CREATE EXTERNAL TABLE

Creates an external table. External tables serve as connectors to your external data sources. External tables contain no data within Firebolt other than metadata virtual columns that are automatically populated with metadata. For more information, see Working with external tables. Data that you ingest must be in an Amazon S3 bucket in the same AWS Region as the Firebolt database.

• PARTITION
• CREDENTIALS
  • Syntax–authenticating using an access key and secret
  • Syntax–authenticating using an IAM role
• URL and OBJECT_PATTERN
• TYPE
  • CSV Types
  • JSON Types
  • Other Types
• COMPRESSION

Syntax

CREATE EXTERNAL TABLE [IF NOT EXISTS] <table>
(
    <column_name> <column_type>[ PARTITION('<regex>')]
    [, <column_name2> <column_type2> [PARTITION('<regex>')]]
    [,...<column_name2> <column_type2> [PARTITION('<regex>')]]
)
[CREDENTIALS = (<awsCredentials>)]
URL = 's3://<bucket_name>[/<folder>][/...]/'
OBJECT_PATTERN = '<object_pattern>'
TYPE = (<type>)
[ <type option> ]
[COMPRESSION = <compression_type>]

Parameters

PARTITION

In some applications, such as Hive partitioning, table partitions are stored in S3 folders and files using a folder naming convention that identifies the partition. The PARTITION keyword allows you to specify a regular expression, <regex>, to extract a portion of the file path and store it in the specified column when Firebolt uses the external table to ingest partitioned data.

Using PARTITION in this way is one method of extracting partition data from file paths. Another method is to use the table metadata column, source_file_name, during the INSERT operation. For more information, see Example–extracting partition values using INSERT

Guidelines for creating the regex

• The regular expression is matched against the object prefix, not including the s3://<bucket_name>/ portion of the prefix.
• You can specify as many PARTITION columns as you need, each extracting a different portion of the object prefix.
• For each PARTITION column, you must specify a regular expression that contains a capturing group, which determines the column value.
• When <column_type> is DATE, Firebolt requires three capturing groups that must be in the order of year, month, and day.
• Firebolt tries to convert the captured string to the specified <column_type>. If the type conversion fails, a NULL is entered.

In most cases, the easiest way to build a regular expression is as follows:

1. Count the number of folders in the path, not including the bucket name.
2. Concatenate the string [^\/]+\/ according to the number of folders.
3. Prefix the regex with an additional [^\/]+ for the file name.
4. Wrap the [^\/]+ in the right folder with a capturing group parenthesis, such as ([^\/]+).

For more information, see Match groups on the RegexOne website. To test your regular expressions, online tools such as regex101 are available.

Example–extract Hive-compatible partitions

The example below demonstrates a CREATE EXTERNAL TABLE statement that creates the table my_ext_table. This table is used to ingest all files with a *.parquet file extension in any sub-folder of the S3 bucket s3://my_bucket.

Consider an example where folders and files in the bucket have the following consistent pattern, which is common for Hive partitions:

s3://my_bucket/c_type=xyz/year=2018/month=01/part-00001.parquet
s3://my_bucket/c_type=xyz/year=2018/month=01/part-00002.parquet
s3://my_bucket/c_type=abc/year=2018/month=01/part-00001.parquet
s3://my_bucket/c_type=abc/year=2018/month=01/part-00002.parquet
[...]

In the example CREATE EXTERNAL TABLE statement below, the PARTITION keyword in the column definition for c_type specifies a regular expression. This expression extracts the portion of the S3 path name that correspond to the xyz or abc within c_type=xyz or c_type=abc.

CREATE EXTERNAL TABLE my_ext_table (
  c_id    INTEGER,
  c_name  TEXT,
  c_type  TEXT PARTITION('[^\/]+\/c_type=([^\/]+)\/[^\/]+\/[^\/]+')
)
CREDENTIALS = (AWS_ROLE_ARN = 'arn:aws:iam::123456789012:role/MyRoleForFireboltS3Access1')
URL = 's3://my_bucket/'
OBJECT_PATTERN= '*.parquet'
TYPE = (PARQUET)

When Firebolt ingests the data from a Parquet file stored in that path, the c_type column for each row contains the extracted portion of the path. For the files listed above, the extraction results in the following values. c_id and c_name are values stored within the respective Parquet files, while c_type are values extracted from the file path.

CREDENTIALS

The credentials for accessing your AWS S3. Firebolt enables using either access key & secret or IAM role.

Syntax–authenticating using an access key and secret

CREDENTIALS = (AWS_KEY_ID = '<ID>' AWS_SECRET_KEY = '<secret>' )

Parameters

In case you don’t have the access key and secret to access your S3 bucket, read more here on how to obtain them.

Syntax–authenticating using an IAM role

Read more on how to configure the AWS role here.

CREDENTIALS = (AWS_ROLE_ARN = '<arn' [AWS_ROLE_EXTERNAL_ID = '<external_ID'])

Parameters

URL and OBJECT_PATTERN

TheURLandOBJECT_PATTERNparameters are used together to match the set of files from within the specified bucket that you wish to include as the data for the external table. The S3 bucket that you reference must be in the same AWS Region as the Firebolt database.

Syntax

URL = 's3://<bucket>[/<folder>][/...]/'
OBJECT_PATTERN = '<object_pattern>'

Parameters

The following wildcards are supported:

• '*' matches any sequence of characters
• '?' matches any single character
• [SET] matches any single character in the specified set
• [!SET] matches any character, not in the specified set.

Example

In the following layout of objects in a bucket, the data is partitioned according to client type, year, and month, with multiple parquet files in each partition. The examples demonstrate how choosing both URL and OBJECT_PATTERN impacts the objects that are retrieved from S3.

s3://bucket/c_type=xyz/year=2018/month=01/part-00001.parquet
s3://bucket/c_type=xyz/year=2018/month=01/part-00002.parquet
...
s3://bucket/c_type=xyz/year=2018/month=12/part-00001.parquet
s3://bucket/c_type=xyz/year=2018/month=12/part-00002.parquet
...
s3://bucket/c_type=xyz/year=2019/month=01/part-00001.parquet
s3://bucket/c_type=xyz/year=2019/month=01/part-00002.parquet
...
s3://bucket/c_type=xyz/year=2020/month=01/part-00001.parquet
s3://bucket/c_type=xyz/year=2020/month=01/part-00002.parquet
...
s3://bucket/c_type=abc/year=2018/month=01/part-00001.parquet
s3://bucket/c_type=abc/year=2018/month=01/part-00002.parquet
...

Following are some common use cases for URL and object pattern combinations:

TYPE

Specifies the type of the files in S3. The following types and type options are supported.

CSV Types

TYPE = (CSV [ <type option> ])

or

TYPE = (CSV)
[ <type option> ]

The following type options allow configuration for ingesting different CSV file formats.

• [ALLOW_DOUBLE_QUOTES = {TRUE|FALSE}]
[ALLOW_SINGLE_QUOTES = {TRUE|FALSE}]
  With ALLOW_DOUBLE_QUOTES = TRUE or ALLOW_SINGLE_QUOTES = TRUE you define that unescaped double or single quotes in CSV input file will not cause an error to be generated on ingest. By default ALLOW_DOUBLE_QUOTES and ALLOW_SINGLE_QUOTES are set to TRUE.

• [ALLOW_COLUMN_MISMATCH = {TRUE|FALSE}]
  With ALLOW_COLUMN_MISMATCH = TRUE the number of delimited columns in a CSV input file can be fewer than the number of columns in the corresponding table. By default, ALLOW_COLUMN_MISMATCH is set to FALSE, and an error is generated if the number of columns is fewer than the number of columns defined in the external table. If set to TRUE, and an input file record contains fewer columns than defined in the external table, the non-matching columns in the table are loaded with NULL values.

• [ALLOW_UNKNOWN_FIELDS = {TRUE|FALSE}]
  With ALLOW_UNKNOWN_FIELDS = TRUE the number of delimited columns in a CSV input file can be more than the number of columns in the corresponding table. By default, ALLOW_UNKNOWN_FIELDS is set to FALSE, and an error is generated if the number of columns is more than the number of columns defined in the external table. If set to TRUE, and an input file record contains more columns than defined in the external table, the non-matching columns in the table are loaded with NULL values.

• [ESCAPE_CHARACTER = {‘<character>’|NONE}
  With ESCAPE_CHARACTER = '<character>' you can define which character is used to escape, to change interpretations from the original. By default, the ESCAPE_CHARACTER value is set to \. If, for example, you want to use " as a value and not as delimiter for string, you can escape like \", with the default escape character.

• [FIELD_DELIMITER = '<field_delimeter>']
  With FIELD_DELIMITER = '<field_delimeter>', you can define a custom field delimiter to separate fields for ingest. By default, the FIELD_DELIMITER is set as ,.

• [NEW_LINE_CHARACTER = '<new_line_character>']
  With NEW_LINE_CHARACTER = '<new_line_character>', you can define a custom new line delimiter to separate entries for ingest. By default, the NEW_LINE_CHARACTER is set as the end of line character \n, but also supports other end of line conventions, such as \r\n, \n\r, and \r, as well as multi-character delimiters, such as #*~.

• [NULL_CHARACTER = '<null_character>']
  With NULL_CHARACTER = '<null_character>' you can define which character is interpreted as NULL. By default, the NULL_CHARACTER value is set to \\N.

• [SKIP_BLANK_LINES {TRUE|FALSE}]
  With SKIP_BLANK_LINES = TRUE any blank lines encountered in the CSV input file will be skipped. By default, SKIP_BLANK_LINES is set to FALSE, and an error is generated if blank lines are enountered on ingest.

• [SKIP_HEADER_ROWS = {1|0}]
  With SKIP_HEADER_ROWS = 1, Firebolt assumes that the first row in each file read from S3 is a header row and skips it when ingesting data. When set to 0, which is the default if not specified, Firebolt ingests the first row as data.

JSON Types

• TYPE = (JSON [PARSE_AS_TEXT = {'TRUE'|'FALSE'}])
  With TYPE = (JSON PARSE_AS_TEXT = 'TRUE'), Firebolt ingests each JSON object literal in its entirety into a single column of type TEXT. With TYPE = (JSON PARSE_AS_TEXT = 'FALSE'), Firebolt expects each key in a JSON object literal to map to a column in the table definition. During ingestion, Firebolt inserts the key’s value into the corresponding column.

Other Types

• TYPE = (ORC)
• TYPE = (PARQUET)
• TYPE = (AVRO)
• TYPE = (TSV)

All type options for CSV above, except for FIELD_DELIMITER, are also supported for the TSV file type.

Example

Creating an external table that reads parquet files from S3 is being done with the following statement:

CREATE EXTERNAL TABLE my_external_table
(
    c_id INTEGER,
    c_name TEXT
)
CREDENTIALS = (AWS_KEY_ID = '*****' AWS_SECRET_KEY = '******')
URL = 's3://bucket/'
OBJECT_PATTERN= '*.parquet'
TYPE = (PARQUET)

COMPRESSION

Specifies the compression type of the files matching the specified OBJECT_PATTERN in S3.

Syntax

[COMPRESSION = <compression_type>]

Parameters

Example

The example below creates an external table to ingest CSV files from S3 that are compressed using gzip. The credentials for an IAM user with access to the bucket are provided.

CREATE EXTERNAL TABLE my_external_table
(
    c_id INTEGER,
    c_name TEXT
)
CREDENTIALS = (AWS_KEY_ID = 'AKIAIOSFODNN7EXAMPLE' AWS_SECRET_KEY = 'wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY')
URL = 's3://mybucket/'
OBJECT_PATTERN= '*.csv.gz'
TYPE = (CSV)
COMPRESSION = GZIP