XQSharp Change Log
1.5.0
New features
- XML version (1.0 or 1.1) can now be specified in the static context.
- Results can now be serialized as XML 1.1.
- Query plans can now be output as XML.
Changes
- Mathematical functions due to be included in the next version of XPath and XQuery Functions and Operators have been moved into the
math namespace.
- Rules for determining the validity of NCNames and QNames have been updated to reflect changes introduced in the 5th edition of XML 1.0.
- The types of external parameters can be specified via the static context.
- Improved efficiency of variables which are only evaluated once.
- Improved performance of user defined functions.
- Improved performance of functions which recurse on a subsequence of one of their arguments.
- Small
for loops will be unrolled.
- Prevented optimizer from optimizing away runtime errors in some circumstances.
- Added handling for MIME types in the
DocumentSet. This is required for correct behaviour of the XSLT document function.
- Introduced a new optimization level (
OptimizationLevel.Specialization. This enables potentially expensive function specialization. This can be selected using -O4 with the command line tool.
Bug fixes
- In some circumstances, namespace declarations could be erroneously omitted.
- Fixed nested XML Schema validation.
- Fixed output of empty XHTML elements during serialization.
- Many small fixes.
See the breaking changes in this version.
1.4.0
- Modules of functions can now be written in .NET and used in XQuery programs and XPath expressions.
-
Intermediate (constructed) nodes are now serialized directly into XdmDocuments, rather than building temporary
trees. This should lead to improved performance of queries performing further processing on constructed nodes.
- Some more improvements to join analysis - left outer joins and group-by expressions are spotted (and hence optimized) more often.
- Typeswitch expressions are now more efficient when the argument is a single item.
- XPathAtomicValue.ToString() now returns the value cast to a string.
- Many small fixes.
See the breaking changes in this version.
1.3.0
- The XQSharp DLL can now be precompiled with ngen. Previously ngen would crash.
- Improved unboxed evaluation, especially for aggregate functions.
- Many small performance improvements.
- Internal changes in preparation for compiling queries to dynamic methods
1.2.0
- Queries can now be evaluated directly to a serializer, avoiding building an intermediate representation.
- Push-based evaluation has been implemented for all expressions, and is used for all expressions that are determined to be strict.
- Added the
(# xqsharp:eager #) pragma to force eager evaluation.
import schema now ignores location hints if the namespace is already statically available.
- Imported schemas that fail to compile are no longer added to the available schema set.
- Fixed an exception when evaluating
treat as empty-sequence() in some circumstances, and improved optimization of type constraints.
- Corrected behaviour of
MoveToId on XPathNavigator produced for constructed nodes.
- Updated liveness analysis to fix problems when using partially evaluated lazy variables in join operations.
- Avoided
distinct-doc-order for unions like (@*|node()).
- Minor improvements to error messages.
- Many small performance improvements.
1.1.0
- Completely new serialization implementation.
- Serialization interface now much closer to
System.Xml.XmlWriter. There are some breaking API changes in the XQSharp.Serialization namespace.
- A new serialization method has been added which allows for serialization of Canonical XML 1.0.
XmlWriter instances can be created with the full set of serialization settings.
- Some new serialization parameters have been added.
- XQSharp now only requires .NET Framework 2.0, rather than 3.5.
- Added XQSharp extension methods for .NET Framework 3.5 to give a simple way of executing XPath 2.0 expressions.
- Added some more constructors to
XdmDocument, to match XPathDocument.
- Removed
IXPathNavigable overloads for methods taking XPathItem, as this was causing calls with XPathNavigator to be ambiguous.
- Reverse axes were previously being cached and reversed. This is now skipped if the results are not needed in document order - eg. for
/ancestor::*[1]
- Improved optimization of
fn:data applied to constructed nodes.
- Conformance to latest errata regarding in-scope schema types from modules.
-
Bug fixes:
- XQuery parsing changed to reflect erratum regarding the leading-lone-slash constraint.
- The order comparison of navigators over
XdmDocument instances was incorrect for namespace nodes.
- The
fn:base-uri, fn:data and fn:node-name functions behaved incorrectly for namespace node arguments.
- The parsing of URI ports was incorrect in
AnyUri.
- The
UserInfo property on AnyUri was truncating the first character
- Empty document nodes were incorrectly causing spaces to be inserted between sequences of atomic values when constructing an element's content
XPathNavigator instances were sometimes not being cloned when evaluating a query to a single node, leading to the returned navigator pointing to the wrong node.
- Fixed some places where the static types of expressions were not being correctly updated during optimization, causing a
NotSupportedException or ArgumentNullException at runtime.
- In some cases function specialization was causing
fn:subsequence to throw a NotSupportedException at runtime.
- The optimized form of
fn:sum($x to $y) was producing xs:decimal values instead of xs:integer values.
fn:distinct-values was missing an evaluator to produce XPathItem for upstream functions, leading to a NotSupportedException at runtime.
- The in-scope namespaces of a constructed element did not include the empty prefix in some situations.
- The
fn:element-with-id function gave incorrect results for schema-defined union types including xs:ID.
1.0.2
- Set default type checking mode to optimistic, as per documentation.
- Fixed compatibility with XmlDocument
- Fixed NullReferenceException when using data() caused by XmlDocument returning an IXmlSchemaInfo instance with null SchemaType.
- Fixed infinite loop caused by XPathNodeIterators over XmlDocuments not consistantly returning false after the end of the node set.
- Fixed the XQuery version declaration for cultures where '.' is not the decimal point.
- Fixed normalization rules of general comparisons as per erratum XQ.E18.
- For increased performance, the command line tool now runs in x86 mode by default.
1.0.1
First full release.