Enhancements To Do

There are lots of potential features that will be going into the draft. Rather than trying to enumerate these, I'll enumerate what I plan to do that is not driven by the draft.

Performance is an ongoing task....
Work on incremental rendering via transformations as the source tree is arriving.
XSL Tutorial!
I want to put in an editor's interface, for use by XML/XSL editors. This would (1) allow a stream interface on the result-tree DOM, and (2) allow structural changes in the source tree to be incrementally updated in the result tree, with proper events thrown, etc.
Directly related to the above item, a Rule Locator (for want of better term). This needs to be able to locate closest matching rules for actions when a given source node changes, for editing situations. This is easy when only patterns are used, but harder when selects and modes are used. May require some special interfacing to the output DOM.
Better internationalization.
Better documentation.

11.4. Changes Since Last Version

September 2, 1999 (Version 0.18.1)

Added getStylesheet() accessor for the stylesheet root to the XSLProcessor class.
Added processStylesheet(Reader xslReader, String styleBaseURI) to XSLProcessor class.
Added processStylesheet(Document styleTree, String styleBaseURI) to XSLProcessor class.
Added WINDOWS-1250 encoding to FormatterToXML.
Test for ISO prefix handling for ISO8859_X encodings.
If the doctype-public attribute of xsl:output starts with "-//W3C//DTD XHTML", put a space before the closing '/>'.
Go to default encoding if encoding is not supported.
Pass "Accept-Language" from HTTP request header to processor, when liaison does connect, do a setRequestProperty on the connection.

August 28, 1999 (Version 0.18.0d14)

Now allow XSLT elements to have attributes from non-XSLT namespaces.
Fixed bug with passing modes through the default rule.
Ripped old dispatchers out.
Made template names and modes QNames.
Fixed encoding so that it properly puts out system encoding instead of UTF-8.
Removed escape for \' in attributes, since attributes are always put out with \".
Removed escape for > in attributes.
Fixed exception when processing-instruction had no arguments.
Upped the maximum token queue size to 500.
Fixed cdata handling of ']]<'.
Fixed '//' whe using psuedo document fragments.

August 26, 1999 (Version 0.18.0d13)

lotusxsl.jar is now uncompressed for better load performance.
Merged in updates to ApplyXSL.java and DefaultApplyXSL.java.
Fixed bug where as error was flagged when parameters were passed via setStylesheetParam that weren't declared in the stylesheet.
Added XObject creation methods to the XSLProcessor so that simple clients who want to push params won't have to know about liaisons.
Added simple example program to show how to push stylesheet parameters.
Integrated Sanjiva code that lets you use simple extensions without declaring a lxslt:component element.
Bug in translate function fixed.
Implemented the unparsed-entity-url function, but this will have to wait 'till after V0.18.0 for a fully robust implementation. URLs are not resolved to absolute URLs, as is specified in the draft, and, if the SYSTEM identifier is not present, it is resolved to the PUBLIC identifier.
Flag early error when the ".[predicate]" syntax is attempted.
Finished work on cdata-section-elements for xsl:output.
BSF stuff updated to do avoid security exceptions in browser.
Client browser demos now work.
Added dummy transform in start method in applet to get the classes loaded and JITed.
Query stuff sort of fixed... still need to fix '//' from doc fragment.
Added make files, thanks to Sanjiva!.
Fixed simple template stylesheets.

August 24, 1999 (Version 0.18.0d11)

bsf.jar and xpath.jar have been merged into lotusxsl.jar. If you have bsf.jar and/or xpath.jar on your classpath they should be removed.
Default namespaces for attributes fixed.
Bug with testing for SCRIPT and STYLE tags for CDATA stripping fixed.
Output encoding finished up (to the extent I'll take it for 0.18.0).
cdata-section-elements for xsl:output implemented. I still hava a bit more work to do on this.
Extensions finished, but not heavily tested yet.
Multiple output documents implemented. See Redirect class.
Fixed errors on all functions that take optional args when arg is not present.
Fixed serialization (but the serialization files are WAY too big!).
Hopefully fixed diagnostics for xpath exceptions.
Merged new servlet code.
Added a bunch of servlet publishing of request/response & query params to param variables. Probably too much. Will have to be reviewed.

August 19, 1999 (Version 0.18.0d10)

The W3C Working Draft 13 August 1999 XSL Transformations (XSLT) should be used as the syntax reference.
The substring function is now complient with the spec.
Extensions are mostly implemented... we have some cleaning up to do.
The -indent switch now works again. You can specify it with or without an argument.
DTDs are now handled properly... it -validate is used, a derived DOM parser validates documents based on the presence (or absence) of a DOCTYPE line in the file.
Problem with infinate loops with XLocators fixed.
Fixed local-part() and local-name() throwing ArrayIndexOutOfBoundsException if no arg.
The namespace and local-part functions have been renamed to namespace-uri and local-name as per the draft.
Attribute sets should work again.
Better error reporting when XML4J returns SAX errors without a message.

August 16, 1999 (Version 0.18.0d9)

The W3C Working Draft 13 August 1999 XSL Transformations (XSLT) should be used as the syntax reference.
xsl:output mostly implemented, except for the cdata-section-elements attribute. result-ns and indent-result no longer work and will generate an error.
Character encoding implemented. Need to do more work on what get's escaped in the formatters.
Made XPath functions table driven. Added installFunction(String name, Function func) API, so callers can install their own built-in functions. This is mainly for use by the XSLT engine.
Added handling of disable-output-escaping attribute for xsl:text and xsl:value-of.
Removed default space attribute on xsl:stylesheet.
Fixed lots of problems with whitespace stripping.
xsl:strip-space and xsl:preserve-space should now work properly and allow use of WildcardName patterns (in fact, you can use full match patterns as long as they don't contain whitespace, but this should be considered a bug).
Implemented current() function.
Potentially nasty bug with apply templates selecting twice fixed.
Implemented debugger interface SelectionEvent for select events.
Whitespace tweaked so that it better matches XT.
URL and filename handling is MUCH improved.
I believe I fixed XML relative DTD URLs.
Fixed problems with base URLs for included stylesheets.
Fixed problem with top-level params not being reset.
Sanjiva's BSF stuff partly integrated.
Auto reset removed. You now need to call this again when using the processor multiple times. I need to add good error handling for when you don't.

August 08, 1999 (Version 0.18.0d8)

Attribute sets now fully implemented to July (August) specification.
Stylesheet parameters fixed. Also only allow parameters if declared.
Added SAX DocumentHandler process() APIs.
Fixed a bunch of bugs that occured when using various process() api calls.
Fixed some bugs to be threadsafe.
Optimized Attribute Value Templates.
Somewhat reduced memory usage.
Fixed Literal Result Element as Stylesheet so that it now works (section 2.3).
Now allow you to not have to specify a XML file when running on the command line.
Experemental implementation of XML4J 1.1.16-specific XLocator class.

August 02, 1999 (Version 0.18.0d7)

Namespaced names for templates, modes, attr sets, keys, locales, variables, parameters)
per-mill on xsl:locale renamed to per-mille
Make XSLProcessor class reuseable.
Added debugging interface.
Fixed key() traversal of attributes.
Fixed document-relative use of key().
Made StylesheetRoot and Stylesheet classes serializable.
Made it an error to create illegal name with xsl:element or xsl:attribute
Implemented line number support for stylesheet.
Made number applied to string that's not a number return a NaN.
Moved handleXXX functions to classes, implemented compiling stylesheet into proprietary tree structure.

July 18, 1999 (Version 0.18.0d6)

Fixed "//a" in match patterns.
Fixed apply-imports.
Fixed problems with import/include cacheing.
Made recursive include/import an error.
Fixed RTF copy to result tree.
Made setParam take object instead of expression string.
Error to make PI with illegal target name
pi() renamed to processing-instruction()
quo operator removed
qname renamed to name()
second argument for document() function
xsl:param-variable has been renamed to xsl:param
xsl:param renamed to xsl:set-param

July 14, 1999 (Version 0.18.0d5)

Fixed problem with namespaces with NodeTest.

July 14, 1999 (Version 0.18.0d4)

Implementation of com.lotus.locators.RowSetLocator extension object.

July 13, 1999 (Version 0.18.0d3)

element extension mechanism
function-available renamed to extension-function-available
Rename xsl:functions to lxsl:functions

July 12, 1999 (Version 0.18.0d2)

Rearchitecture for XPath. XPath is now in a separate package and jar, the old QueryEngine class is gone, and the XML Liaison stuff and the Formatter stuff is broken into the com.lotus.xml package.
Moved Stylesheet class from XSLProcessor to it's own file, and added a StylesheetRoot class.
Consolidated functionality behind process functions into a single function (thank god).
Made xsl-stylesheet PI robust (I hope).
HTML Formatter now puts out HTML-style name-only attributes for 'disabled', etc.
Fixed whitespace stripping so that it only counts #x20, #x9, #xD or #xA.
New draft support: != operator.
New draft support: Added built-in template rule for attribute nodes
New draft support: Embedded Stylesheets
New draft support: Default rules for modes (modes are sticky again).
New draft support: Arg to boolean no longer optional
New draft support: digit-group-sep has been renamed to grouping-seperator
New draft support: expr attribute on xsl:number, xsl:variable, xsl:param, and xsl:set-param renamed to select.
New draft support: Following axis excludes descendants, and preceding excludes ancestors.
New draft support: idref functionality moved to id()
New draft support: merge doc and docref functions to document
New draft support: merge key and keyref functions
New draft support: n-digits-per-group has been renamed to grouping size
New draft support: sequence-src attr on xsl:number has been removed (well, it was never supported).
New draft support: xml:lang on xsl:number renamed to lang

June 24, 1999 (Version 0.17.3)

New code provides base servlet support to apply XSL stylesheets (retrieved from various sources) to XML (also retrieved from various sources).
Fixed: rather nasty bug having to do with union with match patterns that are non-simple.
Fixed: casting of variables in equality expressons had some problems, leading to fatal errors for valid syntax.
Fixed: when duplicate attributes were being added, this was not checked.
Fixed: when attributes were added after a child of the element had been added, the attributes were being added to the next occuring element.
Fixed: idref() was working incorrectly -- it didn't parse multiple IDs in the same attribute.
Fixed: With system paths meant for Unix, the direction of the slashes could cause hard to diagnose problems. The slashes are now normalized.

June 15, 1999 (Version 0.17.2)

Fixed nasty problem with equality expressions, where the cast wasn't being done correctly, which would yeald incorrect results.
Fixed problem with using full path expressions (including '.') after a '//'. (The problem is the differences of handling '//' vs. from-descendents(), which are similar enough to share code, yet their arguments turn out to be different).
Made NOREPORTPARSEERROR, etc., public.
Added hack for IMG tag for HTML FormatterListener so it doesn't add extra linefeeds around this tag.
Fixed being able to turn indenting off via use of indent-result="no" (this still will not work if result-ns is not set and the -HTML flag is used, unfortunately).
Fixed problem with copying result-tree fragments from variables to the result tree
Fixed problem with param context being wrong.
Override getNamespaceOfNode when using XML4JLiaison4dom and make namespace prefix resolution not work with attributes, since the DOM (and the XML4J DOM) is broken in regards to getting the parent of an attribute.
Fixed problem with attributes being propagated up when used with result tree fragments.
Fixed problem with parameters being passed to extensions.
Fixed several parsing problems.
Fixed problem with optimization of simple queries, where it would select the wrong thing.
Hopefully fixed problem with using JavaScript in threaded situations.
Fixed whitespace stripping problems.
Fixed substring-after bug.

May 24, 1999 (Version 0.17.1)

Carriage return/linefeed normalization is now off by default.
Added -STRIPCDATA and -ESCAPECDATA command line switches.
No longer force case change when in HTML mode.
Only escape "\n\r" for HTML attributes.
Fixed several defects with xsl:copy-of.
Fixed several defects with expression parsing when using predicates with axes identifiers.
Several low-hanging-fruit performance enhancements.
Fixed bug where non-elements that were named the same as elements would match on element nodetests. The primary manifestation of this was the doc rule firing twice when a DOCTYPE declaration was present.
Updated the xsl-html DTD (Thanks Henrique!).
Added run.ksh (Thanks Bill!).

May 3, 1999 (Version 0.17.0)

Validation is now off by default. -VALIDATE will turn it on.
You can now set the formatter from the command line.
Added setStylesheetParam() function. For the moment, only one param is supported.
Better trace methods.
-edump switch to tell if a stack dump is done.
Major rewrite to support the April 22 XSLT draft.

April 7, 1999 (Version 0.16.4)

Upgraded to new version of the JavaScript engine.
Use wait/notify in LotusXSLControl to call the trusted thread.

April 2, 1999 (Version 0.16.4)

Completed SAX output feature. Adapted process() calls that take a writer to stream directly to SAX.
Fixed problem with xml: prefix in the generic DOM liaison.

March 22, 1999 (Version 0.16.4)

In ProcessXSL, Automatically launch the XML4J 2.0.X TX compatibility liaison if the version check fails. I can't do the inverse because the 2.0.X liaison won't load if XML4J 1.1.14 is present.
Bottle-necked all printing to the XMLParserLiasonDefault, and thus to FormatterToXML and FormatterToHTML. You now must use -xml4jprint if you want to use the XML4J PrintVisitor classes.
Better whitespace handling.
The caller can now control what characters are escaped in attributes.
Did some work to result-tree namespace handling... it now works much better with xsl:copy.

March 11, 1999

Fixed first-of-type and last-of-type: they were using getElementsByTagName, which get's elements by descendents rather than by children, thus failing if the list had children of the children being tested of the given type.

March 11, 1999 (Version 0.16.4d1)

Added undocumented -SSATV switch that lets you have AVTs interpreted in selects.

March 11, 1999 (Version 0.16.3)

Fixed problem with XMLParserLiaisonDefault in getNamespaceOfNode where it wasn't finding the parents of attributes.
Fixed problem in FormatterToHTML where it wasn't closing non-empty tags that didn't have children.
Fixed regression in query engine for the "/|etc" bug.
Fixed defect reported by Gunnar Teege: the select pattern "tml:re/id('st')" produces a StringIndexOutOfBoundsException in the QueryEngine tokenizer. Fix made in QueryEngine's tokenize method.
Added Bourne scripts.

March 10, 1999

Updated XML4J DOM liaison to XML4J 2.0.3.
Fixed XML4J versioning checks.
Catch exception when add of text nodes to Document node fails, and give warning.

March 8, 1999

Fixed nasty defect where patterns like "/|etc" would not work... anything where the root ('/') was followed by the '|' symbol.
When used with the HTML formatters, or the FormatterToXML class (used for non-tx serialization to XML), the single quote and 0x0A and 0x0D are escaped in attributes.
Fixed ugly defect where anything besides the document element root could not be selected.
Fixed bug where the close tag would sometimes not be added in the FormatterToHTML.
Moved indent handling up to the level of FormatterToXML.

March 3, 1999

In the TrustedAgent inside the LotusXSLControl, fall to sleep while waiting so you don't take all the CPU time (thanks be to Jinbo Chen for this).

March 2, 1999

Fixed problem with comma tokens in query engine.
Made extension calls not found into warnings in query engine. This fix and the one above fixes problems with the call() function.
Used element instead of DocumentFragment when doing early evaluation of xsl:macro contents, and also temporarily parent the element to the current result tree node. This fixes a defect where a java.lang.RuntimeException: Child does not have parent! error could occur with directives inside xsl:invoke.
Moved the LotusXSLControl class into the com.lotus.xsl.client package.
Fixed most JavaDoc errors.
Use Jikes to compile release version.

February 28, 1999

Fixed problem with use of base URI.
Added constructors with XML error listeners on XMLParserLiaisons.

February 26, 1999

Added -Q flag for quiet operation.
Fixed the defect where extra space was being added in xsl:attribute.
Fixed three-deep-macro bug... the late evaluation of xsl:contents was causing the incorrect contents to be evaluated. The fix, to do early evaluation and then clone the fragment into the result tree is less effecient. I don't think this is worth any more work since xsl:contents is going away in the next draft.
The above also contains a fix for a related bug having to do with the incorrect xsl:contents evaluation context.
Require value-of and for-each to have a select attribute.
For the doc name param to ParseXML in the XMLParserLiaison, allow null, also added another signature that doesn't have the doc name param.
Flag error when xsl:import occurs after some other stylesheet element.
Fixed use of relative URIs (I hope).
Fixed base directory to be the directory of the stylesheet instead of the processor.
Began work on Formatter interface, which will eventually become a SAX streaming output interface. Moved print function to be implemented by this interface instead of by the XMLParserLiaison. Made constructors for XSLProcessor that take both a liaison and a formatter.
Implemented xml4j2dom.XML4JLiaison4dom and xml4j2tx.XML4JLiaison for compatibility with XML4J 2.0.
Fixed bugs in XMLParserLiaisonDefault. Added formatter implementations for this also.

February 08, 1999 (Version 0.16.2)

Check attributes for non-existence by using getAttributeNode and testing for NULL, necessary because XML4J 1.1.4 (correctly) changed the return to an empty string when an attribute doesn't exist.
Rewrote pattern tokenizer by hand instead of using the Java version, trying to get better performance...
Added ID handling stuff, including going to a new XMLParserLiaison interface to get the ID, if xsl:id isn't used. This fixes a nasty bug when stylesheet writer expects the ID to be recognized from the stylesheet. Notice that XML4J doesn't recognize attributes specified as IDs unless in validation mode. When using one of the process methods that use a URL for the XML, thus making the liaison do the parsing, scan for xsl:id's while the parse is occurring.
Added -NV flag to turn off validation..., which is on by default.
Pre expand namespace prefixes for XSL tag lookups, then do lookups based on value from getExpandedName... a small increase in performance.
If the caller says it's OK, use the UserData field in the TX Child class to cache the expanded names of prefixes... a good performance improvement. This is off by default.
Made default for XML4J version 1.1.14. 1.1.9 should give off a warning.
Did various other optimizations, esp. in query engine and XMLParserLiaison.
Fixed defect in cascading for attribute sets.
Removed carriage return/linefeed normalization as default. Use -CR and -LF to normalize.
Fixed defect on secondary key sorts with xsl:sort.
Added xsl-html.dtd. Thanks be to Henrique M. Holschuh for this.
Added readme.dtd. Thanks be to Don Day for this.

February 01, 1999

Use mergesort instead of quicksort for sorting. Because of glitch I still have to sort by doc order when elements are equal. When strings are used in a number sort, I imitate XT, though I'm not absolutely sure this is the right handling (a phone number is interpreted as a number...). Sorting should have no bugs left in it, though performance is not what I would like yet.
Worked on xsl:number performance... achieved some improvement. Tried to implement count cache, but realized it's nearly impossible, and not the right thing to do.
Pass file name to ProcessXSL#parseXMLStream for better error messages.
Added QC switch to suppress pattern conflict warnings.
Do version check for >= 1.1.9 of XML4J. I might have to revisit this a bit for next version of XML4J.
Put switches in and API functions to control if Carriage returns or linefeeds are used in the output. Default is CR/LF.
Tried to clarify the docs for process API that input trees are imutable and result tree is the same one that is handed in, if one is handed in.
Implemented generic XMLProcessorLiaison, and made com.lotus.xsl.xml4j.ProcessXSL derive from it. Though functions are written to handle namespaces with a generic DOM, I have not tested them yet.
Added more error handling in several places.
Improved error reporting for patterns.
Fixed baz[foo=""].
Fixed handling of multiple input IDs when they come as the result of a select pattern.
Fixed handling of AndExpr and OrExpr, which I incorrectly interpreted from the BNF. Thanks be to Craig for this.
Fixed silly 'XML' that should be 'XSL' when warning or flagging an error.
Implemented XML4J-style version numbering.

January 25, 1999

xsl:apply-imports has been implemented.
Namespace quoting for the result tree, and general result tree namespace handling has been implemented.
Fixed bug with file opening for nested imports.
Fixed bug where '/' root anchors didn't work for match patterns.
Whitespace normalized to LF/CR for character-level fidelity with XT.
Whitespace bug fixed where whitespace outside of root was being added.

January 21, 1999

xmlns="http://www.w3.org/TR/REC-html40" result-ns="" now works for the LotusXSLControl applet class.
Fixed: Kill extra linefeeds. For the most part, I now have character-by-character whitespace equality with XT output. There's one small glitch when comments are found in the templates - an extra space is added.
Fixed: Bug with 'from' attribute in xsl:number
Fixed: Bug with count attribute default.
Fixed: Incorrect match in test17. Something to do with //.
Fixed: Attribute value templates should be constrained to the following types of attributes:
- All literal result element attributes.
- name attribute in xsl:element.
- name attribute in xsl:attribute.
- name attribute in xsl:pi.
- order, lang, data-type, case-order in xsl:sort.
- Number-to-string conversion attributes in xsl:number, xsl:counter, and xsl:counters.
- value attribute in xsl:constant.
- value attribute of xsl:arg.
- default attribute of xsl:macro-arg (I think).
Fixed: <xsl:if test=""> generates a null pointer exception.
Fixed: Better Error message for when XML4J not found.
Fixed: Comments inside macros should not be written to result tree.
Fixed: Fixed realestate demo. It now uses a newly added setStyleSheetAttribute() method.
Fixed: Kill extra linefeeds.
Fixed: Bugs in ancestor() and ancestor-or-self().
Fixed: Null pointers for empty select or match expressions.
Fixed: ID not found trying to process parent of document element.
Fixed: Indent result attribute not processed correctly.
A bit better error detection for bad patterns. Still have some major work to do on this front.
Fixed: Compile error with Java 1.2.
Fixed: Getting the attribute of an ancestor doesn't work in the new version..
Fixed: When you don't specify an output file on the command line, a null pointer exception is thrown rather than printing the output to System.out as it used to do.
Added -v switch to give version info.
Started errors portion of testsuite (but only have three cases so far).

January 15, 1999

Major API cleanup
Handling of namespaces in patterns
Importance fixed for Macro, Constant, etc.
Made nodes non-sticky, as they're supposed to be.
Empty attributes are now properly copied into the result tree.
Stylesheet comments are no longer being automatically copied into the result tree.
Some XT compatibility tested (more to go).
Use HTML Print Vistor when result-ns="HTML"
Parent of document element is incorrect is fixed
Worked on whitespace
Added SelectExpr for uri()
bug - test1, a & a[emph] should have conflict fixed
Bug with not showing conflict with "*" and "foo" fixed
Don't use default namespace for element type names in patterns
Fix Attr return null for getParentNode
Implemented ancestor-or-self
Implement Error Handlers (test7)
Implement style PI in ProcessXML
Removed attribute copy on xsl:copy
Removed old syntax support
Reviewed and add source code comments
Turned Entity Expansion on in XML parser
Implemented XSL/XML Exception throwing
And a bunch of other bugs fixed along the way.

December 20, 1998

Fixed result tree counting so it doesn't take the name attribute as a result tree element name.
Implemented counter-scope.
Fixed a problem with conflict resolution with "*" and element patterns.

December 17, 1998

Implemented handling of the stylesheet PI in the XML doc from ProcessXSL.
Fixed xsl:copy so that it does not automatically copy attributes that are not xmlns.

December 15, 1998

Added proper ID syntax and semantics for new draft -- IDs must be quoted, or they will be interpeted as patterns.
Added lotusxsl:css-style-conversion support as stylesheet element, and removed -T support.
Fixed bug with select qualifiers on text(), attributes, pi's, doc nodes, and comments.
Implemented result tree counting. Not sure I have it right.
Implemented sorting (xsl:sort).

December 11, 1998

Added Process() APIs on com.lotus.xsl.xml4.ProcessXSL.
Did some stuff with whitespace. Defaulted indent to zero, made it so whitespace is only stripped from attributes and value-of based on the normal XSL whitespace dance, did callback to XMLParserLiaison for isIgnorableWhitespace, so it can strip these nodes.
Added public getQueryEngine() call to XSLProcessor.

November 28, 1998 to 18-May-1998

Details deleted.

18-May-1998 to 30-June-1998

Old submission version. Details deleted.

12. Contact Info and Bug Reports

Contact information for LotusXSL: Scott Boag

13. Glossary

XSL Instruction: Any tag with an XSL namespace prefix.
XSL Template Instruction: Any tag with an XSL namespace prefix that occurs inside an xsl:template element.
Template Child: Any node that is a child of an xsl:template element.
Source Tree: The tree input to the XSL process.
Result Tree: The tree that is output by the XSL process.
Stylesheet Tree: The stylesheet tree produced from the XSL file.
Pattern List: A parsed query or match pattern.

Release Notes for LotusXSL