Skip to main content

COPY INTO <table>

Introduced or updated: v1.2.666

COPY INTO allows you to load data from files located in one of the following locations:

  • User / Internal / External stages: See What is Stage? to learn about stages in Databend.
  • Buckets or containers created in a storage service.
  • Remote servers from where you can access the files by their URL (starting with "https://...").
  • IPFS.

See also: COPY INTO <location>

Syntax​

COPY INTO [<database_name>.]<table_name>
     FROM { userStage | internalStage | externalStage | externalLocation |
            ( SELECT [<file_col> ... ]
              FROM { userStage | internalStage | externalStage } ) }
[ FILES = ( '<file_name>' [ , '<file_name>' ] [ , ... ] ) ]
[ PATTERN = '<regex_pattern>' ]
[ FILE_FORMAT = (
         FORMAT_NAME = '<your-custom-format>'
         | TYPE = { CSV | TSV | NDJSON | PARQUET | ORC } [ formatTypeOptions ]
       ) ]
[ copyOptions ]

FROM ...​

The FROM clause specifies the source location (user stage, internal stage, external stage, or external location) from which data will be loaded into the specified table using the COPY INTO command. You can also nest a SELECT ... FROM subquery to transform the data you want to load. For more information, see Transforming Data on Load.

note

When you load data from a staged file and the stage path contains special characters such as spaces or parentheses, you can enclose the entire path in single quotes, as demonstrated in the following SQL statements:

COPY INTO mytable FROM 's3://mybucket/dataset(databend)/' ... COPY INTO mytable FROM 's3://mybucket/dataset databend/' ...

userStage​

userStage ::= @~[/<path>]

internalStage​

internalStage ::= @<internal_stage_name>[/<path>]

externalStage​

externalStage ::= @<external_stage_name>[/<path>]

externalLocation​

externalLocation ::=
  's3://<bucket>[<path>]'
  CONNECTION = (
        <connection_parameters>
  )

For the connection parameters available for accessing Amazon S3-like storage services, see Connection Parameters.

FILES​

FILES specifies one or more file names (separated by commas) to be loaded.

PATTERN​

A PCRE2-based regular expression pattern string, enclosed in single quotes, specifying the file names to match. For PCRE2 syntax, see http://www.pcre.org/current/doc/html/pcre2syntax.html. See Example 4: Filtering Files with Pattern for examples and useful tips about filtering files with the PATTERN parameter.

FILE_FORMAT​

See Input & Output File Formats for details.

copyOptions​

copyOptions ::=
  [ SIZE_LIMIT = <num> ]
  [ PURGE = <bool> ]
  [ FORCE = <bool> ]
  [ DISABLE_VARIANT_CHECK = <bool> ]
  [ ON_ERROR = { continue | abort | abort_N } ]
  [ MAX_FILES = <num> ]
  [ RETURN_FAILED_ONLY = <bool> ]
  [ COLUMN_MATCH_MODE =  { case-sensitive | case-insensitive } ]
ParameterDescriptionRequired
SIZE_LIMITSpecifies the maximum rows of data to be loaded for a given COPY statement. Defaults to 0 meaning no limits.Optional
PURGEIf true, the command will purge the files in the stage after they are loaded successfully into the table. Default: false.Optional
FORCECOPY INTO ensures idempotence by automatically tracking and preventing the reloading of files for a default period of 12 hours. This can be customized using the load_file_metadata_expire_hours setting to control the expiration time for file metadata.
This parameter defaults to false meaning COPY INTO will skip duplicate files when copying data. If true, duplicate files will not be skipped.		Optional
DISABLE_VARIANT_CHECKIf true, invalid JSON data is replaced with null values during COPY INTO. If false (default), COPY INTO fails on invalid JSON data.Optional
ON_ERRORDecides how to handle a file that contains errors: continue to skip and proceed, abort (default) to terminate on error, abort_N to terminate when errors ≥ N. Note: abort_N not available for Parquet files.Optional
MAX_FILESSets the maximum number of files to load that have not been loaded already. The value can be set up to 15,000; any value greater than 15,000 will be treated as 15,000.Optional
RETURN_FAILED_ONLYWhen set to true, only files that failed to load will be returned in the output. Default: false.Optional
COLUMN_MATCH_MODE(For Parquet only) Determines if column name matching during COPY INTO is case-sensitive or case-insensitive (default).Optional
tip

When importing large volumes of data, such as logs, it is recommended to set both PURGE and FORCE to true. This ensures efficient data import without the need for interaction with the Meta server (updating the copied-files set). However, it is important to be aware that this may lead to duplicate data imports.

Output​

COPY INTO provides a summary of the data loading results with these columns:

ColumnTypeNullableDescription
FILEVARCHARNOThe relative path to the source file.
ROWS_LOADEDINTNOThe number of rows loaded from the source file.
ERRORS_SEENINTNONumber of error rows in the source file
FIRST_ERRORVARCHARYESThe first error found in the source file.
FIRST_ERROR_LINEINTYESLine number of the first error.

If RETURN_FAILED_ONLY is set to true, the output will only contain the files that failed to load.

Distributed COPY INTO​

The COPY INTO feature in Databend activates distributed execution automatically in cluster environments, enhancing data loading efficiency and scalability.

Examples​

Example 1: Loading from Stages​

These examples showcase data loading into Databend from various types of stages:

COPY INTO mytable
    FROM @~
    PATTERN = '.*[.]parquet'
    FILE_FORMAT = (TYPE = PARQUET);

Example 2: Loading from External Locations​

These examples showcase data loading into Databend from various types of external sources:

This example establishes a connection to Amazon S3 using AWS access keys and secrets, and it loads 10 rows from a CSV file:

-- Authenticated by AWS access keys and secrets.
COPY INTO mytable
    FROM 's3://mybucket/data.csv'
    CONNECTION = (
        ACCESS_KEY_ID = '<your-access-key-ID>',
        SECRET_ACCESS_KEY = '<your-secret-access-key>'
    )
    FILE_FORMAT = (
        TYPE = CSV,
        FIELD_DELIMITER = ',',
        RECORD_DELIMITER = '\n',
        SKIP_HEADER = 1
    )
    SIZE_LIMIT = 10;

This example connects to Amazon S3 using AWS IAM role authentication with an external ID and loads CSV files matching the specified pattern from 'mybucket':

-- Authenticated by AWS IAM role and external ID.
COPY INTO mytable
    FROM 's3://mybucket/'
    CONNECTION = (
        ENDPOINT_URL = 'https://<endpoint-URL>',
        ROLE_ARN = 'arn:aws:iam::123456789012:role/my_iam_role',
        EXTERNAL_ID = '123456'
    )
    PATTERN = '.*[.]csv'
        FILE_FORMAT = (
        TYPE = CSV,
        FIELD_DELIMITER = ',',
        RECORD_DELIMITER = '\n',
        SKIP_HEADER = 1
    );

Example 3: Loading Compressed Data​

This example loads a GZIP-compressed CSV file on Amazon S3 into Databend:

COPY INTO mytable
    FROM 's3://mybucket/data.csv.gz'
    CONNECTION = (
        ENDPOINT_URL = 'https://<endpoint-URL>',
        ACCESS_KEY_ID = '<your-access-key-ID>',
        SECRET_ACCESS_KEY = '<your-secret-access-key>'
    )
    FILE_FORMAT = (
        TYPE = CSV,
        FIELD_DELIMITER = ',',
        RECORD_DELIMITER = '\n',
        SKIP_HEADER = 1,
        COMPRESSION = AUTO
    );

Example 4: Filtering Files with Pattern​

This example demonstrates how to load CSV files from Amazon S3 using pattern matching with the PATTERN parameter. It filters files with 'sales' in their names and '.csv' extensions:

COPY INTO mytable
    FROM 's3://mybucket/'
    PATTERN = '.*sales.*[.]csv'
    FILE_FORMAT = (
        TYPE = CSV,
        FIELD_DELIMITER = ',',
        RECORD_DELIMITER = '\n',
        SKIP_HEADER = 1
    );

Where .* is interpreted as zero or more occurrences of any character. The square brackets escape the period character . that precedes a file extension.

To load from all the CSV files:

COPY INTO mytable
    FROM 's3://mybucket/'
    PATTERN = '.*[.]csv'
    FILE_FORMAT = (
        TYPE = CSV,
        FIELD_DELIMITER = ',',
        RECORD_DELIMITER = '\n',
        SKIP_HEADER = 1
    );

When specifying the pattern for a file path including multiple folders, consider your matching criteria:

  • If you want to match a specific subpath following a prefix, include the prefix in the pattern (e.g., 'multi_page/') and then specify the pattern you want to match within that subpath (e.g., '_page_1').
-- File path: parquet/multi_page/multi_page_1.parquet
COPY INTO ... FROM @data/parquet/ PATTERN = 'multi_page/.*_page_1.*') ...
  • If you want to match any part of the file path that contains the desired pattern, use '.*' before and after the pattern (e.g., '.*multi_page_1.*') to match any occurrences of 'multi_page_1' within the path.
-- File path: parquet/multi_page/multi_page_1.parquet
COPY INTO ... FROM @data/parquet/ PATTERN ='.*multi_page_1.*') ...

Example 5: Loading to Table with Extra Columns​

This section demonstrates data loading into a table with extra columns, using the sample file books.csv:

books.csv
Transaction Processing,Jim Gray,1992
Readings in Database Systems,Michael Stonebraker,2004

Alt text

By default, COPY INTO loads data into a table by matching the order of fields in the file to the corresponding columns in the table. It's essential to ensure that the data aligns correctly between the file and the table. For example,

CREATE TABLE books
(
    title VARCHAR,
    author VARCHAR,
    date VARCHAR
);

COPY INTO books
    FROM 'https://datafuse-1253727613.cos.ap-hongkong.myqcloud.com/data/books.csv'
    FILE_FORMAT = (TYPE = CSV);

If your table has more columns than the file, you can specify the columns into which you want to load data. For example,

CREATE TABLE books_with_language
(
    title VARCHAR,
    language VARCHAR,
    author VARCHAR,
    date VARCHAR
);

COPY INTO books_with_language (title, author, date)
    FROM 'https://datafuse-1253727613.cos.ap-hongkong.myqcloud.com/data/books.csv'
    FILE_FORMAT = (TYPE = CSV);

If your table has more columns than the file, and the additional columns are at the end of the table, you can load data using the FILE_FORMAT option ERROR_ON_COLUMN_COUNT_MISMATCH. This allows you to load data without specifying each column individually. Please note that ERROR_ON_COLUMN_COUNT_MISMATCH currently works for the CSV file format.

CREATE TABLE books_with_extra_columns
(
    title VARCHAR,
    author VARCHAR,
    date VARCHAR,
    language VARCHAR,
    region VARCHAR
);

COPY INTO books_with_extra_columns
    FROM 'https://datafuse-1253727613.cos.ap-hongkong.myqcloud.com/data/books.csv'
    FILE_FORMAT = (TYPE = CSV, ERROR_ON_COLUMN_COUNT_MISMATCH = false);
note

Extra columns in a table can have default values specified by CREATE TABLE or ALTER TABLE COLUMN. If a default value is not explicitly set for an extra column, the default value associated with its data type will be applied. For instance, an integer-type column will default to 0 if no other value is specified.

Example 6: Loading JSON with Custom Format​

This example loads data from a CSV file "data.csv" with the following content:

1,"U00010","{\"carPriceList\":[{\"carTypeId":10,\"distance":5860},{\"carTypeId":11,\"distance\":5861}]}"
2,"U00011","{\"carPriceList\":[{\"carTypeId":12,\"distance":5862},{\"carTypeId":13,\"distance\":5863}]}"

Each line contains three columns of data, with the third column being a string containing JSON data. To load CSV data correctly with JSON fields, we need to set the correct escape character. This example uses the backslash \ as the escape character, as the JSON data contains double quotes ".

Step 1: Create custom file format.​

-- Define a custom CSV file format with the escape character set to backslash \
CREATE FILE FORMAT my_csv_format
    TYPE = CSV
    ESCAPE = '\\';

Step 2: Create target table.​

CREATE TABLE t
  (
     id       INT,
     seq      VARCHAR,
     p_detail VARCHAR
  );

Step 3: Load with custom file format.​

COPY INTO t FROM @t_stage FILES=('data.csv')
FILE_FORMAT=(FORMAT_NAME='my_csv_format');

Example 7: Loading Invalid JSON​

When loading data into a Variant column, Databend automatically checks the data's validity and throws an error in case of any invalid data. For example, if you have a Parquet file named invalid_json_string.parquet in the user stage that contains invalid JSON data, like this:

SELECT *
FROM @~/invalid_json_string.parquet;

┌────────────────────────────────────┐
│        a        │         b        │
├─────────────────┼──────────────────┤
│               5 │ {"k":"v"}        │
│               6 │ [1,              │
└────────────────────────────────────┘

DESC t2;

┌──────────────────────────────────────────────┐
│  Field │   Type  │  Null  │ Default │  Extra │
├────────┼─────────┼────────┼─────────┼────────┤
│ a      │ VARCHAR │ YES    │ NULL    │        │
│ b      │ VARIANT │ YES    │ NULL    │        │
└──────────────────────────────────────────────┘

An error would occur when attempting to load the data into a table:

root@localhost:8000/default>  COPY INTO t2 FROM @~/invalid_json_string.parquet FILE_FORMAT = (TYPE = PARQUET) ON_ERROR = CONTINUE;
error: APIError: ResponseError with 1006: EOF while parsing a value, pos 3 while evaluating function `parse_json('[1,')`

To load without checking the JSON validity, set the option DISABLE_VARIANT_CHECK to true in the COPY INTO statement:

COPY INTO t2 FROM @~/invalid_json_string.parquet
FILE_FORMAT = (TYPE = PARQUET)
DISABLE_VARIANT_CHECK = true
ON_ERROR = CONTINUE;

┌───────────────────────────────────────────────────────────────────────────────────────────────┐
│             File            │ Rows_loaded │ Errors_seen │    First_error   │ First_error_line │
├─────────────────────────────┼─────────────┼─────────────┼──────────────────┼──────────────────┤
│ invalid_json_string.parquet │           2 │           0 │ NULL             │             NULL │
└───────────────────────────────────────────────────────────────────────────────────────────────┘

SELECT * FROM t2;
-- Invalid JSON is stored as null in the Variant column.
┌──────────────────────────────────────┐
│         a        │         b         │
├──────────────────┼───────────────────┤
│ 5                │ {"k":"v"}         │
│ 6                │ null              │
└──────────────────────────────────────┘
Did this page help you?
Yes
No
Join our growing community
GitHub
🌟 8.0K Stars
Slack
Twitter
YouTube
Explore Databend Cloud for FREE
Low-cost
Fast Analytics
Easy Data Ingestion
Elastic Scaling
Try it today