cfscript: support the ${ … } ordered-struct literal, keeping $[ … ] a subscript (#80)

[ghedwards]

Closes #80. Supersedes #107 (now closed), which took the opposite decision on $[ … ].

animals = ${ Aardwolf: "…", aardvark: "…" }; yields an ordered_struct, the same node the empty [:] and [=] forms already produced.

The decision the issue was blocked on

$ stays an ordinary identifier. Only the brace form is the literal, so $[ … ] remains an ordinary array-style reference — a subscript_expression — and every jQuery-shaped spelling keeps the tree it had:

input	tree
animals = ${ a: "x", b: "y" };	ordered_struct with pairs ← new
x = ${};	ordered_struct ← new
x = $[ 1 ];	subscript_expression
$ = 1; / x = $; / y = $.foo; / $( "sel" )	identifier, member, call
a = [:]; / b = [=];	ordered_struct
x = `a ${ b } c`;	template_substitution
x = "…?n=${trackingNumber}";	string

The reason is which postfix operators exist, not how common each spelling is. [ subscripts any expression, so $[ … ] already means something and a literal reading would take that meaning away; { is not a postfix operator on anything, so a $ followed by a brace cannot be a reference at all and ${ … } has no competing reading to lose. Stated that way the trade needs no corpus counting, and it cannot be re-argued from prevalence: even if nobody subscripts a variable named $, the expression is still well-formed CFML and the literal still is not.

That also settles Lucee's bracket spelling of the same literal, $[ a: "…", b: "…" ], as deliberately unsupported: one key: value inside brackets is Lucee's slice syntax and parses, while several comma-separated pairs are not an array reference at all. cfscript/subscript_multiple_pairs.cfc now pins a decision rather than a gap.

Verification — re-run on today's master

Rebased onto 098916e (after #122 and #131 merged) and re-measured, because #122 moved the corpus baseline by one node.

check	result
npm test	347/347 (117 cfml, 158 cfscript, 72 cfquery)
npm run probe	50/57, no drift
npm run lint / npm run fuzz	clean
corpus scan, 15,392 files	641 → 646 error nodes across 122 files
npm run treediff vs master	zero changed trees in both grammars
STATE_COUNT	5315 → 5361 (+46), no new conflicts

Per-file, the whole delta is two files:

  4 -> 3    lucee_Lucee/test/tickets/LDEV3133/test.cfm          the ${ half fixed, the $[ half declined
 12 -> 18   cfwheels/tools/vscode-ext/assets/templates/controller.cfc

The increase is a VS Code snippet template, not CFML — its ${modelNamePlural} placeholders sit in code position. It failed before (12 nodes) and fails after (18): recovery reshaped, validity unchanged, which docs/FAILING-PATTERNS.md already describes as not a reason to back out a fix. Nothing that parses moved, as the zero-change tree diff confirms.

'${' was already a token of this grammar — template_substitution inside a backtick string — so this admits an existing lexical form in a new position rather than adding one. Every other ${…} in the corpus sits inside a string literal (Java-style placeholders in Slatwall's shipping URLs, JS template literals in TestBox's coverage browser), and a string lexes as one token.

Also in the diff

• Corpus test and probe cfscript/ordered_struct_brace.cfc pin $ as a variable — $[ 1 ], $ = 1, $( "x" ) — alongside the literal, because that is the half this decision protects.
• LIMITATIONS.md and docs/FAILING-PATTERNS.md: the ${ entry becomes a record, and the subscript-multiple-pairs row states a decision instead of describing a gap.
• CHANGELOG.md under ## [Unreleased], and the in-flight entry in docs/TODO.md.

🤖 Generated with Claude Code

https://claude.ai/code/session_0117EvdX3EgLgDevVNrWZ5hr