s3dol

s3 (through boto3) with a simple (dict-like or list-like) interface

To install: pip install s3dol

Documentation

import s3dol

s = s3dol.s3_store("my-bucket")  # MutableMapping[str, bytes]
s["hello.txt"] = b"world"
s["hello.txt"]  # b'world'
list(s)
"hello.txt" in s
del s["hello.txt"]

s = s3dol.s3_store("my-bucket", prefix="logs/")  # scoped to a prefix
s = s3dol.s3_store("open-data", anon=True)  # public bucket, no credentials
s = s3dol.s3_store("b", preset="minio", endpoint_url="http://localhost:9000")

Starting from nothing (the default never creates a bucket from a typo):

s = s3dol.s3_store("brand-new-bucket", on_missing_bucket="create")

Big objects go through the same interface — no upload_multipart method:

s["video.mp4"] = s3dol.Filepath("/tmp/video.mp4")  # streamed, multipart above 8 MiB
s["stream.bin"] = s3dol.Chunks(chunk_iterator)  # never fully in memory

Keyed capabilities are stores you index, not methods (a dol key wrapper hands a method the unmapped key — so s3dol has none):

s3dol.handles(s)["video.mp4"].read(offset=0, length=1024)  # ranged read
s3dol.handles(s)["video.mp4"].url(expires_in=3600)  # presigned URL
s3dol.urls(s)["video.mp4"]  # ...or directly
s3dol.info(s)["video.mp4"].size  # one HeadObject

Everything else is a free function taking the store first:

s3dol.sub(s, "folder/")
s3dol.prefixes(s)
s3dol.delete_many(s, ["a", "b"])
s3dol.delete_bucket(endpoint, "name", force=True)

Test without a cloud — no network, no docker, no moto:

from s3dol.testing import mock_s3, run_conformance


def test_my_service():
    assert MediaService(mock_s3(data={"a.mp4": b"..."})).play("a.mp4")

run_conformance(make_store) is exported too: run the same laws against your own store (or a sibling *dol package's).

A major, behaviour-clarifying v1 is in progress (design: misc/docs/architecture.md and the ADRs in misc/docs/decisions/). Today's S3Store will keep working through a compatibility shim, but a few v0 behaviours were bugs (e.g. an explicit endpoint_url= being silently dropped whenever AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY are exported) and v1 fixes them — which can move where your data goes.

Before upgrading to v1, run the diagnosis with the same arguments you pass S3Store today, in the environment you deploy in:

import s3dol

s3dol.diagnose(bucket_name="my-bucket", endpoint_url="...", path="...")

It prints what resolves (endpoint, region, signing, credential source — never a secret), where each value came from, and a v0-vs-v1 divergence table telling you whether the upgrade changes anything for your call. It never raises; a failing resolution is part of the report.

Recommended prerequisite to make getting started easier but not required.

export AWS_ACCESS_KEY_ID=AKIAIOSFODNN7EXAMPLE
export AWS_SECRET_ACCESS_KEY=wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY
export AWS_DEFAULT_REGION=us-west-2

Add credentails in ~/.aws/credentials

[default]
aws_access_key_id = AKIAIOSFODNN7EXAMPLE
aws_secret_access_key = wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY

Install AWS CLI

brew instal awscli

Set credentails with CLI

aws configure
AWS Access Key ID [None]: AKIAIOSFODNN7EXAMPLE
AWS Secret Access Key [None]: wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY
Default region name [None]: us-west-2
Default output format [None]: