aiohttp-s3-client

The simple module for putting and getting object from Amazon S3 compatible endpoints

Installation

pip install aiohttp-s3-client

Usage

from http import HTTPStatus

from aiohttp import ClientSession
from aiohttp_s3_client import S3Client


async with ClientSession(raise_for_status=True) as session:
    client = S3Client(
        url="http://s3-url",
        session=session,
        access_key_id="key-id",
        secret_access_key="hackme",
        region="us-east-1"
    )

    # Upload str object to bucket "bucket" and key "str"
    async with client.put("bucket/str", "hello, world") as resp:
        assert resp.status == HTTPStatus.OK

    # Upload bytes object to bucket "bucket" and key "bytes"
    async with await client.put("bucket/bytes", b"hello, world") as resp:
        assert resp.status == HTTPStatus.OK

    # Upload AsyncIterable to bucket "bucket" and key "iterable"
    async def gen():
        yield b'some bytes'

    async with client.put("bucket/file", gen()) as resp:
        assert resp.status == HTTPStatus.OK

    # Upload file to bucket "bucket" and key "file"
    async with client.put_file("bucket/file", "/path_to_file") as resp:
        assert resp.status == HTTPStatus.OK

    # Check object exists using bucket+key
    async with client.head("bucket/key") as resp:
        assert resp == HTTPStatus.OK

    # Get object by bucket+key
    async with client.get("bucket/key") as resp:
        data = await resp.read()

    # Delete object using bucket+key
    async with client.delete("bucket/key") as resp:
        assert resp == HTTPStatus.NO_CONTENT

    # List objects by prefix
    async for result in client.list_objects_v2("bucket/", prefix="prefix"):
        # Each result is a list of metadata objects representing an object
        # stored in the bucket.
        do_work(result)

Bucket may be specified as subdomain or in object name:

client = S3Client(url="http://bucket.your-s3-host", ...)
async with client.put("key", gen()) as resp:
    ...

client = S3Client(url="http://your-s3-host", ...)
async with client.put("bucket/key", gen()) as resp:
    ...

client = S3Client(url="http://your-s3-host/bucket", ...)
async with client.put("key", gen()) as resp:
    ...

Auth may be specified with keywords or in URL:

client = S3Client(url="http://your-s3-host", access_key_id="key_id",
                  secret_access_key="access_key", ...)

client = S3Client(url="http://key_id:access_key@your-s3-host", ...)

Temporary credentials are supported by using a token.

client = S3Client(url="http://your-s3-host", access_key_id="key_id",
                  secret_access_key="access_key", session_token="token",
                  ...)

Multipart upload

For uploading large files multipart uploading can be used. It allows you to asynchronously upload multiple parts of a file to S3. S3Client handles retries of part uploads and calculates part hash for integrity checks.

client = S3Client()
await client.put_file_multipart(
    "test/bigfile.csv",
    headers={
    	"Content-Type": "text/csv",
    },
    workers_count=8,
)

Parallel download to file

S3 supports GET requests with Range header. It's possible to download objects in parallel with multiple connections for speedup. S3Client handles retries of partial requests and makes sure that file won't changed during download with ETag header. If your system supports pwrite syscall (linux, macos, etc) it will be used to write simultaneously to a single file. Otherwise, each worker will have own file which will be concatenated after downloading.

client = S3Client()
await client.get_file_parallel(
    "dump/bigfile.csv",
    "/home/user/bigfile.csv",
    workers_count=8,
)