CR-1a: ee[-] Parser-Fix (Silvain) — lookbehind trennt e[-] vom führenden Allel

normalizeToken: lookbehind-Regex /(?<=[A-Za-z])e\[-\]/→'?' greift wenn e[-]
von einem Buchstaben (erstes Allel) gefolgt wird (ee[-] → e?); standalone e[-]
fällt durch auf die generische \[-\]→? Regel (→ e?). Beide Pfade liefern
splitToken 2 Allele [e,?]. Vorher: ee[-]→ee?→ 3 Allele → Fehler.
Fixture: Silvain 'aa c[chm]c[chm] Dd ee[-] Gg Pp Spsp' → E=[e,?] ✓.
Gate: build ✓  eslint ✓  vitest 92/92 ✓  e2e 126/126 ✓

gerbil-manager-web/src/genetics/__tests__/genetics.test.ts

@@ -602,12 +602,21 @@ describe('GEN-3h: breeder bracket-notation display + E-locus e-before-ef order',
     )
   })
 
-  it('[-] bracket-unknown parses as wildcard (e[-] → E=[e,?], displays e-)', () => {
-    // NOTE: oracle for Silvain shows "ee[-]" which contains 3 E-allele tokens
-    // (e + e + [-]) and cannot be parsed. Flagged to god — see done-report.
-    // This test documents what [e,?] at E produces: "e-".
+  it('e[-] standalone: parses as [e,?], displays e-', () => {
     const g = fromDisplayString('aa CC DD e[-] GG PP spsp rere')
     expect(g.E).toEqual(['e', '?'])
     expect(toDisplayString(g)).toBe('aa CC DD e- GG PP spsp rere')
   })
+
+  it('CR-1a: Silvain oracle ee[-] parses without crash → [e,?], displays e-', () => {
+    // Real herdbook notation: ee[-] = fox allele e + unknown e-type second allele.
+    // The lookbehind rule strips the second e[-] → '?', leaving 'e?' for splitToken.
+    const input = 'aa c[chm]c[chm] Dd ee[-] Gg Pp Spsp'
+    const g = fromDisplayString(input)
+    expect(g.E).toEqual(['e', '?'])
+    expect(g.C).toEqual(['cchm', 'cchm'])
+    expect(g.D).toEqual(['D', 'd'])
+    expect(g.Sp).toEqual(['Sp', 'sp'])
+    expect(toDisplayString(g)).toBe('aa c[chm]c[chm] Dd e- Gg Pp Spsp rere')
+  })
 })

gerbil-manager-web/src/genetics/genotype.ts

@@ -179,11 +179,17 @@ function normalizeToken(tok: string): string | null {
   t = t.replace(/S\(l\)/g, 'Sl').replace(/s\(l\)/g, 'sl')
   t = t.replace(/Uw/g, 'G').replace(/uw/g, 'g')
   // GEN-3h: accept bracket display notation → canonical internal symbols.
-  // Order matters: [-] must be replaced before the plain-dash rule below.
-  t = t.replace(/\[-\]/g, '?')       // bracket-unknown [-] → internal wildcard
-  t = t.replace(/e\[f\]/g, 'ef')     // Schimmel display form → internal
+  t = t.replace(/e\[f\]/g, 'ef')     // Schimmel allele display form → internal
   t = t.replace(/c\[chm\]/g, 'cchm') // Colourpoint display form → internal
   t = t.replace(/c\[h\]/g, 'ch')     // Himalayan display form → internal
+  // CR-1a: allele-prefixed bracket-unknown like ee[-] (Silvain).
+  // When e[-] is PRECEDED by a letter it is the second unknown allele in a
+  // 2-allele token (e.g. ee[-] → e + e[-] → e + ?).  Lookbehind strips only
+  // the e[-] part; the leading allele stays. Standalone e[-] falls through to
+  // the generic [-]→? rule below (which makes the bracket-dash a wildcard,
+  // leaving the leading allele intact for splitToken).
+  t = t.replace(/(?<=[A-Za-z])e\[-\]/g, '?')
+  t = t.replace(/\[-\]/g, '?')       // bare/standalone bracket-unknown → wildcard
   // GEN-3c: plain dash is the breeder's UNKNOWN marker on input; store internally as '?'.
   t = t.replace(/-/g, '?')
   return t