Filter out null keys and values from document metadata for TextSplitter

- Update TextSplitter to filter out metadata entries with null keys or values
- Simplify text substring call in TextSplitterTests
- Add test case for document splitting with null metadata values

Author: ogbozoyan

spring-ai-core/src/main/java/org/springframework/ai/transformer/splitter/TextSplitter.java

@@ -88,7 +88,8 @@ private List<Document> createDocuments(List<String> texts, List<ContentFormatter
 				// only primitive values are in here -
 				Map<String, Object> metadataCopy = metadata.entrySet()
 					.stream()
-					.collect(Collectors.toMap(e -> e.getKey(), e -> e.getValue()));
+					.filter(e -> e.getKey() != null && e.getValue() != null)
+					.collect(Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue));
 				Document newDoc = new Document(chunk, metadataCopy);
 
 				if (this.copyContentFormatter) {

spring-ai-core/src/test/java/org/springframework/ai/transformer/splitter/TextSplitterTests.java

@@ -42,7 +42,7 @@ protected List<String> splitText(String text) {
 			List<String> chunks = new ArrayList<>();
 
 			chunks.add(text.substring(0, chuckSize));
-			chunks.add(text.substring(chuckSize, text.length()));
+			chunks.add(text.substring(chuckSize));
 
 			return chunks;
 		}
@@ -213,4 +213,35 @@ public void pageWithChunkSplit() {
 				() -> assertThat(splitedDocument.get(3).getMetadata().get("page_number")).isEqualTo(3));
 	}
 
+	@Test
+	public void testSplitTextWithNullMetadata() {
+
+		var contentFormatter = DefaultContentFormatter.defaultConfig();
+
+		var doc = new Document("In the end, writing arises when man realizes that memory is not enough.");
+
+		doc.getMetadata().put("key1", "value1");
+		doc.getMetadata().put("key2", null);
+
+		doc.setContentFormatter(contentFormatter);
+
+		List<Document> chunks = testTextSplitter.apply(List.of(doc));
+
+		assertThat(testTextSplitter.isCopyContentFormatter()).isTrue();
+
+		assertThat(chunks).hasSize(2);
+
+		// Doc chunks:
+		assertThat(chunks.get(0).getContent()).isEqualTo("In the end, writing arises when man");
+		assertThat(chunks.get(1).getContent()).isEqualTo(" realizes that memory is not enough.");
+
+		// Verify that the same, merged metadata is copied to all chunks.
+		assertThat(chunks.get(0).getMetadata()).isEqualTo(chunks.get(1).getMetadata());
+		assertThat(chunks.get(1).getMetadata()).containsKeys("key1");
+
+		// Verify that the content formatters are copied from the parents to the chunks.
+		assertThat(chunks.get(0).getContentFormatter()).isSameAs(contentFormatter);
+		assertThat(chunks.get(1).getContentFormatter()).isSameAs(contentFormatter);
+	}
+
 }