COPY INTO (Transact-SQL)

Applies to: Azure Synapse Analytics

This article explains how to use the COPY statement in Azure Synapse Analytics for loading from external storage accounts. The COPY statement provides the most flexibility for high-throughput data ingestion into Azure Synapse Analytics.

Note

For Warehouse in Microsoft Fabric, visit COPY INTO.

Use COPY for the following capabilities:

• Use lower privileged users to load without needing strict CONTROL permissions on the data warehouse
• Execute a single T-SQL statement without having to create any other database objects
• Properly parse and load CSV files where delimiters (string, field, row) are escaped within string delimited columns
• Specify a finer permission model without exposing storage account keys using Share Access Signatures (SAS)
• Use a different storage account for the ERRORFILE location (REJECTED_ROW_LOCATION)
• Customize default values for each target column and specify source data fields to load into specific target columns
• Specify a custom row terminator, field terminator, and field quote for CSV files
• Use SQL Server Date formats for CSV files
• Specify wildcards and multiple files in the storage location path
• Automatic schema discovery simplifies the process of defining and mapping source data into target tables
• The automatic table creation process automatically creates the tables and works alongside with automatic schema discovery
• Directly load complex data types from Parquet files such as Maps and Lists into string columns, without using other tools to preprocess the data

Note

To load complex data types from Parquet files, automatic table creation must be turned on by using AUTO_CREATE_TABLE.

Visit the following documentation for comprehensive examples and quickstarts using the COPY statement:

• Quickstart: Bulk load data using the COPY statement
• Quickstart: Examples using the COPY statement and its supported authentication methods
• Quickstart: Creating the COPY statement using the rich Synapse Studio UI

Note

Microsoft Entra ID was previously known as Azure Active Directory (Azure AD).

Syntax

COPY INTO [ schema. ] table_name
[ (Column_list) ]
FROM '<external_location>' [ , ...n ]
WITH
 (
 [ FILE_TYPE = { 'CSV' | 'PARQUET' | 'ORC' } ]
 [ , FILE_FORMAT = EXTERNAL FILE FORMAT OBJECT ]
 [ , CREDENTIAL = (AZURE CREDENTIAL) ]
 [ , ERRORFILE = ' [ http(s)://storageaccount/container ] /errorfile_directory [ / ] ] '
 [ , ERRORFILE_CREDENTIAL = (AZURE CREDENTIAL) ]
 [ , MAXERRORS = max_errors ]
 [ , COMPRESSION = { 'Gzip' | 'DefaultCodec' | 'Snappy' } ]
 [ , FIELDQUOTE = 'string_delimiter' ]
 [ , FIELDTERMINATOR =  'field_terminator' ]
 [ , ROWTERMINATOR = 'row_terminator' ]
 [ , FIRSTROW = first_row ]
 [ , DATEFORMAT = 'date_format' ]
 [ , ENCODING = { 'UTF8' | 'UTF16' } ]
 [ , IDENTITY_INSERT = { 'ON' | 'OFF' } ]
 [ , AUTO_CREATE_TABLE = { 'ON' | 'OFF' } ]
)

Arguments

schema_name

Optional if the default schema for the user performing the operation is the schema of the specified table. If schema isn't specified, and the default schema of the user performing the COPY operation is different from the schema of the specified table, COPY is canceled, and an error message is returned.

table_name

The name of the table to COPY data into. The target table can be a temporary or permanent table and must already exist in the database. For automatic schema detection mode, don't provide a column list.

(column_list)

An optional list of one or more columns used to map source data fields to target table columns for loading data.

Don't specify a column_list when AUTO_CREATE_TABLE = 'ON'.

column_list must be enclosed in parentheses and delimited by commas. The column list is of the following format:

[(Column_name [default Default_value] [Field_number] [,...n])]

• Column_name - the name of the column in the target table.
• Default_value - the default value that replaces any NULL value in the input file. Default value applies to all file formats. COPY attempts to load NULL from the input file when a column is omitted from the column list or when there's an empty input file field. Default value precedes the keyword 'default'
• Field_number - the input file field number that is mapped to the target column.
• The field indexing starts at 1.

When a column list isn't specified, COPY maps columns based on the source and target order: Input field 1 goes to target column 1, field 2 goes to column 2, etc.

External locations

Is where the files containing the data is staged. Currently Azure Data Lake Storage (ADLS) Gen2 and Azure Blob Storage are supported:

• External location for Blob Storage: https://<account\>.blob.core.windows.net/<container\>/<path\>
• External location for ADLS Gen2: https://<account\>.dfs.core.windows.net/<container\>/<path\>

Note

The .blob endpoint is available for ADLS Gen2 as well and currently yields the best performance. Use the .blob endpoint when .dfs is not required for your authentication method.

• Account - The storage account name

• Container - The blob container name

• Path - the folder or file path for the data. The location starts from the container. If a folder is specified, COPY retrieves all files from the folder and all its subfolders. COPY ignores hidden folders and doesn't return files that begin with an underline (_) or a period (.) unless explicitly specified in the path. This behavior is the same even when specifying a path with a wildcard.

Wildcards cards can be included in the path where

• Wildcard path name matching is case-sensitive
• Wildcard can be escaped using the backslash character (\)
• Wildcard expansion is applied recursively. For instance, all CSV files under Customer1 (including subdirectories of Customer1) are loaded in the following example: Account/Container/Customer1/*.csv

Note

For best performance, avoid specifying wildcards that would expand over a larger number of files. If possible, list multiple file locations instead of specifying wildcards.

Multiple file locations can only be specified from the same storage account and container via a comma-separated list such as:

• https://<account>.blob.core.windows.net/<container\>/<path\>, https://<account\>.blob.core.windows.net/<container\>/<path\>

FILE_TYPE = { 'CSV' | 'PARQUET' | 'ORC' }

FILE_TYPE specifies the format of the external data.

• CSV: Specifies a comma-separated values file compliant to the RFC 4180 standard.
• PARQUET: Specifies a Parquet format.
• ORC: Specifies an Optimized Row Columnar (ORC) format.

Note

The file type 'Delimited Text' in PolyBase is replaced by the 'CSV' file format where the default comma delimiter can be configured via the FIELDTERMINATOR parameter.

FILE_FORMAT = external_file_format_name

FILE_FORMAT applies to Parquet and ORC files only and specifies the name of the external file format object that stores the file type and compression method for the external data. To create an external file format, use CREATE EXTERNAL FILE FORMAT.

CREDENTIAL (IDENTITY = '', SECRET = '')

CREDENTIAL specifies the authentication mechanism to access the external storage account. Authentication methods are:

	CSV	Parquet	ORC
Azure Blob Storage	SAS/MSI/SERVICE PRINCIPAL/KEY/AAD	SAS/KEY	SAS/KEY
Azure Data Lake Gen2	SAS/MSI/SERVICE PRINCIPAL/KEY/AAD	SAS (blob 1 )/MSI (dfs 2 )/SERVICE PRINCIPAL/KEY/AAD	SAS (blob 1 )/MSI (dfs 2 )/SERVICE PRINCIPAL/KEY/AAD

1: The .blob endpoint (.blob.core.windows.net) in your external location path is required for this authentication method.

2: The .dfs endpoint (.dfs.core.windows.net) in your external location path is required for this authentication method.

Note

• When authenticating using Microsoft Entra ID or to a public storage account, CREDENTIAL does not need to be specified.
• If your storage account is associated with a VNet, you must authenticate using a managed identity.

• Authenticating with Shared Access Signatures (SAS)

  • IDENTITY: A constant with a value of 'Shared Access Signature'
  • SECRET: The shared access signature provides delegated access to resources in your storage account.
  • Minimum permissions required: READ and LIST

• Authenticating with Service Principals

  • IDENTITY: <ClientID>@<OAuth_2.0_Token_EndPoint>
  • SECRET: Microsoft Entra application service principal key
  • Minimum RBAC roles required: Storage blob data contributor, Storage blob data contributor, Storage blob data owner, or Storage blob data reader

• Authenticating with Storage account key

  • IDENTITY: A constant with a value of 'Storage Account Key'
  • SECRET: Storage account key

• Authenticating with Managed Identity (VNet Service Endpoints)

  • IDENTITY: A constant with a value of 'Managed Identity'
  • Minimum RBAC roles required: Storage blob data contributor or Storage blob data owner for the Microsoft Entra registered logical server in Azure. When using a dedicated SQL pool (formerly SQL DW) that is not associated with a Synapse Workspace this RBAC role is not required, but the managed identity requires Access Control List (ACL) permissions on the target objects to enable read access to the source files

• Authenticating with a Microsoft Entra user

  • CREDENTIAL isn't required
  • Minimum RBAC roles required: Storage blob data contributor or Storage blob data owner for the Microsoft Entra user

ERRORFILE = Directory Location

ERRORFILE only applies to CSV. Specifies the directory within the COPY statement where the rejected rows and the corresponding error file should be written. The full path from the storage account can be specified or the path relative to the container can be specified. If the specified path doesn't exist, one is created on your behalf. A child directory is created with the name "_rejectedrows". The "_" character ensures that the directory is escaped for other data processing unless explicitly named in the location parameter.

Note

When a relative path is passed to ERRORFILE, the path is relative to the container path specified in external_location.

Within this directory, there's a folder created based on the time of load submission in the format YearMonthDay -HourMinuteSecond (Ex. 20180330-173205). In this folder, two types of files are written, the reason (Error) file and the data (Row) file each preappending with the queryID, distributionID, and a file guid. Because the data and the reason are in separate files, corresponding files have a matching prefix.

If ERRORFILE has the full path of the storage account defined, then the ERRORFILE_CREDENTIAL is used to connect to that storage. Otherwise, the value mentioned for CREDENTIAL is used. When the same credential that is used for the source data is used for ERRORFILE, restrictions that apply to ERRORFILE_CREDENTIAL also apply

ERRORFILE_CREDENTIAL = (IDENTITY= '', SECRET = '')

ERRORFILE_CREDENTIAL only applies to CSV files. Supported data source and authentication methods are:

• Azure Blob Storage - SAS/SERVICE PRINCIPAL/AAD

• Azure Data Lake Gen2 - SAS/MSI/SERVICE PRINCIPAL/AAD

• Authenticating with Shared Access Signatures (SAS)

  • IDENTITY: A constant with a value of 'Shared Access Signature'
  • SECRET: The shared access signature provides delegated access to resources in your storage account.
  • Minimum permissions required: READ, LIST, WRITE, CREATE, DELETE

• Authenticating with Service Principals

  • IDENTITY: <ClientID>@<OAuth_2.0_Token_EndPoint>
  • SECRET: Microsoft Entra application service principal key
  • Minimum RBAC roles required: Storage blob data contributor or Storage blob data owner

Note

Use the OAuth 2.0 token endpoint V1

• Authenticating with Managed Identity (VNet Service Endpoints)

  • IDENTITY: A constant with a value of 'Managed Identity'
  • Minimum RBAC roles required: Storage blob data contributor or Storage blob data owner for the Microsoft Entra registered SQL Database server

• Authenticating with a Microsoft Entra user

  • CREDENTIAL isn't required
  • Minimum RBAC roles required: Storage blob data contributor or Storage blob data owner for the Microsoft Entra user

Using a storage account key with ERRORFILE_CREDENTIAL is not supported.

Note

If you are using the same storage account for your ERRORFILE and specifying the ERRORFILE path relative to the root of the container, you do not need to specify the ERROR_CREDENTIAL.

MAXERRORS = max_errors

MAXERRORS specifies the maximum number of reject rows allowed in the load before the COPY operation fails. Each row that can't be imported by the COPY operation is ignored and counted as one error. If max_errors isn't specified, the default is 0.

MAXERRORS cannot be used with AUTO_CREATE_TABLE.

When FILE_TYPE is 'PARQUET', exceptions that are caused by data type conversion errors (e.g., Parquet binary to SQL integer) still cause COPY INTO will to fail, ignoring MAXERRORS.

COMPRESSION = { 'DefaultCodec ' | 'Snappy' | 'GZIP' | 'NONE'}

COMPRESSION is optional and specifies the data compression method for the external data.

• CSV supports GZIP
• Parquet supports GZIP and Snappy
• ORC supports DefaultCodec and Snappy.
• Zlib is the default compression for ORC

The COPY command autodetects the compression type based on the file extension when this parameter isn't specified:

• .gz - GZIP
• .snappy – Snappy
• .deflate - DefaultCodec (Parquet and ORC only)

The COPY command requires that gzip files do not contain any trailing garbage to operate normally. The gzip format strictly requires that files be composed of valid members without any additional information before, between, or after them. Any deviation from this format, such as the presence of trailing non-gzip data, will result in the failure of the COPY command. Make sure to verify there's no trailing garbage at the end of gzip files to ensure COPY can successfully process these files.

FIELDQUOTE = 'field_quote'

FIELDQUOTE applies to CSV and specifies a single character that is used as the quote character (string delimiter) in the CSV file. If not specified, the quote character (") is used as the quote character as defined in the RFC 4180 standard. Hexadecimal notation is also supported for FIELDQUOTE. Extended ASCII and multi-byte characters aren't supported with UTF-8 for FIELDQUOTE.

Note

FIELDQUOTE characters are escaped in string columns where there is a presence of a double FIELDQUOTE (delimiter).

FIELDTERMINATOR = 'field_terminator'

FIELDTERMINATOR Only applies to CSV. Specifies the field terminator that is used in the CSV file. The field terminator can be specified using hexadecimal notation. The field terminator can be multi-character. The default field terminator is a (,). Extended ASCII and multi-byte characters aren't supported with UTF-8 for FIELDTERMINATOR.

ROW TERMINATOR = 'row_terminator'

ROW TERMINATOR Only applies to CSV. Specifies the row terminator that is used in the CSV file. The row terminator can be specified using hexadecimal notation. The row terminator can be multi-character. By default, the row terminator is \r\n.

The COPY command prefixes the \r character when specifying \n (newline) resulting in \r\n. To specify only the \n character, use hexadecimal notation (0x0A). When specifying multi-character row terminators in hexadecimal, don't specify 0x between each character.

Extended ASCII and multi-byte characters aren't supported with UTF-8 for ROW TERMINATOR.

FIRSTROW = First_row_int

FIRSTROW applies to CSV and specifies the row number that is read first in all files for the COPY command. Values start from 1, which is the default value. If the value is set to two, the first row in every file (header row) is skipped when the data is loaded. Rows are skipped based on the existence of row terminators.

DATEFORMAT = { 'mdy' | 'dmy' | 'ymd' | 'ydm' | 'myd' | 'dym' }

DATEFORMAT only applies to CSV and specifies the date format of the date mapping to SQL Server date formats. For an overview of all Transact-SQL date and time data types and functions, see Date and Time Data Types and Functions (Transact-SQL). DATEFORMAT within the COPY command takes precedence over DATEFORMAT configured at the session level.

ENCODING = 'UTF8' | 'UTF16'

ENCODING only applies to CSV. Default is UTF8. Specifies the data encoding standard for the files loaded by the COPY command.

IDENTITY_INSERT = 'ON' | 'OFF'

IDENTITY_INSERT specifies whether the identity value or values in the imported data file are to be used for the identity column. If IDENTITY_INSERT is OFF (default), the identity values for this column are verified, but not imported. Azure Synapse Analytics automatically assigns unique values based on the seed and increment values specified during table creation. Note the following behavior with the COPY command:

• If IDENTITY_INSERT is OFF, and table has an identity column
  • A column list must be specified which doesn't map an input field to the identity column.
• If IDENTITY_INSERT is ON, and table has an identity column
  • If a column list is passed, it must map an input field to the identity column.
• Default value isn't supported for the IDENTITY COLUMN in the column list.
• IDENTITY_INSERT can only be set for one table at a time.

AUTO_CREATE_TABLE = { 'ON' | 'OFF' }

AUTO_CREATE_TABLE specifies if the table could be automatically created by working alongside with automatic schema discovery. It's available only for Parquet files.

• ON: Enables automatic table creation. The COPY INTO process creates a new table automatically by discovering the structure of the file to be loaded. Can also be used with preexisting tables to take advantage of automatic schema discovery of Parquet files.
• OFF: Automatic table creation isn't enabled. Default.

Note

The automatic table creation works alongside with automatic schema discovery. The automatic table creation is NOT enabled by default.

Don't load into hash distributed tables from Parquet files using COPY INTO with AUTO_CREATE_TABLE = 'ON'.

If Parquet files are to be loaded into hash distributed tables using COPY INTO, load it into a round robin staging table followed by INSERT ... SELECT from that table to the target hash distributed table.

Permissions

The user executing the COPY command must have the following permissions:

• ADMINISTER DATABASE BULK OPERATIONS
• INSERT

Requires INSERT and ADMINISTER BULK OPERATIONS permissions. In Azure Synapse Analytics, INSERT, and ADMINISTER DATABASE BULK OPERATIONS permissions are required.

Additionally, if the user executing the COPY command also intends to generate a new table and load data into it, they require CREATE TABLE and ALTER ON SCHEMA permissions.

For example, to allow mike@contoso.com to use COPY to create a new table in the HR schema, and insert the data from a Parquet file, use the following Transact-SQL sample:

GRANT ADMINISTER DATABASE BULK OPERATIONS to [mike@contoso.com];
GRANT INSERT to [mike@contoso.com];

GRANT CREATE TABLE to [mike@contoso.com];
GRANT ALTER on SCHEMA::HR to [mike@contoso.com];

Remarks

The COPY statement accepts only UTF-8 and UTF-16 valid characters for row data and command parameters. Source files or parameters (such as ROW TERMINATOR or FIELD TERMINATOR) that use invalid characters may be interpreted incorrectly by the COPY statement and cause unexpected results such as data corruption, or other failures. Make sure your source files and parameters are UTF-8 or UTF-16 compliant before you invoke the COPY statement.

Examples

A. Load from a public storage account

The following example is the simplest form of the COPY command, which loads data from a public storage account. For this example, the COPY statement's defaults match the format of the line item csv file.

COPY INTO dbo.[lineitem]
FROM 'https://unsecureaccount.blob.core.windows.net/customerdatasets/folder1/lineitem.csv'
WITH (FIELDTERMINATOR = '|')

The default values of the COPY command are:

• DATEFORMAT = Session DATEFORMAT

• MAXERRORS = 0

• COMPRESSION default is uncompressed

• FIELDQUOTE = '"'

• FIELDTERMINATOR = ','

• ROWTERMINATOR = '\n'

Important

COPY treats \n as \r\n internally. For more information, see the ROWTERMINATOR section.

• FIRSTROW = 1

• ENCODING = 'UTF8'

• FILE_TYPE = 'CSV'

• IDENTITY_INSERT = 'OFF'

B. Load authenticating via Share Access Signature (SAS)

The following example loads files that use the line feed as a row terminator such as a UNIX output. This example also uses a SAS key to authenticate to Azure Blob Storage.

COPY INTO test_1
FROM 'https://myaccount.blob.core.windows.net/myblobcontainer/folder1/'
WITH (
    FILE_TYPE = 'CSV',
    CREDENTIAL=(IDENTITY= 'Shared Access Signature', SECRET='<Your_SAS_Token>'),
    --CREDENTIAL should look something like this:
    --CREDENTIAL=(IDENTITY= 'Shared Access Signature', SECRET='?sv=2018-03-28&ss=bfqt&srt=sco&sp=rl&st=2016-10-17T20%3A14%3A55Z&se=2021-10-18T20%3A19%3A00Z&sig=IEoOdmeYnE9%2FKiJDSHFSYsz4AkNa%2F%2BTx61FuQ%2FfKHefqoBE%3D'),
    FIELDQUOTE = '"',
    FIELDTERMINATOR=';',
    ROWTERMINATOR='0X0A',
    ENCODING = 'UTF8',
    DATEFORMAT = 'ymd',
    MAXERRORS = 10,
    ERRORFILE = '/errorsfolder',--path starting from the storage container
    IDENTITY_INSERT = 'ON'
)

C. Load with a column list with default values authenticating via Storage Account Key

This example loads files specifying a column list with default values.

--Note when specifying the column list, input field numbers start from 1
COPY INTO test_1 (Col_one default 'myStringDefault' 1, Col_two default 1 3)
FROM 'https://myaccount.blob.core.windows.net/myblobcontainer/folder1/'
WITH (
    FILE_TYPE = 'CSV',
    CREDENTIAL=(IDENTITY= 'Storage Account Key', SECRET='<Your_Account_Key>'),
    --CREDENTIAL should look something like this:
    --CREDENTIAL=(IDENTITY= 'Storage Account Key', SECRET='x6RWv4It5F2msnjelv3H4DA80n0PQW0daPdw43jM0nyetx4c6CpDkdj3986DX5AHFMIf/YN4y6kkCnU8lb+Wx0Pj+6MDw=='),
    FIELDQUOTE = '"',
    FIELDTERMINATOR=',',
    ROWTERMINATOR='0x0A',
    ENCODING = 'UTF8',
    FIRSTROW = 2
)

D. Load Parquet or ORC using existing file format object

This example uses a wildcard to load all Parquet files under a folder.

COPY INTO test_parquet
FROM 'https://myaccount.blob.core.windows.net/myblobcontainer/folder1/*.parquet'
WITH (
    FILE_FORMAT = myFileFormat,
    CREDENTIAL=(IDENTITY= 'Shared Access Signature', SECRET='<Your_SAS_Token>')
)

E. Load specifying wild cards and multiple files

COPY INTO t1
FROM
'https://myaccount.blob.core.windows.net/myblobcontainer/folder0/*.txt',
    'https://myaccount.blob.core.windows.net/myblobcontainer/folder1'
WITH (
    FILE_TYPE = 'CSV',
    CREDENTIAL=(IDENTITY= '<client_id>@<OAuth_2.0_Token_EndPoint>',SECRET='<key>'),
    FIELDTERMINATOR = '|'
)

F. Load using MSI credentials

COPY INTO dbo.myCOPYDemoTable
FROM 'https://myaccount.blob.core.windows.net/myblobcontainer/folder0/*.txt'
WITH (
    FILE_TYPE = 'CSV',
    CREDENTIAL = (IDENTITY = 'Managed Identity'),
    FIELDQUOTE = '"',
    FIELDTERMINATOR=','
)

G. Load using automatic schema detection

COPY INTO [myCOPYDemoTable]
FROM 'https://myaccount.blob.core.windows.net/customerdatasets/folder1/lineitem.parquet'
WITH (
    FILE_TYPE = 'Parquet',
    CREDENTIAL = ( IDENTITY = 'Shared Access Signature',  SECRET='<key>'),
    AUTO_CREATE_TABLE = 'ON'
)

FAQ

What is the performance of the COPY command compared to PolyBase?

The COPY command has better performance depending on your workload.

• Compressed files can't be split automatically. For best loading performance, consider splitting your input into multiple files when loading compressed CSVs.

• Large uncompressed CSV files can be split and loaded in parallel automatically, so there's no need to manually split uncompressed CSV files in most cases. In certain cases where auto file splitting isn't feasible due to data characteristics, manually splitting large CSVs may still benefit performance.

What is the file splitting guidance for the COPY command loading compressed CSV files?

Guidance on the number of files is outlined in the following table. Once the recommended number of files are reached, you have better performance the larger the files. The number of files is determined by number of compute nodes multiplied by 60. For example, at 6000DWU we have 12 compute nodes and 12*60 = 720 partitions. For a simple file splitting experience, refer to How to maximize COPY load throughput with file splits.

DWU	#Files
100	60
200	60
300	60
400	60
500	60
1,000	120
1,500	180
2,000	240
2,500	300
3,000	360
5,000	600
6,000	720
7,500	900
10,000	1200
15,000	1800
30,000	3600

What is the file splitting guidance for the COPY command loading Parquet or ORC files?

There's no need to split Parquet and ORC files because the COPY command automatically splits files. Parquet and ORC files in the Azure storage account should be 256 MB or larger for best performance.

Are there any limitations on the number or size of files?

There are no limitations on the number or size of files; however, for best performance, we recommend files that are at least 4 MB.

Are there any known issues with the COPY statement?

If you have an Azure Synapse workspace that was created prior to December 7, 2020, you may run into a similar error message when authenticating using Managed Identity: com.microsoft.sqlserver.jdbc.SQLServerException: Managed Service Identity has not been enabled on this server. Please enable Managed Service Identity and try again.

Follow these steps to work around this issue by re-registering the workspace's managed identity:

1. Install Azure PowerShell. Refer to Install PowerShell.
2. Register your workspace's managed identity using PowerShell:

   Connect-AzAccount
   Select-AzSubscription -SubscriptionId <subscriptionId>
   Set-AzSqlServer -ResourceGroupName your-database-server-resourceGroup -ServerName your-SQL-servername -AssignIdentity
   

Related content

• Loading overview with Azure Synapse Analytics