AWS.jl
Overview
A Julia interface for Amazon Web Services.
This package replaces AWSCore.jl and AWSSDK.jl which previously provided low-level and high-level APIs respectively. It includes automated code generation to ensure all new AWS services are available, as well as keeping existing services up to date.
To see an overview of the architecture see the design document.
Installation
julia> Pkg.add("AWS")Usage
AWS.jl can be used with low-level and high-level API requests.
Please note when passing parameters for a request they must be a subtype of AbstractDict{String, <:Any}.
Low-Level
To use the low-level API, you must know how to perform the request you are making.
If you do not know how to perform a request you can reference the AWS Documentation.
Alternatively you can look at /src/services/{Service}.jl to find a list of available requests, as well as their required and optional parameters.
For example, to list the objects in an S3 bucket you must pass in the request method ("GET") and the endpoint ("/${bucket}"):
using AWS.AWSServices: s3
s3("GET", "/your-bucket")High-Level
To use the high-level API, you only need to know the name of the request you wish to make. For example again, to list the objects in an S3 bucket:
using AWS: @service
@service S3
S3.list_objects("your-bucket") # note: no '/' in front of bucket nameThe high-level function calls are wrapped around the low-level function calls, meaning you can still pass along any low-level kwargs such as aws_config when making these requests.
Note: When calling the @service macro you CANNOT match the predefined constant for the low level API. The low level API constants are named in all lowercase, and spaces are replaced with underscores.
using AWS.AWSServices: secrets_manager
using AWS: @service
# This matches the constant and will error!
@service secrets_manager
> ERROR: cannot assign a value to variable AWSServices.secrets_manager from module Main
# This does NOT match the filename structure and will error!
@service secretsmanager
> ERROR: could not open file /.julia/dev/AWS.jl/src/services/secretsmanager.jl
# All of the examples below are valid!
@service Secrets_Manager
@service SECRETS_MANAGER
@service sECRETS_MANAGERLimitations
Currently there are a few limitations with the high-level APIs. For example, with S3's DeleteMultipleObjects call. To remove multiple objects you must pass in an XML string (see below) in the body of the request.
Low-Level API Example:
using AWS.AWSServices: s3
body = """
<Delete xmlns="http://s3.amazonaws.com/doc/2006-03-01/">
<Object>
<Key>test.txt</Key>
</Object>
</Delete>
"""
bucket_name = "example-bucket"
s3("POST", "/$bucket_name?delete", Dict("body" => body)) # Delete multiple objectsThere is no-programatic way to see this from the aws-sdk-js, so the high-level function will not work.
High-Level API Example:
using AWS: @service
@service S3
body = """
<Delete xmlns="http://s3.amazonaws.com/doc/2006-03-01/">
<Object>
<Key>test.txt</Key>
</Object>
</Delete>
"""
bucket_name = "example-bucket"
S3.DeleteObjects(bucket_name, body) # Delete multiple objects
> ERROR: AWS.AWSExceptions.AWSException("MissingRequestBodyError", "Request Body is empty")There are most likely other similar functions which require more intricate details in how the requests are performed, both in the S3 definitions and in other services.
Modifying Functionality
There are sometimes situations, in which default behavior of AWS.jl might be overridden, for example when this package is used to access S3-compatible object storage of a different cloud service provider, which might have different ways of joining the endpoint url, encoding the region in the signature etc.
In many cases this can be achieved by creating a user-defined subtype of AbstractAWSConfig where some of the default methods are overwritten.
For example, if you want to use the S3 high-level interface to access public data from GCS without authorisation, you could define:
struct AnonymousGCS <:AbstractAWSConfig end
struct NoCredentials end
AWS.region(aws::AnonymousGCS) = "" # No region
AWS.credentials(aws::AnonymousGCS) = NoCredentials() # No credentials
AWS.check_credentials(c::NoCredentials) = c # Skip credentials check
AWS.sign!(aws::AnonymousGCS, ::AWS.Request) = nothing # Don't sign request
function AWS.generate_service_url(aws::AnonymousGCS, service::String, resource::String)
service == "s3" || throw(ArgumentError("Can only handle s3 requests to GCS"))
return string("https://storage.googleapis.com.", resource)
end
AWS.global_aws_config(AnonymousGCS())which skips some of the signature and credentials checking and modifies the generation of the endpoint url. A more extended example would be to use this package to access a custom minio server, we can define:
struct MinioConfig <: AbstractAWSConfig
endpoint::String
region::String
creds
end
AWS.region(c::MinioConfig) = c.region
AWS.credentials(c::MinioConfig) = c.credsand we define our own credentials type:
struct SimpleCredentials
access_key_id::String
secret_key::String
token::String
end
AWS.check_credentials(c::SimpleCredentials) = cas well as a custom url generator:
function AWS.generate_service_url(aws::MinioConfig, service::String, resource::String)
service == "s3" || throw(ArgumentError("Can only handle s3 requests to Minio"))
return string(aws.endpoint, resource)
end
AWS.global_aws_config(MinioConfig("http://127.0.0.1:9000", "aregion", SimpleCredentials("minio", "minio123", "")))Now we are ready to use AWS.jl to do S3-compatible requests to a minio server.
Alternative Solutions
There are a few alternatives to this package, the two below are being deprecated in favour of this package:
- AWSCore.jl - Low-level AWS interface
- AWSSDK.jl - High-level AWS interface
As well as some hand-written packages for specific AWS services: