@qt4cg statuses

90% XQuery, 10% human eyes

Closes #2891

(PR is coming)

Splits the supports-dtd option into supports-dtd-validation and supports-dtd-attribute-typing on the grounds that some commonly used XML parsers (for example the Microsoft .NET parser) support one and not the other.

Fix #2887

The current equivalence rules require a processor to work out which components of the captured focus a function body can actually observe. This can become fairly expensive and will only be required for edge cases, like function items that capture different focuses yet always return the same result. An example:

1 ! fn:last#0
2 ! fn:last#0

I believe that comparing the captured focus as a whole makes the rules simpler. A processor will still be allowed to recognize those cases.

XSLT Specification 7.2 The xsl:iterate Instruction - Example

The example data contains lines like this: { "date": "2008-09-01", credit: 12.00 },

Saxon 13.0 produces this when using the data: FOJS0001 Invalid JSON input on line 2: Property name must be a string literal (found <UNQUOTED_STRING>

Invalid JSON input on line 2: Property name must be a string literal (found <UNQUOTED_STRING>

Should I raise an issue with Saxonica about the missing closing ) in the error?

Summary of the most important changes:

• I have removed the explicit listing of the default function properties both in the code as well as in the rendered representation (deterministic, context-independent, focus-independent); they are stated once in a new sectin §1.5.4.
• The arity prose has been changed to a single sentence naming the omitted argument, for example: “If $node is absent, this function is context-dependent and focus-dependent.”
• Various wrong statements were fixed (fn:format-integer, fn:format-number, fn:civil-timezone, fn:contains-subsequence, fn:load-xquery-module, fn:in-scope-prefixes`, and others).
• Functions like fn:transform now name the component they depend on.

Closes #1459

In this PR, numeric specific subtypes are changed back to the general xs:integer (see #2849 for the rationale).

It also fixes an inconsistency in the CSV rules…

parsed-csv-structure-record?column-index was map(xs:string, xs:integer) with the prose “column positions (as 1-based positive integers)”, while ?get in the same record type uses xs:positiveInteger.

…and it now allows users to also supply negative input for fn:unix-dateTime (similarly: “If the value is out of range …”).

Transferred from #2837

Thanks very much for these changes! Of course I have a few small comments:

To answer your question, the constructor attribute was (and is) missing from the xsl:record-type element. Sorry I wasn't more specific. xsl:item-type allows mixed content, but I don't think it should. xsl:item-type has an assertion requiring the "as" attribute, but the spec says it's optional. The shadow attribute "_main_module" has two underscores, but it should be "_main-module" with an underscore then a hyphen. (This is a new one) xsl:merge-key is in the substitution group for xsl:instruction but I don't think it should be. Looks like I'm not able to reopen this issue...please let me know if you'd like me to create a new issue.

Thanks again!

As we provide the option parse-xml: strip-space=conditional I think we should make it work consistently with the same option on fn:doc and fn:document - it should strip whitespace according to the XSLT strip-space/preserve-space declarations.

The default should remain "none" to preserve backwards compatibility.