READ_FILES

A table-valued function (TVF) that reads a file of any supported format, inferring the format and dispatching to the matching reader (READ_PARQUET, READ_CSV, READ_JSON, READ_AVRO). Use it when you do not want to name the format up front. Like the other read TVFs, it accepts a LOCATION object or direct credentials, and reads a single file or a glob of files. READ_FILES also has a shorthand: a string literal in the FROM clause is read as READ_FILES, so

SELECT * FROM '/data/events/2026-06-01.parquet';

is equivalent to SELECT * FROM READ_FILES('/data/events/2026-06-01.parquet'). This works for local paths and for object storage that does not need credentials (for example public Amazon S3 buckets); to pass credentials or a LOCATION, call READ_FILES(...) directly.

Format inference

The format is inferred by listing the objects the path points at and inspecting the most recent one’s name, after stripping any compression suffix (.gz, .zst, .br, …). The path itself does not need an extension: a directory or glob resolves from the files it matches, so s3://bucket/events/* containing only .json.gz objects reads as gzip-compressed JSON. Recognized extensions:

Extension	Format
`.parquet`	Parquet
`.csv`	CSV
`.tsv`	TSV (read as CSV with a tab delimiter)
`.json`, `.jsonl`, `.ndjson`	JSON
`.avro`	Avro
`.orc`	ORC

If the matched objects have no usable extension, READ_FILES cannot infer the format; use the format-specific TVF (READ_PARQUET, READ_CSV, …) instead.

Syntax

-- Using LOCATION object (recommended)
READ_FILES (
  LOCATION => 'location_name'
  [, PATTERN => <pattern>]
  [, COMPRESSION => <file_compression>]
)
|
-- Using static credentials
READ_FILES (
  URL => <url>
  [, COMPRESSION => <file_compression>]
  [, AWS_ACCESS_KEY_ID => <aws_access_key_id>]
  [, AWS_SECRET_ACCESS_KEY => <aws_secret_access_key>]
  [, AWS_SESSION_TOKEN => <aws_session_token>]
  [, AWS_ROLE_ARN => <aws_role_arn>]
  [, AWS_ROLE_EXTERNAL_ID => <aws_role_external_id>]
)
|
-- Shorthand: a string literal in the FROM clause
SELECT * FROM '<path>';

When you pass AWS_ROLE_ARN, set the optional AWS_ROLE_EXTERNAL_ID to add a customer-controlled condition to your role’s trust policy.

For role-based AWS access you can additionally set an external ID. An external ID is a value you choose and control that AWS checks when Firebolt assumes your role, adding a second condition on top of your account’s unique IAM principal. Configuring one is a recommended best practice. See IAM roles.

Parameters

Parameter	Description	Supported input types
`LOCATION`	The name of a location object that contains the Amazon S3 URL and credentials. Firebolt recommends using `LOCATION` to store credentials for authentication. For a comprehensive guide, see LOCATION objects.	`TEXT`
`PATTERN`	When using `LOCATION`, an optional glob pattern to filter files within the location’s URL path, applied relative to the location’s base path.	`TEXT`
`URL`	The location of your files. The expected format is `s3://{bucket_name}/{full_file_path_glob_pattern}`. A local path is supported only where local file system access is enabled, and may be an absolute or relative path (relative paths resolve against the engine’s working directory), optionally prefixed with `file://`.	`TEXT`
`COMPRESSION`	The compression type of the input. If unset, it is inferred from the most recent object’s extension.	`TEXT`
`AWS_ACCESS_KEY_ID`	The AWS access key ID.	`TEXT`
`AWS_SECRET_ACCESS_KEY`	The AWS secret access key.	`TEXT`
`AWS_SESSION_TOKEN`	The AWS session token.	`TEXT`
`AWS_ROLE_ARN`	The AWS role arn.	`TEXT`
`AWS_ROLE_EXTERNAL_ID`	The AWS role external ID.	`TEXT`

Parquet, JSON, and Avro read nested objects as STRUCT; ORC flattens them into separate columns. CSV and TSV are read with default dialect options (comma or tab delimiter, no header row, types inferred). To control the CSV dialect (header, quote, delimiter, null string), use READ_CSV.

Return Type

A table with the data from the file. Column names and types are inferred the same way the format-specific TVF infers them.

Examples

Read a single Parquet file, inferring the format from the extension:

SELECT * FROM READ_FILES('s3://firebolt-publishing-public/help_center_assets/firebolt_sample_dataset/levels.parquet');

The bare-path shorthand reads the same file, and works across deployment modes for public data:

SELECT * FROM 's3://firebolt-publishing-public/help_center_assets/firebolt_sample_dataset/levels.parquet';

Read all matching files with a glob. The format is inferred from the matched objects, so the pattern itself needs no extension:

SELECT * FROM READ_FILES('s3://firebolt-publishing-public/help_center_assets/firebolt_sample_dataset/*');

SELECT * FROM 's3://firebolt-publishing-public/help_center_assets/firebolt_sample_dataset/levels*';

Read a gzip-compressed CSV; both the format and the compression are inferred from the object name:

SELECT * FROM READ_FILES('s3://my-bucket/exports/report.csv.gz');

Overview

Performance and Observability

Security

Self-Managed

Managed service

Guides

SQL reference

Release notes

API reference

Legal

Format inference

Syntax

Parameters

Return Type

Examples

​Format inference

​Syntax

​Parameters

​Return Type

​Examples

Format inference

Syntax

Parameters

Return Type

Examples