From 460d9e8dd6ea86e11a00f01e7e09c6ad73bcd3dd Mon Sep 17 00:00:00 2001
From: Virgil <virgil@lethean.io>
Date: Wed, 1 Apr 2026 23:08:59 +0000
Subject: [PATCH] fix(reversal): include contraction auxiliaries in fallback
 signals

Co-Authored-By: Virgil <virgil@lethean.io>
---
 reversal/tokeniser.go      | 19 ++++++++++++++-----
 reversal/tokeniser_test.go | 12 ++++++++++++
 2 files changed, 26 insertions(+), 5 deletions(-)

diff --git a/reversal/tokeniser.go b/reversal/tokeniser.go
index 7ed8189..730e64b 100644
--- a/reversal/tokeniser.go
+++ b/reversal/tokeniser.go
@@ -546,11 +546,7 @@ func (t *Tokeniser) buildSignalIndex() {
 			t.verbAux[core.Lower(w)] = true
 		}
 	} else {
-		for _, w := range []string{
-			"is", "are", "was", "were", "has", "had", "have",
-			"do", "does", "did", "will", "would", "could", "should",
-			"can", "may", "might", "shall", "must",
-		} {
+		for _, w := range defaultVerbAuxiliaries() {
 			t.verbAux[w] = true
 		}
 	}
@@ -564,6 +560,19 @@ func (t *Tokeniser) buildSignalIndex() {
 	}
 }
 
+func defaultVerbAuxiliaries() []string {
+	return []string{
+		"am", "is", "are", "was", "were",
+		"has", "had", "have",
+		"do", "does", "did",
+		"will", "would", "could", "should",
+		"can", "may", "might", "shall", "must",
+		"don't", "can't", "won't", "shouldn't", "couldn't", "wouldn't",
+		"doesn't", "didn't", "isn't", "aren't", "wasn't", "weren't",
+		"hasn't", "hadn't", "haven't",
+	}
+}
+
 func defaultWeights() map[string]float64 {
 	return map[string]float64{
 		"noun_determiner":   0.35,
diff --git a/reversal/tokeniser_test.go b/reversal/tokeniser_test.go
index 762041d..44fae39 100644
--- a/reversal/tokeniser_test.go
+++ b/reversal/tokeniser_test.go
@@ -694,6 +694,18 @@ func TestTokeniser_Disambiguate_ContractionAux(t *testing.T) {
 	}
 }
 
+func TestTokeniser_Disambiguate_ContractionAux_FallbackDefaults(t *testing.T) {
+	tok := NewTokeniserForLang("zz")
+	tokens := tok.Tokenise("don't run the tests")
+	// The hardcoded fallback auxiliaries should still recognise contractions
+	// even when no locale grammar data is loaded.
+	for _, token := range tokens {
+		if token.Lower == "run" && token.Type != TokenVerb {
+			t.Errorf("'run' after \"don't\": Type = %v, want TokenVerb", token.Type)
+		}
+	}
+}
+
 func TestTokeniser_WithSignals_Breakdown(t *testing.T) {
 	setup(t)
 	tok := NewTokeniser(WithSignals())