Add the script-syntax tags that had no token at all

[ghedwards]

Closes #47.

I enumerated instead of guessing

The issue said "cfimap and others" without saying which. The repo ships the answer — cfml.dictionary's syntax dictionaries. Cross-referencing them against cfmlFunction:

• 118 tags across cf8/9/10/11, cfml7, railo3/4, lucee4.5/5
• 60 had no token
• 21 of those are control flow — cfif, cftry, cfreturn, cfparam, cffunction, cfcomponent, cflock, cftransaction … — and already have dedicated grammar rules. Adding them to cfmlFunction would be actively wrong, so they stay out.
• leaving 39

Prefixed spelling only, and that's the whole design

The issue worried that each new token is a new reserved word. It is — and the bare names here are the worst possible candidates: dump, trace, map, xml, flush, timer, invoke, login, document, textarea, calendar are ordinary functions and variables in real code. dump(...) alone appears in four fixtures in this repo. A bare DUMP token would have reclassified every one of those calls as a tag.

So these 39 get cfdump, cftrace, cfmap … and no bare token. That neutralises the hazard by construction rather than by hoping.

It's a deliberate departure from the aliases added in #29, which map cfX -> type(X) onto an existing bare token. Where no safe bare token can exist, the prefixed spelling stands alone. imap is the single exception — it's the one bare spelling the corpora actually use, so it gets both, with CFIMAP aliasing onto it.

Every new name also goes into the identifier rule, in the same commit. That's the fourth-time-lucky insurance: APPLICATION, INSTANCEOF and CT/NCT each broke working code because the entry came later.

The fixture pins both directions

imap action="close" connection="testImap";     // the tags parse
cfspreadsheet action="read" src="report.xls";
cfdocument format="pdf" { echo("body"); }

dump(results);   trace = 1;   map = {};        // and the names still work
xml = "<a/>";    flush = false;   cfdump = 1;

Verification

• 313 cfparser tests, ./gradlew build
• Differential harness 8 → 7, nothing newly broken
• CFLint 675 against a clean build — 40 new tokens renumber every constant, so this is exactly the case the recompile note in CLAUDE.md is about
• Fixture fails with cfml.parsing/src/main stashed

Still not complete, by design

The 21 control-flow tags are excluded permanently. The 39 here have prefixed spellings only. If a corpus later needs a bare chart or timer, that's a deliberate decision with a known cost, not an oversight — and the enumeration above makes it a five-minute change rather than an investigation.

---

Generated by Claude Code