Fix BulkWriter bug for Array type field

Signed-off-by: yhmo <yihua.mo@zilliz.com>
milvus-io · Feb 10, 2025 · 7dfd5d4 · 7dfd5d4
1 parent 768a2dd
commit 7dfd5d4
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 4 deletions.
diff --git a/examples/bulk_import/example_bulkwriter.py b/examples/bulk_import/example_bulkwriter.py
@@ -179,7 +179,7 @@ def local_writer_simple(schema: CollectionSchema, file_type: BulkFileType):
             file_type=file_type,
     ) as local_writer:
         # read data from csv
-        read_sample_data("./data/train_embeddings.csv", local_writer)
+        read_sample_data("./train_embeddings.csv", local_writer)
 
         # append rows
         for i in range(100000):
@@ -208,7 +208,7 @@ def remote_writer_simple(schema: CollectionSchema, file_type: BulkFileType):
             file_type=file_type,
     ) as remote_writer:
         # read data from csv
-        read_sample_data("./data/train_embeddings.csv", remote_writer)
+        read_sample_data("./train_embeddings.csv", remote_writer)
 
         # append rows
         for i in range(10000):

diff --git a/pymilvus/bulk_writer/constants.py b/pymilvus/bulk_writer/constants.py
@@ -67,8 +67,8 @@
     DataType.INT64.name: np.dtype("int64"),
     DataType.FLOAT.name: np.dtype("float32"),
     DataType.DOUBLE.name: np.dtype("float64"),
-    DataType.VARCHAR.name: None,
-    DataType.JSON.name: None,
+    DataType.VARCHAR.name: np.dtype("str"),
+    DataType.JSON.name: np.dtype("str"),  # in numpy/parquet file, json object are stored as string
     DataType.FLOAT_VECTOR.name: np.dtype("float32"),
     DataType.BINARY_VECTOR.name: np.dtype("uint8"),
     DataType.FLOAT16_VECTOR.name: np.dtype("uint8"),