task-51: accepted 行序列化与 hash 批量生成

CollectDataResultDetailCodec.encodeRefsWithHash 一次迭代为整 chunk 生成
全部行引用 JSON 及其 SHA-256（offset 从 startOffset 递增），带
MAX_REFS_PER_BATCH 上限与参数校验；CollectDataResultItemBatchWriter
接入批量生成，删除逐行 encodeRef+hash 循环，同 chunk 重提结果逐字节一致。

backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/util/CollectDataResultDetailCodec.java

@@ -6,6 +6,9 @@ import com.fasterxml.jackson.databind.ObjectMapper;
 import com.nanri.aiimage.modules.collectdata.model.vo.CollectDataResultRowVo;
 import org.springframework.stereotype.Component;
 
+import java.security.MessageDigest;
+import java.nio.charset.StandardCharsets;
+import java.util.ArrayList;
 import java.util.List;
 
 /**
@@ -26,6 +29,13 @@ public class CollectDataResultDetailCodec {
     public static final String REF_FIELD_OFFSET = "offset";
     public static final String REF_FIELD_PAYLOAD = "payload";
 
+    /** 单次批量引用生成的上限：与结果明细 batch size 同量级，拒绝前不分配结果。 */
+    public static final int MAX_REFS_PER_BATCH = 100_000;
+
+    /** 一批引用及各自 hash：refJson 为行引用 JSON，payloadHash 为其 SHA-256。 */
+    public record RefWithHash(String refJson, String payloadHash) {
+    }
+
     private final ObjectMapper objectMapper;
 
     public CollectDataResultDetailCodec(ObjectMapper objectMapper) {
@@ -101,9 +111,50 @@ public class CollectDataResultDetailCodec {
         }
     }
 
+    /**
+     * 批量生成整 chunk 的引用 JSON 及其 SHA-256 hash：offset 从 startOffset
+     * 递增共 count 行，一次迭代完成序列化 + hash，替代调用方逐行两轮循环。
+     * 非法参数（负 chunk/负 startOffset/负 count/null 或空白指针）抛
+     * IllegalArgumentException；count 超过 MAX_REFS_PER_BATCH 时在分配
+     * 结果前拒绝，不发生无界内存增长。
+     */
+    public List<RefWithHash> encodeRefsWithHash(int chunkIndex, int startOffset, int count, String pointer) {
+        if (chunkIndex < 0 || startOffset < 0 || count < 0 || pointer == null || pointer.isBlank()) {
+            throw new IllegalArgumentException("invalid chunk detail ref batch chunk=" + chunkIndex
+                    + " startOffset=" + startOffset + " count=" + count + " pointer=" + pointer);
+        }
+        if (count > MAX_REFS_PER_BATCH) {
+            throw new IllegalArgumentException("chunk detail ref batch too large count=" + count
+                    + " max=" + MAX_REFS_PER_BATCH);
+        }
+        if (count == 0) {
+            return List.of();
+        }
+        List<RefWithHash> refs = new ArrayList<>(count);
+        for (int offset = startOffset; offset < startOffset + count; offset++) {
+            String refJson = encodeRef(chunkIndex, offset, pointer);
+            refs.add(new RefWithHash(refJson, sha256(refJson)));
+        }
+        return refs;
+    }
+
     /** 结果明细行引用：chunk 序号 + 行内 offset + chunk 明细对象指针。 */
     public record ChunkRef(@JsonProperty(REF_FIELD_CHUNK) int chunkIndex,
                            @JsonProperty(REF_FIELD_OFFSET) int offset,
                            @JsonProperty(REF_FIELD_PAYLOAD) String pointer) {
     }
+
+    private static String sha256(String value) {
+        try {
+            MessageDigest digest = MessageDigest.getInstance("SHA-256");
+            byte[] bytes = digest.digest((value == null ? "" : value).getBytes(StandardCharsets.UTF_8));
+            StringBuilder sb = new StringBuilder(bytes.length * 2);
+            for (byte b : bytes) {
+                sb.append(String.format("%02x", b));
+            }
+            return sb.toString();
+        } catch (Exception ex) {
+            throw new IllegalStateException("chunk detail ref hash failed", ex);
+        }
+    }
 }

backend-java/src/main/java/com/nanri/aiimage/modules/collectdata/util/CollectDataResultItemBatchWriter.java

@@ -55,7 +55,7 @@ public class CollectDataResultItemBatchWriter {
         if (rows == null || rows.isEmpty()) {
             return new UpsertCounts(0, 0);
         }
-        String scopeHash = hash(scopeKey);
+        String scopeHash = sha256(scopeKey);
         // 一次性取回本 scope 现有行，构建 item_key → 现有行 映射（hash 相等即跳过）。
         List<TaskResultItemEntity> existingList = taskResultItemMapper.selectList(
                 new LambdaQueryWrapper<TaskResultItemEntity>()
@@ -72,6 +72,9 @@ public class CollectDataResultItemBatchWriter {
         List<TaskResultItemEntity> toUpsert = new ArrayList<>();
         int skipped = 0;
         LocalDateTime now = LocalDateTime.now();
+        // 批量生成整 chunk 的引用 JSON + hash（一次迭代），替代逐行 encodeRef + hash。
+        List<CollectDataResultDetailCodec.RefWithHash> refsWithHash =
+                resultDetailCodec.encodeRefsWithHash(chunkIndex, 0, rows.size(), storedDetail);
         for (int offset = 0; offset < rows.size(); offset++) {
             CollectDataResultRowVo row = rows.get(offset);
             if (row == null || row.getAsin() == null || row.getAsin().isBlank()) {
@@ -80,8 +83,8 @@ public class CollectDataResultItemBatchWriter {
             String itemKey = "asin:" + row.getAsin();
             // chunk 级引用共享同一 RustFS 对象（deterministic key，同 chunk 重提
             // 覆盖同一对象，引用 pointer 稳定，无需删除）。
-            String refJson = resultDetailCodec.encodeRef(chunkIndex, offset, storedDetail);
-            String payloadHash = hash(refJson);
+            String refJson = refsWithHash.get(offset).refJson();
+            String payloadHash = refsWithHash.get(offset).payloadHash();
             TaskResultItemEntity existing = existingByKey.get(itemKey);
             if (existing != null && Objects.equals(existing.getPayloadHash(), payloadHash)) {
                 skipped++;
@@ -124,7 +127,7 @@ public class CollectDataResultItemBatchWriter {
         return new UpsertCounts(written, skipped);
     }
 
-    private String hash(String value) {
+    private static String sha256(String value) {
         try {
             java.security.MessageDigest digest = java.security.MessageDigest.getInstance("SHA-256");
             byte[] bytes = digest.digest((value == null ? "" : value).getBytes(java.nio.charset.StandardCharsets.UTF_8));