[bpamiri]

Fixes #4187

Summary

The singularize rule ((a)naly|(b)a|(d)iagno|(p)arenthe|(p)rogno|(s)ynop|(t)he)ses$ → \1\2sis added group 2, the leading a of analy, a second time. So psychoanalyses came back as psychoanalyasis. The rule is now (analy|ba|diagno|parenthe|progno|synop|the)ses$ → \1sis.

For the other alternatives group 2 was empty, and a bare analyses is caught by the earlier (^analy)ses$ rule. So output changes only for a word that ends in analyses without starting with it.

Upgrade notes

• The changelog fragment is a changed one (changelog.d/4187-singularize-prefixed-analyses.changed.md), because singularize() output changes for words ending in analyses without starting with it.
• The 4.1 to 4.2 upgrade guide gets an entry under "Behaviour changes to check".

Test plan

• New spec. vendor/wheels/tests/specs/global/SingularizeSesRuleSpec.cfc checks that prefixed -analyses words are fixed. It also pins analyses, bases, diagnoses, parentheses, prognoses, synopses, theses and hypotheses to their current output.
• RED on develop: Lucee 7 and BoxLang each fail 1 spec (psychoanalyses → psychoanalyasis).
• GREEN: 2/2 on Lucee 7 and BoxLang.
• global directory (pass / fail / error):

Engine	Result
Lucee 7	307 / 0 / 0
Lucee 6	307 / 0 / 0
Adobe ColdFusion 2023	306 / 0 / 0
Adobe ColdFusion 2025	306 / 0 / 0
BoxLang	307 / 0 / 0
RustCFML v0.693.0	306 / 0 / 0

• Full suite (tools/test-local.sh, Lucee 7 + SQLite): 6524 passed.

• After merging develop (for 4.2): global specs (SQLite) 310 / 0 / 0 on Lucee 7, Lucee 6 and BoxLang, 309 / 0 / 0 on Adobe 2023, Adobe 2025 and RustCFML v0.693.0; full core suite, Lucee 7 + SQLite: 6664 passed; guides build passes.

🤖 Generated with Claude Code

[bpamiri]

rev1-r1 review @ c0caee1 (draft, 4.3): APPROVE on code; merge gated on compat-matrix run 37113979661 and on 4.2.0 shipping.

((a)naly|(b)a|(d)iagno|(p)arenthe|(p)rogno|(s)ynop|(t)he)ses$ → \1\2sis becomes (analy|ba|diagno|parenthe|progno|synop|the)ses$ → \1sis.

• The old \2 was a bug. Group 1 already holds the whole stem, and group 2 is the (a) of the analy alternative only. So "…analyses" (anything the anchored (^analy)ses$ rule above doesn't catch, such as psychoanalyses) became "…analyasis". For every other stem \2 was unset: empty on Lucee and Adobe, and BoxLang's unset-group error (the same class as pluralize() throws on BoxLang for words ending in -ex, -lf or -rf #4185) for bases, theses, diagnoses, parentheses, prognoses, synopses.
• The new rule gives stem + sis for all of them. Plain analyses is unaffected, since the earlier anchored rule still takes it.
• The spec covers the prefixed analyses case and the other stems, and is RED on Lucee 7 and BoxLang.

[github-actions]

Wheels Test Results

216 679 tests   212 555 ✅  1h 8m 8s ⏱️
 16 576 suites    4 124 💤
     32 files          0 ❌

Results for commit 331c195.

♻️ This comment has been updated with latest results.

[bpamiri]

rev1-r1 correction: in my review above I said the unset \2 made BoxLang throw for bases/theses/diagnoses/…. That's wrong. BoxLang only throws when group 1 is unset (#4185), and on develop those words already return basis/thesis/diagnosis on BoxLang. So #4188 fixes wrong output on every engine (psychoanalyses → psychoanalyasis), not a crash. Matrix 37113979661 on c0caee1 is fully green: all 32 legs discovered 497 bundles with 0 failures and 0 errors, SingularizeSesRuleSpec is 2/2 on every leg, and RustCFML and the gate are green. OK to merge once it leaves draft after the 4.2.0 cut.

[bpamiri]

Final: APPROVE @ 331c195 (lifted into 4.2; matrix 37263380571, all 32 legs read)

Delta since my approval at c0caee1

• The code is identical.
• The develop merge has one combined hunk in global/strings.cfm, because fix: stop pluralize() throwing on BoxLang for -ex, -lf and -rf words #4186 changed the adjacent pluralize rule list. It is a pure take-both: the merged pluralize line equals develop's, and the singularize line equals c0caee1's.
• The new 4.1→4.2 guide entry is accurate. Only a non-anchored …analyses word was misspelled. bases, theses, diagnoses and the rest come out unchanged.
• The fragment is retyped to changed.

Matrix

• 32/32 legs have 0 fail, 0 error.
• SingularizeSesRuleSpec passes 2/2, and fix: stop pluralize() throwing on BoxLang for -ex, -lf and -rf words #4186's InflectionRuleGroupsSpec 3/3, with 0 skipped in every leg (lucee6/7, adobe2023/2025, boxlang × every database). So both rule-list changes hold together.
• The set of skipped specs is identical to a develop-based run (perf(router): index dynamic routes by method and first two path segments #4183's matrix 37261921912) on lucee7-mysql, adobe2025-sqlite and boxlang-oracle.

No open findings.