From e46e9757ccc603dd56d91e71745f68593595f200 Mon Sep 17 00:00:00 2001
From: Shehroz Riaz <zorhehs@gmail.com>
Date: Fri, 21 Aug 2026 14:34:20 +0500
Subject: [PATCH] fix: classify large modifications as feat instead of fix

A 1-line typo fix and a 200+ line rewrite were both landing on 'fix:'
because inferType() only looked at file status (A/M/D), never diff
size. Modified files whose total changed lines (additions + deletions)
reach largeChangeThreshold (50) now classify as feat.

Found via dogfooding on a real 224-line change (src/app.py in
ai-research-summarizer) that was incorrectly labeled fix.
---
 internal/generator/heuristic/heuristic.go     | 21 ++++++++++----
 .../generator/heuristic/heuristic_test.go     | 28 +++++++++++++++++++
 2 files changed, 44 insertions(+), 5 deletions(-)

diff --git a/internal/generator/heuristic/heuristic.go b/internal/generator/heuristic/heuristic.go
index 622ab44..cad852c 100644
--- a/internal/generator/heuristic/heuristic.go
+++ b/internal/generator/heuristic/heuristic.go
@@ -24,6 +24,12 @@ func New() *Generator {
 
 func (g *Generator) Name() string { return "heuristic" }
 
+// largeChangeThreshold is the total number of changed lines (additions +
+// deletions) across modified-only files above which we treat the change
+// as substantial enough to be "feat" rather than "fix". A one-line typo
+// fix and a 200-line rewrite should not both be labeled "fix".
+const largeChangeThreshold = 50
+
 // Generate drafts a Conventional Commits message from the staged diff
 // using only path/status heuristics — no network, no LLM.
 func (g *Generator) Generate(diff *gitutil.StagedDiff) (generator.Message, error) {
@@ -43,11 +49,12 @@ func (g *Generator) Generate(diff *gitutil.StagedDiff) (generator.Message, error
 // files. Rules are checked in priority order; the first match wins.
 func inferType(files []gitutil.FileChange) string {
 	var (
-		allTest    = true
-		allDocs    = true
-		hasNewFile = false
-		hasDeleted = false
-		hasCI      = false
+		allTest            = true
+		allDocs            = true
+		hasNewFile         = false
+		hasDeleted         = false
+		hasCI              = false
+		totalModifiedLines = 0
 	)
 
 	for _, f := range files {
@@ -68,6 +75,8 @@ func inferType(files []gitutil.FileChange) string {
 			hasNewFile = true
 		case "D":
 			hasDeleted = true
+		case "M":
+			totalModifiedLines += f.Additions + f.Deletions
 		}
 	}
 
@@ -82,6 +91,8 @@ func inferType(files []gitutil.FileChange) string {
 		return "chore"
 	case hasNewFile:
 		return "feat"
+	case totalModifiedLines >= largeChangeThreshold:
+		return "feat"
 	default:
 		return "fix"
 	}
diff --git a/internal/generator/heuristic/heuristic_test.go b/internal/generator/heuristic/heuristic_test.go
index 6a0b521..6dc1358 100644
--- a/internal/generator/heuristic/heuristic_test.go
+++ b/internal/generator/heuristic/heuristic_test.go
@@ -72,6 +72,34 @@ func TestGenerate(t *testing.T) {
 			},
 			wantPrefix: "feat(a): add 1, update 1, remove 1 files",
 		},
+		{
+			name: "small modification stays fix",
+			files: []gitutil.FileChange{
+				{Path: "src/app.py", Status: "M", Additions: 10, Deletions: 5},
+			},
+			wantPrefix: "fix",
+		},
+		{
+			name: "large modification becomes feat",
+			files: []gitutil.FileChange{
+				{Path: "src/app.py", Status: "M", Additions: 200, Deletions: 24},
+			},
+			wantPrefix: "feat(src): update src/app.py",
+		},
+		{
+			name: "modification exactly at threshold becomes feat",
+			files: []gitutil.FileChange{
+				{Path: "src/app.py", Status: "M", Additions: 40, Deletions: 10},
+			},
+			wantPrefix: "feat",
+		},
+		{
+			name: "modification just under threshold stays fix",
+			files: []gitutil.FileChange{
+				{Path: "src/app.py", Status: "M", Additions: 30, Deletions: 19},
+			},
+			wantPrefix: "fix",
+		},
 	}
 
 	g := New()