register ( snowflakeQuerySchema, 'featureSet2', primaryKey = 'keyName' ) // Ingest to cache myFeatureSet1. register ( snowflakeTableSchema, 'featureSet1', primaryKey = 'keyName' ) val myFeatureSet2 = project. extractSchemaFromSource ( snowflakeQuery ) // Register the feature set val myFeatureSet1 = project. extractSchemaFromSource ( snowflakeTable ) // Create a snowflake query source val snowflakeQuery = SnowflakeTable ( '', 'warehouse name', 'database name', 'schema name', query = '' ) val snowflakeQuerySchema = client. create ( 'demo' ) // Create a snowflake table source val proxy = new Proxy ( 'url', 'port', 'username', 'password' ) val snowflakeTable = SnowflakeTable ( '', 'warehouse name', 'database name', 'schema name', 'table name', insecure = false, proxy = proxy, role = 'role', account = 'account' ) val snowflakeTableSchema = client. login () // Set project specifics val project = client. Initialise feature store client val client = Client ( 'url' ) client. In terms of performance between query JSON and query relational table, they may produce similar performance results in some cases, in other cases, query performance against JSON. ingest ( snowflake_query ) # Retrieve feature set ref = my_feature_set_1. Snowflake optimized query performance on JSON by extracting as many JSON elements as columnar form, which means those extracted elements’ metadata will be collected. ingest ( snowflake_table ) my_feature_set_2. register ( snowflake_query_schema, 'feature_set_2', primary_key = 'key_name' ) # Ingest to cache my_feature_set_1. register ( snowflake_table_schema, 'feature_set_1', primary_key = 'key_name' ) my_feature_set_2 = project. extract_schema_from_source ( snowflake_query ) # Register the feature set my_feature_set_1 = project. extract_schema_from_source ( snowflake_table ) # Create a snowflake query source snowflake_query = SnowflakeTable ( '', 'warehouse name', 'database name', 'schema name', query = '' ) snowflake_query_schema = client. create ( 'demo' ) # Create a snowflake table source proxy = Proxy ( 'url', 'port', 'username', 'password' ) snowflake_table = SnowflakeTable ( '', 'warehouse name', 'database name', 'schema name', 'table name', insecure = False, proxy = proxy, role = 'role', account = 'account name' ) snowflake_table_schema = client. login () # Set project specifics project = client. From featurestore import Client, SnowflakeTable # Initialise feature store client client = Client ( 'ip:port' ) client. There is a work around you can use the “enabling Staging” option and that should do the trick. O filePattern in copy activity sink is left as default or set to setOfObjects.“* For JSON format, direct copy only supports the case that source Snowflake table or query result only has single column and the data type of this column is VARIANT, OBJECT, or ARRAY.O quoteChar is double quote, single quote, or empty string (no quote char).

O encodingName is left as default or set to utf-8. O compression can be no compression, gzip, bzip2, or deflate. O rowDelimiter is , or any single character.

For Parquet format, the compression codec is None, Snappy, or Lzo.

*“ The sink data format is of Parquet, delimited text, or JSON with the following configurations: This is due to the fact that at this time the SF connector supports only DelimitedText, Parquet or JSON. The error which are getting is belowĭirect copying data from Snowflake is only supported when sink dataset is DelimitedText, Parquet or JSON with Azure Blob Storage linked service, for other dataset or linked service, please enable staging.

Snowflake json query