Implement declare decimal-format (XQuery 3.1)

[joewiz]

Summary

• Add parser support for XQuery 3.1 declare decimal-format and declare default decimal-format prolog declarations (spec section 4.10)
• The runtime infrastructure was already in place (DecimalFormat, XQueryContext storage, FnFormatNumbers 3-arg support) — this adds the missing parser recognition and tree walker processing
• Enables users to customize number formatting (e.g., European comma-as-decimal-separator conventions) via fn:format-number

Changes

File	Change
XQuery.g	Add tokens, grammar rules for named/default forms, property keywords in reservedKeywords
XQueryTree.g	Walk AST, validate properties (single-char, zero-digit, distinctness), register formats in context
ErrorCodes.java	Add XQST0097 (duplicate declarations) and XQST0098 (invalid/conflicting property values)
XQueryContext.java	Add setDefaultStaticDecimalFormat() convenience method
format-numbers.xql	8 new tests for named/default formats, custom NaN/infinity, and error cases

Closes #56

See also: eXist-db/exist-xqts-runner#44 (fix copy-paste typo in XQTS runner for exponent-separator attribute)

Test plan

• All 161 existing format-numbers.xql tests pass (0 failures, 0 errors)
• New tests pass: named format, default format, custom NaN, custom infinity
• New error tests pass: duplicate named (XQST0097), duplicate default (XQST0097), non-distinct properties (XQST0098), invalid zero-digit (XQST0098)
• ANTLR grammar generates without errors
• Full project compiles successfully

🤖 Generated with Claude Code

[line-o]

Thank you @joewiz for this welcome and valuable addition to exist-db.
From what I can see and understand the changes make sense.

A few questions I asked myself going through the changes:

• Why is the default UNNAMED?
• The newly added function requireSingleChar sounds like a general purpose one but it is very closely tied to decimal formats looking at the exception that is thrown
• Same is true for validateZeroDigit
• validateDistinctPictureChars is less ambiguous but I think all new functions would benefit from a clear prefix like df or so

[joewiz]

(CI note: Codacy keeps stalling on this PR, despite repeated restarts of the job. If only there were a way to restart one stalled job without restarting them all.)

[joewiz]

[Co-authored with Claude Code.]

Thanks for the review, @line-o! I've pushed a commit (bb4a1a1) addressing your points:

df prefix on helper methods — Great suggestion. Renamed all three validation helpers in XQueryTree.g:

• requireSingleChar → dfRequireSingleChar
• validateZeroDigit → dfValidateZeroDigit
• validateDistinctPictureChars → dfValidateDistinctPictureChars

Why UNNAMED? — The name comes directly from the XQuery 3.1 spec (§4.10), which defines the concept of a "default unnamed decimal format" that is always available. I've added Javadoc comments on both DecimalFormat.UNNAMED and UNNAMED_DECIMAL_FORMAT in XQueryContext citing the spec section. That said, if you think DEFAULT would be clearer as a constant name (while keeping the spec reference in Javadoc), I'm happy to rename it — just let me know.

[joewiz]

(CI observation: Looks like the XQTS failure needs us to implement one of the workarounds/fixes we discussed yesterday.)

[line-o]

@joewiz Please rebase this onto current develop. I am curious about the XQTS results

[line-o]

This PR lets us pass 36 additional test cases of XQTS.

The two new failures are

• re00062: this test case is flaky - there is an open issue check XQTS case re00062 #6056
• numberformat127 - this one looks related at first but then it is apparent that it only applies to XQ30 and not XQ31

  <test-case name="numberformat127" covers-30="fn-format-number">
      <description>
         PURPOSE: exponent separator rejected in 3.0
      </description>
      <created by="Michael Kay" on="2015-04-22"/>
      <modified by="O'Neil Delpratt" on="2016-09-08" change="Changed covers attribute to covers-30" />
    <dependency type="spec" value="XQ30"/>
    <test>
      declare decimal-format exp-E exponent-separator="E";
      fn:format-number(12345.678, '9.9999E999', 'exp-E')
    </test>
    <result>
      <error code="XPST0003"/>
    </result>
  </test-case>

[line-o]

just for reference basex has the same result as we would after this PR is applied
https://fiddle.basex.org/#/?share=('query!'declare%20decimal-format%20exp-E%20exponent-separator%3D%5C'E%5C'%3B%5Cn***fn%3Aformat-number%7B12345.67809.9999E999%220exp-E%22%7D')*%20%200%2C%20%22%010*_

[joewiz]

[This comment was co-authored with Claude Code. -Joe]

Closing — superseded by #6217 (v2/declare-decimal-format).

This work has been consolidated into a clean v2/ branch as part of the eXist-db 7.0 PR reorganization. The new PR includes all commits from this PR plus additional related work, with reviewer feedback incorporated where applicable. See the reviewer guide for the full context.