👁 Image
Blobject.Disk 5.0.19

Blobject

Blobject is a common, consistent storage interface for Microsoft Azure, Amazon S3, S3 compatible storage (i.e. Minio, Less3, View), CIFS (Windows file shares), NFS (Linux and UNIX file shares), Google Cloud Storage, and local filesystem written in C#.

Help, Feedback, Contribute

If you have any issues or feedback, please file an issue here in Github. We'd love to have you help by contributing code for new features, optimization to the existing codebase, ideas for future releases, or fixes!

Overview

This project was built to provide a simple interface over external storage to help support projects that need to work with potentially multiple storage providers. It is by no means a comprehensive interface, rather, it supports core methods for creation, retrieval, deletion, metadata, and enumeration.

Contributors

• @phpfui for adding the original code for BLOB copy functionality
• @Revazashvili for fixes related to byte array instantiation, Azure, and refactoring
• @courtzzz for keeping the region list updated

Dependencies

Though this library is MIT licensed, it is dependent upon other libraries, some of which carry a different license. Each of these libraries are included by reference, that is, none of their code has been modified.

New in v5.0.x

• Rename from BlobHelper to Blobject
• Added support for CIFS, NFS, and Google Cloud Storage
• Remove use of continuation tokens for disk
• Add S3Lite variant, not dependent on AWSSDK
• Enumerate APIs now return an IEnumerable<BlobMetadata>, no pagination required
• Blobject.AmazonS3 and Blobject.AmazonS3Lite v5.0.19 normalize AWS region aliases such as USEast2 to DNS-safe names such as us-east-2 and default AWS S3 settings to HTTPS
• Blobject.Core v5.0.19 fixes streaming-first copy, bounded bulk operations, empty-write consistency, filter cloning, and case handling
• Provider patch releases v5.0.19/v5.0.20 align stream APIs, empty writes, filter matching, and shared bulk behavior
• Provider packages now share common WriteManyAsync and EmptyAsync behavior with configurable MaxConcurrency
• Refactor

Example Project

Refer to the Test project for exercising the library.

Automated Tests

The repository includes Touchstone-based automated contract tests. Test.Shared contains the runner-agnostic provider contract definitions. By default, the tests run against Blobject.Disk in a temporary directory and clean up after each case.

dotnet run --project src/Test.Automated/Test.Automated.csproj --framework net10.0
dotnet test src/Test.Xunit/Test.Xunit.csproj --framework net10.0
dotnet test src/Test.Nunit/Test.Nunit.csproj --framework net10.0

The default suite contains 80 contract cases. Add --include-stress true to Test.Automated to include large enumeration and large WriteManyAsync coverage.

dotnet run --project src/Test.Automated/Test.Automated.csproj --framework net10.0 -- --include-stress true

Test.Automated accepts provider-specific command-line overrides. Remote providers are isolated by a generated prefix and cleaned up by default.

dotnet run --project src/Test.Automated/Test.Automated.csproj --framework net10.0 -- \
 --provider s3 \
 --s3-access-key <access-key> \
 --s3-secret-key <secret-key> \
 --s3-region us-east-2 \
 --s3-bucket <bucket> \
 --prefix blobject-contract-tests

For S3-compatible storage, also supply --s3-endpoint, --s3-ssl, and --s3-base-url. Supported providers are disk, s3, s3lite, azure, gcp, cifs, and nfs; run Test.Automated --help for the full argument list. The xUnit and NUnit adapters use the same shared suite and can be configured with the equivalent BLOBJECT_TEST_* environment variables, such as BLOBJECT_TEST_PROVIDER, BLOBJECT_TEST_S3_BUCKET, and BLOBJECT_TEST_PREFIX.

dotnet test src/Test.Xunit/Test.Xunit.csproj --framework net10.0 \
 -e BLOBJECT_TEST_PROVIDER=s3 \
 -e BLOBJECT_TEST_S3_ACCESS_KEY=<access-key> \
 -e BLOBJECT_TEST_S3_SECRET_KEY=<secret-key> \
 -e BLOBJECT_TEST_S3_REGION=us-east-2 \
 -e BLOBJECT_TEST_S3_BUCKET=<bucket>

The existing Test.* console applications remain available for interactive provider-specific testing.

Getting Started - AWS S3

using Blobject;

AwsSettings settings = new AwsSettings(
	accessKey,
	secretKey,
	"us-west-1",
	bucket);

BlobClient blobs = new BlobClient(settings);

AWS S3 region names are normalized against the current Amazon S3 regular endpoint region list. For example, USEast2, us_east_2, and us-east-2 are stored as us-east-2.

Getting Started - AWS S3 Compatible Storage (Minio, Less3, etc)

using Blobject.AmazonS3;

AwsSettings settings = new AwsSettings(
	endpoint, // http://localhost:8000/
	true, // enable or disable SSL
	accessKey,
	secretKey,
	"us-west-1",
	bucket,
	baseUrl // i.e. http://localhost:8000/{bucket}/{key}
	);

AmazonS3BlobClient blobs = new AmazonS3BlobClient(settings);

Getting Started - AWS S3 Lite (non-AWS library to reduce dependency drag)

using Blobject.AmazonS3Lite;

// Initialize settings as above
AmazonS3LiteBlobClient blobs = new AmazonS3LiteBlobClient(settings);

Getting Started - AWS S3 Anonymous Access

For accessing public buckets that don't require authentication, you can omit the access key and secret key:

using Blobject.AmazonS3Lite;

// Anonymous access to a public bucket
AwsSettings settings = new AwsSettings(
 null, // accessKey - null for anonymous access
 null, // secretKey - null for anonymous access
 "us-west-1",
 "public-bucket-name"
 );

AmazonS3LiteBlobClient blobs = new AmazonS3LiteBlobClient(settings);

// Check if credentials are configured
Console.WriteLine("Has credentials: " + settings.HasCredentials); // False

// Read operations work on public buckets
byte[] data = await blobs.GetAsync("public-file.txt");

Note: Write and delete operations require authentication. Anonymous access is read-only.

Getting Started - Azure

using Blobject.AzureBlob;

AzureBlobSettings settings = new AzureBlobSettings(
	accountName,
	accessKey,
	"https://[accountName].blob.core.windows.net/",
	containerName);

AzureBlobClient blobs = new AzureBlobClient(settings);

Getting Started - Google Cloud

Important - you must have the JSON credentials for the service account, which includes the private key. Creation of these credentials requires the organization policy administrator role.

using Blobject.GoogleCloud;

GcpBlobSettings settings = new GcpBlobSettings(
	projectId,
	bucket,
	"... JSON credentials for service account ..."

GcpBlobClient blobs = new GcpBlobClient(settings);

Getting Started - CIFS

using Blobject.CIFS;

CifsSettings settings = new CifsSettings(
	"localhost",
	username,
	password,
	sharename);

CifsBlobClient blobs = new CifsBlobClient(settings);

Getting Started - Disk

using Blobject.Disk;

DiskSettings settings = new DiskSettings("blobs");

DiskBlobClient blobs = new DiskBlobClient(settings);

Getting Started - NFS

using Blobject.NFS;

NfsSettings settings = new NfsSettings(
	"localhost",
	0, // user ID
	0, // group ID,
	sharename,
	NfsVersionEnum.V3 // V2, V3, or V4
	);

NfsBlobClient = new NfsBlobClient(settings);

Getting Started (Byte Arrays for Smaller Objects)

await blobs.WriteAsync("test", "text/plain", "This is some data"); // throws IOException
await blobs.WriteAsync("empty", "application/octet-stream", Array.Empty<byte>());
byte[] data = await blobs.GetAsync("test"); // throws IOException
bool exists = await blobs.ExistsAsync("test");
await blobs.DeleteAsync("test");

Getting Started (Streams for Larger Objects)

// Writing a file using a stream
FileInfo fi = new FileInfo(inputFile);
long contentLength = fi.Length;

using (FileStream fs = new FileStream(inputFile, FileMode.Open))
{
 await _Blobs.WriteAsync("key", "content-type", contentLength, fs); // throws IOException
}

// Downloading to a stream
BlobData blob = await _Blobs.GetStreamAsync(key);
// read blob.ContentLength bytes from blob.Data

Accessing Files within Folders

//
// Use a key of the form [path]/[to]/[file]/[filename].[ext]
//
await blobs.WriteAsync("subdirectory/filename.ext", "text/plain", "Hello!");

Metadata and Enumeration

Enumeration is always full; the library will manage any continuation tokens (e.g. AWS S3, S3 compatible, Azure) and also recurse into subdirectories for file, CIFS, and NFS.

// Get BLOB metadata
BlobMetadata md = await _Blobs.GetMetadataAsync("key");

// Enumerate BLOBs
await foreach (BlobMetadata blob in _Blobs.EnumerateAsync())
 Console.WriteLine(blob.Key + " " + blob.ContentLength + " folder? " + blob.IsFolder);

Enumeration filters are cloned internally so provider calls do not mutate caller-supplied filter instances. Object-storage providers use case-sensitive key matching; disk and CIFS use case-insensitive matching for compatibility with their typical filesystems.

Copying BLOBs from Repository to Repository

If you have multiple storage repositories and wish to move BLOBs from one repository to another, use the BlobCopy class (refer to the Test.Copy project for a full working example).

Thanks to @phpfui for contributing code and the idea for this enhancement!

// instantiate two BLOB clients
BlobCopy copy = new BlobCopy(from, to);
CopyStatistics stats = await copy.StartAsync();
/*
	{
	 "Success": true,
	 "Time": {
	 "Start": "2021-12-22T18:44:42.9098249Z",
	 "End": "2021-12-22T18:44:42.9379215Z",
	 "TotalMs": 28.1
	 },
	 "ContinuationTokens": 0,
	 "BlobsEnumerated": 12,
	 "BytesEnumerated": 1371041,
	 "BlobsRead": 12,
	 "BytesRead": 1371041,
	 "BlobsWritten": 12,
	 "BytesWritten": 1371041,
	 "Keys": [
	 "filename.txt",
	 ...
	 ]
	}
 */

BlobCopy.Start(...) remains available for compatibility and delegates to StartAsync(...). Copy now uses GetStreamAsync and stream writes where the provider supports them.

Bulk Operations

WriteManyAsync and EmptyAsync use bounded concurrency through BlobClientBase.MaxConcurrency, which defaults to 4.

blobs.MaxConcurrency = 8;
await blobs.WriteManyAsync(writes);

Version History

Refer to CHANGELOG.md for version history.