Phase 3: PDF/DOCX extraction, chunking, LLM client with mock interface

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

internal/parse/chunk_test.go

@@ -0,0 +1,52 @@
+package parse_test
+
+import (
+	"strings"
+	"testing"
+
+	"qbank/internal/parse"
+)
+
+func TestChunk(t *testing.T) {
+	t.Run("small text stays in one chunk", func(t *testing.T) {
+		text := "Para one.\n\nPara two.\n\nPara three."
+		chunks := parse.Chunk(text, 1000)
+		if len(chunks) != 1 {
+			t.Errorf("want 1 chunk, got %d: %v", len(chunks), chunks)
+		}
+		if !strings.Contains(chunks[0], "Para one") || !strings.Contains(chunks[0], "Para three") {
+			t.Errorf("content lost: %q", chunks[0])
+		}
+	})
+
+	t.Run("paragraphs split when over limit", func(t *testing.T) {
+		para := strings.Repeat("x", 600)
+		text := para + "\n\n" + para + "\n\n" + para
+		chunks := parse.Chunk(text, 1000)
+		if len(chunks) < 2 {
+			t.Errorf("want ≥2 chunks for 1800-rune input with 1000 limit, got %d", len(chunks))
+		}
+		// No chunk should combine paragraphs past the limit
+		for i, c := range chunks {
+			if len([]rune(c)) > 1200 {
+				t.Errorf("chunk %d is %d runes, too large", i, len([]rune(c)))
+			}
+		}
+	})
+
+	t.Run("single oversized paragraph kept as own chunk", func(t *testing.T) {
+		bigPara := strings.Repeat("x", 2000)
+		chunks := parse.Chunk(bigPara, 1000)
+		if len(chunks) != 1 {
+			t.Errorf("want 1 chunk for single oversized para, got %d", len(chunks))
+		}
+	})
+
+	t.Run("empty paragraphs ignored", func(t *testing.T) {
+		text := "\n\nPara one.\n\n\n\nPara two.\n\n"
+		chunks := parse.Chunk(text, 1000)
+		if len(chunks) != 1 {
+			t.Errorf("want 1 chunk after ignoring blanks, got %d", len(chunks))
+		}
+	})
+}