read_parquet

Read parquet files from a local filesystem or supported object stores.

URLs provided to read_parquet may also include glob characters to allow scanning multiple files when reading from S3, GCS, or a local filesystem. Globbed paths are not allowed when querying parquet files over HTTP/HTTPS.

Syntax

-- Single url or path.
read_parquet(<url>)
-- Multiple urls or paths.
read_parquet([<url>])
-- Using a cloud credentials object.
read_parquet(<url>, <credential_object>)
-- Required named argument for S3 buckets.
read_parquet(<url>, <credentials_object>, region => '<aws_region>')
-- Pass S3 credentials using named arguments.
read_parquet(<url>, access_key_id => '<aws_access_key_id>', secret_access_key => '<aws_secret_access_key>', region => '<aws_region>')
-- Pass GCS credentials using named arguments.
read_parquet(<url>, service_account_key => '<gcp_service_account_key>')

Parameter descriptions and which object store it’s relevant to.

Parameter	Object store	Description
url	All	The URL or path to a parquet file to scan.
credential_object	S3 and GCS	A database object storing credentials for accessing the object or objects.
aws_region	S3	If scanning an object in S3, the region of the bucket.
aws_access_key_id	S3	ID of AWS access key with permissions to read from the bucket.
aws_secret_access_key	S3	Secret associated with the AWS access key.
gcp_service_account_key	GCS	A JSON-encoded GCP service account key with access to the bucket.

Usage

Local files

Local files can be read by passing the path or a globbed path to read_parquet. Paths may be absolute or relative.

-- Read a relative path.
SELECT * FROM read_parquet('./my_data.parquet');
-- Read all parquet files in a directory.
SELECT * FROM read_parquet('./directory_of_data/*.parquet');
-- Read parquet files from multiple directories.
SELECT * FROM read_parquet('./**/*.parquet');
-- Read multiple explicitly provided files.
SELECT * FROM read_parquet(['./directory_of_data/1.parquet', './directory_of_data/2.parquet']);

Objects in GCS

Objects in GCS can be read by providing the path to the object and credentials for the object to read_parquet. The service account key provided should be JSON encoded.

-- Single object.
SELECT * FROM read_parquet('gs://my-bucket/path/object.parquet',
                           service_account_key => '<service_account_key>');
-- Scan objects matching a glob.
SELECT * FROM read_parquet('gs://my-bucket/path/prefix_*.parquet',
                           service_account_key => '<service_account_key>');
-- Read all objects in a directory.
SELECT * FROM read_parquet('gs://my-bucket/path/*',
                           service_account_key => '<service_account_key>');

Optionally, a credentials object can be created and used in place of providing the service account key directly.

-- Create the credentials.
CREATE CREDENTIALS my_gcp_creds PROVIDER gcp
    ( service_account_key '<service_account_key>' );
-- And use them in the scan.
SELECT * FROM read_parquet('gs://my-bucket/path/object.parquet', my_gcp_creds);

Objects in S3

Objects in GCS can be read by providing the path to the object and credentials for the object, and the bucket region to read_parquet.

SELECT * FROM read_parquet('gs://my-bucket/path/object.parquet',
                           access_key_id = '<access_key_id>',
                           secret_access_key = '<secret_access_key>',
                           region = 'us-east-1');
-- Scan objects matching a glob.
SELECT * FROM read_parquet('gs://my-bucket/path/prefix_*.parquet',
                           access_key_id = '<access_key_id>',
                           secret_access_key = '<secret_access_key>',
                           region = 'us-east-1');
-- Read all objects in a directory.
SELECT * FROM read_parquet('gs://my-bucket/path/prefix_*.parquet',
                           access_key_id = '<access_key_id>',
                           secret_access_key = '<secret_access_key>',
                           region = 'us-east-1');

Optionally, a credentials object can be created and used in place of providing the service account key directly.

-- Create the credentials.
CREATE CREDENTIALS my_aws_creds PROVIDER aws
    OPTIONS (
        access_key_id = '<access_key_id>',
        secret_access_key = '<secret_access_key>',
    );
-- And use them in the scan. Note that a region still needs to be provided.
SELECT * FROM read_parquet('gs://my-bucket/path/object.parquet', my_aws_creds, region => 'us-east-1');

HTTP files

Parquet files can also be scanned over HTTP/HTTPS.

-- Read a single file.
SELECT * FROM read_parquet('https://my_domain.com/file.parquet');
-- Read multiple files
SELECT * FROM read_parquet(['https://my_domain.com/1.parquet', 'https://my_domain.com/2.parquet']);

Globbed paths are not supported when reading parquet files over HTTP/HTTPS

Examples

In the following example we use COPY TO to create a parquet file and finally read it with read_parquet.

-- Output the file to 'example.parquet'
COPY (SELECT * FROM generate_series(1, 10)) TO 'example.parquet';

-- And scan it back in.
SELECT * FROM read_parquet('./example.parquet');

-- Output:
-- ┌─────────────────┐
-- │ generate_series │
-- │ ──              │
-- │ Int64           │
-- ╞═════════════════╡
-- │ 1               │
-- │ 2               │
-- │ 3               │
-- │ 4               │
-- │ 5               │
-- │ 6               │
-- │ 7               │
-- │ 8               │
-- │ 9               │
-- │ 10              │
-- └─────────────────┘