The Arrow C++ library includes a generic filesystem interface and specific implementations for some cloud storage systems. This setup allows various parts of the project to be able to read and write data with different storage backends. In the arrow R package, support has been enabled for AWS S3. This vignette provides an overview of working with S3 data using Arrow.

In Windows and macOS binary packages, S3 support is included. On Linux when installing from source, S3 support is not enabled by default, and it has additional system requirements. See vignette("install", package = "arrow") for details.

URIs

File readers and writers (read_parquet(), write_feather(), et al.) accept an S3 URI as the source or destination file, as do open_dataset() and write_dataset(). An S3 URI looks like:

s3://[access_key:secret_key@]bucket/path[?region=]

For example, one of the NYC taxi data files used in vignette("dataset", package = "arrow") is found at

s3://ursa-labs-taxi-data/2019/06/data.parquet

Given this URI, we can pass it to read_parquet() just as if it were a local file path:

df <- read_parquet("s3://ursa-labs-taxi-data/2019/06/data.parquet")

Note that this will be slower to read than if the file were local, though if you're running on a machine in the same AWS region as the file in S3, the cost of reading the data over the network should be much lower.

Creating a FileSystem object

Another way to connect to S3 is to create a FileSystem object once and pass that to the read/write functions. S3FileSystem objects can be created with the s3_bucket() function, which automatically detects the bucket's AWS region. Additionally, the resulting FileSystem will consider paths relative to the bucket's path (so for example you don't need to prefix the bucket path when listing a directory). This may be convenient when dealing with long URIs, and it's necessary for some options and authentication methods that aren't supported in the URI format.

With a FileSystem object, we can point to specific files in it with the $path() method. In the previous example, this would look like:

bucket <- s3_bucket("ursa-labs-taxi-data")
df <- read_parquet(bucket$path("2019/06/data.parquet"))

See the help for FileSystem for a list of options that s3_bucket() and S3FileSystem$create() can take. region, scheme, and endpoint_override can be encoded as query parameters in the URI (though region will be auto-detected in s3_bucket() or from the URI if omitted). access_key and secret_key can also be included, but other options are not supported in the URI.

The object that s3_bucket() returns is technically a SubTreeFileSystem, which holds a path and a file system to which it corresponds. SubTreeFileSystems can be useful for holding a reference to a subdirectory somewhere, on S3 or elsewhere.

One way to get a subtree is to call the $cd() method on a FileSystem

june2019 <- bucket$cd("2019/06")
df <- read_parquet(june2019$path("data.parquet"))

SubTreeFileSystem can also be made from a URI:

june2019 <- SubTreeFileSystem$create("s3://ursa-labs-taxi-data/2019/06")

Authentication

To access private S3 buckets, you need typically need two secret parameters: a access_key, which is like a user id, and secret_key, like a token. There are a few options for passing these credentials:

  1. Include them in the URI, like s3://access_key:secret_key@bucket-name/path/to/file. Be sure to URL-encode your secrets if they contain special characters like “/”.

  2. Pass them as access_key and secret_key to S3FileSystem$create() or s3_bucket()

  3. Set them as environment variables named AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY, respectively.

  4. Define them in a ~/.aws/credentials file, according to the AWS documentation.

You can also use an AccessRole for temporary access by passing the role_arn identifier to S3FileSystem$create() or s3_bucket().

File systems that emulate S3

The S3FileSystem machinery enables you to work with any file system that provides an S3-compatible interface. For example, MinIO is and object-storage server that emulates the S3 API. If you were to run minio server locally with its default settings, you could connect to it with arrow using S3FileSystem like this:

minio <- S3FileSystem$create(
  access_key = "minioadmin",
  secret_key = "minioadmin",
  scheme = "http",
  endpoint_override = "localhost:9000"
)

or, as a URI, it would be

s3://minioadmin:minioadmin@?scheme=http&endpoint_override=localhost%3A9000

(note the URL escaping of the : in endpoint_override).

Among other applications, this can be useful for testing out code locally before running on a remote S3 bucket.