Blobby provides uniform interface for object storage solutions of common cloud providers. It also provides a local filesystem-based implementation and an in-memory implementation for local development and testing.

In addition to the core APIs for manipulating and retrieving binary data, blobby also provides convenient wrappers to write and read pydantic objects serialised as JSON documents.

Provider support

• AWS S3
• Azure Blob Storage
• Filesystem
• Google Cloud Storage
• In-memory

Creating a storage

All storage implementations inherit from blobby.Storage and offer a uniform API.

AWS S3 storage

:warning: Install blobby with the aws extra, i.e. pip install blobby[aws]

The S3 implementation uses a boto3 client, which needs to be passed in when the storage is initialised. An S3 storage object represents a bucket, whose name also needs to be supplied.

import boto3
from blobby.aws import S3Storage

client = boto3.client("s3")
storage = S3Storage(client=client, bucket_name="my-bucket")

Azure Blob Storage

:warning: Install blobby with the azure extra, i.e. pip install blobby[azure]

The Azure implementation leverages the Azure SDK for Python. The storage expects the storage client to be provided.

from azure.storage.blob import BlobServiceClient
from blobby.azure import AzureBlobStorage

url = "<connection_string>"
service_client = BlobServiceClient.from_connection_string(url)
container_client = service_client.create_container("my-container")

storage = AzureBlobStorage(container_client)

Google Cloud Storage

:warning: Install blobby with the gcp extra, i.e. pip install blobby[gcp]

The Google Cloud Storage leverages the official SDK for Cloud Storage. The bucket object needs to be supplied to the storage when it's initialised.

from google.cloud.storage import Client
from blobby.gcp import GoogleCloudStorage

client = Client()
bucket = client.bucket("my-bucket")
storage = GoogleCloudStorage(bucket)

Filesystem storage

When creating a filesystem-based storage, the root directory needs to be provided. All files will be relative to this directory.

from blobby.filesystem import FileSystemStorage

storage = FileSystemStorage(root_dir="/my/storage/", create_missing_dirs=True)

The create_missing_dirs flag controls whether the root directory will be automatically created if it doesn't already exist.

In-memory storage

The in-memory implementation is backed with a simple dictionary stored in memory.

from blobby.memory import MemoryStorage

storage = MemoryStorage()

Common operations

Putting objects

The put operation works with bytes and str inputs. In either case, the object is stored as a binary blob.

key = "my-object"
data = b"hello world"
storage.put(key, data)

In the case of filesystem storage, the key needs to be a valid path.

Getting objects

key = "my-object"
storage.get(key)

Deleting objects

key = "my-object"
storage.delete(key)

Listing objects

Currently, only listing by object prefix is supported. This isn't very natural for filesystems, but the primary focus of this library is object storage solutions, which often don't have the concept of a folder or directory.

prefix = "my/prefix"
storage.list(prefix)

Pydantic objects

Pydantic objects can be written and read using dedicated APIs for convenience.

class MyData(pydantic.BaseModel):
    foo: str
    bar: int

key = "my/data"
data = MyData(foo="hello", bar=1)

storage.put_model_object(key, data)

Error handling

Storage implementations map their internal errors to shared error types, which are contained in blobby.error.

from blobby.error import NoSuchKeyError

try:
    storage.get("test")
except NoSuchKeyError as err:
    # do something with err
    pass