3.27.1

Fixed issues

• fn:format-integer now returns the correct result for empty inputs and negative inputs
• The descendant axis retrieves child elements if the selector looks like self::*, which allows further bucket-style optimization. Used in Fonto to minimize dependencies for certain queries.

3.27.0

Fixed issues

• PathExpressions ending with a map lookup no longer annotate themselves as a map lookup. This fixes a rare static typing error that was unjustified.
• Traces from compiled scripts give a correct position again

New features

• Thanks to @rebp, we now have a minimal implementation of format-number! For now, only the A/a/I/i picture strings are supported!
• Trace messages with elements in them now output the element's XML, assuming the serializer argument is passed.

3.26.2

Fixed issues

• Fixed fn:random-number-generator?permute to work like the spec describes it
• Make fn:serialize output a single string instead of an array

Other improvements

• Simplify types in postFixExprWithStep, thanks to @HectorPeeters
• Add type annotations for single item sequences, again thanks to @HectorPeeters
• Remove slimdom-sax-parser from the dependencies, use the built-in parser from slimdom instead

3.26.1

Bugfixes

• Custom JavaScript functions are now allowed to return maps.
• Custom functions that return dates in maps are now correctly converting the values.

3.26.0

New functionality

The ResultType.ALL type is added. This type is a replacement for the new deprecated ANY type that returned items in an unpredictable way.

Improvements

The child axis is now lazy: it will not get all children if only the first child is used in the query. This narrows down the dependencies that a child::x[1] query would take.

3.25.1

Fixed issues

• the internal function to evaluate XPath and XQuery from within FontoXPath no longer reparses every expression it sees. It will instead cache the results.

3.25.0

New functionality

• The parser has been refactored from using PEGJS to using PRSC. This improved a lot of parsing-related errors as well because we're now parsing a lot more! Update your own buildtools if you need to expose this new library. It should work for 99% of cases, for the other ones, follow how you integrated xspattern. Many thanks to @HectorPeeters and @MTimmerije for the actual work!
• XQuery library modules can now also be parsed using parseScript

3.24.0

New functionality

• fn:serialize! With the help of @rrthomas we now support fn:serialize. Be sure to pass a serializer to your evaluateXPath calls before you can use it
• Better buckets. When parsing a selector like self::*[self::p], the stronger name-p bucket will be reconized as the bucket for the whole selector, improving performance
• Passing an AST to evaluateUpdatingExpression is now supported
• Namespaces are now annotated in an AST, making it possible to nest ASTs with different namespace configurations into each other

3.23.0

New functionality

• With the incredible help of @fhaftmann , fn:distinct-values is implemented!
• fontoxpath:evaluate now accepts XQueryX elements representing a full document

3.22.0

New functionality

• This release adds overloads for all APIs to pass an Abstract Syntax Tree as an XML document instead of a string. This allows one to generate an XPath or XQuery program, or combine multiple scripts into one.
• The name and local-name functions can now be compiled into JavaScript.

---

Bugfixes

• #434: Fix the outermost function returning wrong results when used with a descendant axis. This is now also covered by the QT3 tests.

---

Other improvements

• The and operator now returns the most specific bucket if it finds one. So the self::* and self::my-element returns the bucket for name-my-element instead of the first one. This may improve performance for a lot of nested and operators, or when using the bucket system. For more information on the bucket system, give our paper a read! Or watch us presenting it!