This is the Julia library needed along with code generated by the OpenAPI generator to help define, produce and consume OpenAPI interfaces.
The goal of OpenAPI is to define a standard, language-agnostic interface to REST APIs which allows both humans and computers to discover and understand the capabilities of the service without access to source code, documentation, or through network traffic inspection. When properly defined via OpenAPI, a consumer can understand and interact with the remote service with a minimal amount of implementation logic. Similar to what interfaces have done for lower-level programming, OpenAPI removes the guesswork in calling the service.
Check out OpenAPI-Spec for additional information about the OpenAPI project, including additional libraries with support for other languages and more.
Note: This package supersedes the Swagger.jl package. OpenAPI.jl and the associated generator can address both OpenAPI 2.x (Swagger) and OpenAPI 3.x specifications. Code dependent on Swagger.jl would not directly work with OpenAPI.jl, but migration should not be too difficult.
Use instructions provided for the Julia OpenAPI code generator plugin to generate Julia code.
Requires version 6.3.0 or later of openapi-generator.
Each model from the specification is generated into a file named model_<modelname>.jl
. It is represented as a mutable struct
that is a subtype of the abstract type APIModel
. Models have the following methods defined:
- constructor that takes keyword arguments to fill in values for all model properties.
propertynames
hasproperty
getproperty
setproperty!
In addition to these standard Julia methods, these convenience methods are also generated that help in checking value at a hierarchical path of the model.
function haspropertyat(o::T, path...) where {T<:APIModel}
function getpropertyat(o::T, path...) where {T<:APIModel}
E.g:
# access o.field.subfield1.subfield2
if haspropertyat(o, "field", "subfield1", "subfield2")
getpropertyat(o, "field", "subfield1", "subfield2")
end
# access nested array elements, e.g. o.field2.subfield1[10].subfield2
if haspropertyat(o, "field", "subfield1", 10, "subfield2")
getpropertyat(o, "field", "subfield1", 10, "subfield2")
end
Following validations are incorporated into models:
- maximum value: must be a numeric value less than or equal to a specified value
- minimum value: must be a numeric value greater than or equal to a specified value
- maximum length: must be a string value of length less than or equal to a specified value
- minimum length: must be a string value of length greater than or equal to a specified value
- maximum item count: must be a list value with number of items less than or equal to a specified value
- minimum item count: must be a list value with number of items greater than or equal to a specified value
- unique items: items must be unique
- maximum properties count: number of properties must be less than or equal to a specified value
- minimum properties count: number of properties must be greater than or equal to a specified value
- pattern: must match the specified regex pattern
- format: must match the specified format specifier (see subsection below for details)
- enum: value must be from a list of allowed values
- multiple of: must be a multiple of a specified value
Validations are imposed in the constructor and setproperty!
methods of models.
String, number and integer data types can have an optional format modifier that serves as a hint at the contents and format of the string. Validations for the following OpenAPI defined formats are built in:
Data Type | Format | Description |
---|---|---|
number | float | Floating-point numbers. |
number | double | Floating-point numbers with double precision. |
integer | int32 | Signed 32-bit integers (commonly used integer type). |
integer | int64 | Signed 64-bit integers (long type). |
string | date | full-date notation as defined by RFC 3339, section 5.6, for example, 2017-07-21 |
string | date-time | the date-time notation as defined by RFC 3339, section 5.6, for example, 2017-07-21T17:32:28Z |
string | byte | base64-encoded characters, for example, U3dhZ2dlciByb2Nrcw== |
Validations for custom formats can be plugged in by overloading the OpenAPI.val_format
method.
E.g.:
# add a new validation named `custom` for the number type
function OpenAPI.val_format(val::AbstractFloat, ::Val{:custom})
return true # do some validations and return result
end
# add a new validation named `custom` for the integer type
function OpenAPI.val_format(val::Integer, ::Val{:custom})
return true # do some validations and return result
end
# add a new validation named `custom` for the string type
function OpenAPI.val_format(val::AbstractString, ::Val{:custom})
return true # do some validations and return result
end
Each client API set is generated into a file named api_<apiname>.jl
. It is represented as a struct
and the APIs under it are generated as methods. An API set can be constructed by providing the OpenAPI client instance that it can use for communication.
The required API parameters are generated as regular function arguments. Optional parameters are generated as keyword arguments. Method documentation is generated with description, parameter information and return value. Two variants of the API are generated. The first variant is suitable for calling synchronously. It returns a tuple of the result struct and the HTTP response.
# example synchronous API that returns an Order instance
getOrderById(api::StoreApi, orderId::Int64) -> (result, http_response)
The second variant is suitable for asynchronous calls to methods that return chunked transfer encoded responses, where in the API streams the response objects into an output channel.
# example asynchronous API that streams matching Pet instances into response_stream
findPetsByStatus(
api::PetApi,
response_stream::Channel,
status::Vector{String}) -> (response_stream, http_response)
The HTTP response returned from the API calls, have these properties:
status
: integer status codemessage
: http message corresponding to status codeheaders
: http response headers asVector{Pair{String,String}}
A client context holds common information to be used across APIs. It also holds a connection to the server and uses that across API calls. The client context needs to be passed as the first parameter of all API calls. It can be created as:
Client(root::String;
headers::Dict{String,String}=Dict{String,String}(),
get_return_type::Function=(default,data)->default,
timeout::Int=DEFAULT_TIMEOUT_SECS,
long_polling_timeout::Int=DEFAULT_LONGPOLL_TIMEOUT_SECS,
pre_request_hook::Function,
verbose::Union{Bool,Function}=false,
)
Where:
root
: the root URI where APIs are hosted (should not end with a/
)headers
: any additional headers that need to be passed along with all API callsget_return_type
: optional method that can map a Julia type to a return type other than what is specified in the API specification by looking at the data (this is used only in special cases, for example when models are allowed to be dynamically loaded)timeout
: optional timeout to apply for server methods (defaultOpenAPI.Clients.DEFAULT_TIMEOUT_SECS
)long_polling_timeout
: optional timeout to apply for long polling methods (defaultOpenAPI.Clients.DEFAULT_LONGPOLL_TIMEOUT_SECS
)pre_request_hook
: user provided hook to modify the request before it is sentverbose
: whether to enable verbose logging
The pre_request_hook
must provide the following two implementations:
pre_request_hook(ctx::OpenAPI.Clients.Ctx) -> ctx
pre_request_hook(resource_path::AbstractString, body::Any, headers::Dict{String,String}) -> (resource_path, body, headers)
The verbose
option can be one of:
false
: the default, no verbose loggingtrue
: enables curl verbose logging to stderr- a function that accepts two arguments - type and message (available on Julia version >= 1.7)
- a default implementation of this that uses
@info
to log the arguments is provided asOpenAPI.Clients.default_debug_hook
- a default implementation of this that uses
In case of any errors an instance of ApiException
is thrown. It has the following fields:
status::Int
: HTTP status codereason::String
: Optional human readable stringresp::Downloads.Response
: The HTTP Response for this callerror::Union{Nothing,Downloads.RequestError}
: The HTTP error on request failure
An API call involves the following steps:
- If a pre request hook is provided, it is invoked with an instance of
OpenAPI.Clients.Ctx
that has the request attributes. The hook method is expected to make any modifications it needs to the request attributes before the request is prepared, and return the modified context. - The URL to be invoked is prepared by replacing placeholders in the API URL template with the supplied function parameters.
- If this is a POST request, serialize the instance of
APIModel
provided as thebody
parameter as a JSON document. - If a pre request hook is provided, it is invoked with the prepared resource path, body and request headers. The hook method is expected to modify and return back a tuple of resource path, body and headers which will be used to make the request.
- Make the HTTP call to the API endpoint and collect the response.
- Determine the response type / model, invoke the optional user specified mapping function if one was provided.
- Convert (deserialize) the response data into the return type and return.
- In case of any errors, throw an instance of
ApiException
The server code is generated as a package. It contains API stubs and validations of API inputs. It requires the caller to have implemented the APIs, the signatures of which are provided in the generated package module docstring.
A register
function is made available that when provided with a Router
instance, registers handlers
for all the APIs.
register(router, impl; path_prefix="", optional_middlewares...) -> HTTP.Router
Paramerets:
router
:HTTP.Router
to register handlers in, the same instance is also returnedimpl
: module that implements the server APIs
Optional parameters:
path_prefix
: prefix to be applied to all pathsoptional_middlewares
: Register one or more optional middlewares to be applied to all requests.
Optional middlewares can be one or more of:
init
: called before the request is processedpre_validation
: called after the request is parsed but before validationpre_invoke
: called after validation but before the handler is invokedpost_invoke
: called after the handler is invoked but before the response is sent
The order in which middlewares are invoked is:
init |> read |> pre_validation |> validate |> pre_invoke |> invoke |> post_invoke
The Petstore is a common example that most OpenAPI implementations use to test and demonstrate. Clients and servers generated from both version 2 and 3 specifications are included in this repo.
- Petstore v2:
- Client: docs, implementation
- Server: docs, implementation
- Petstore v3:
- Client: docs, implementation
- Server: docs, implementation
Swagger UI allows visualization and interaction with the API’s resources without having any of the implementation logic in place. OpenAPI.jl includes convenience methods to launch Swagger UI from Julia.
Use OpenAPI.swagger_ui
to open Swagger UI. It uses the standard swaggerapi/swagger-ui
docker image and requires docker engine to be installed.
# specify a specification file to start with
OpenAPI.swagger_ui(
spec::AbstractString; # the OpenAPI specification to use
port::Int=8080, # port to use
use_sudo::Bool=false # whether to use sudo while invoking docker
)
# specify a folder and specification file name to start with
OpenAPI.swagger_ui(
spec_dir::AbstractString; # folder containing the specification file
spec_file::AbstractString; # the specification file
port::Int=8080, # port to use
use_sudo::Bool=false # whether to use sudo while invoking docker
)
It returns the URL that should be opened in a browser to access the Swagger UI. Combining it with a tool like DefaultApplication.jl can help open a browser tab directly from Julia.
DefaultApplication.open(OpenAPI.swagger_ui("/my/openapi/spec.json"))
To stop the Swagger UI container, use OpenAPI.stop_swagger_ui
.
OpenAPI.stop_swagger_ui(;
use_sudo::Bool=false # whether to use sudo while invoking docker
)
Swagger Editor allows editing of OpenAPI specifications and simultaneous visualization and interaction with the API’s resources without having any of the client implementation logic in place. OpenAPI.jl includes convenience methods to launch Swagger Editor from Julia.
Use OpenAPI.swagger_editor
to open Swagger Editor. It uses the standard swaggerapi/swagger-editor
docker image and requires docker engine to be installed.
# specify a specification file to start with
OpenAPI.swagger_editor(
spec::AbstractString; # the OpenAPI specification to use
port::Int=8080, # port to use
use_sudo::Bool=false # whether to use sudo while invoking docker
)
# specify a folder and specification file name to start with
OpenAPI.swagger_editor(
spec_dir::AbstractString; # folder containing the specification file
spec_file::AbstractString; # the specification file
port::Int=8080, # port to use
use_sudo::Bool=false # whether to use sudo while invoking docker
)
# start without specifying any initial specification file
OpenAPI.swagger_editor(
port::Int=8080, # port to use
use_sudo::Bool=false # whether to use sudo while invoking docker
)
It returns the URL that should be opened in a browser to access the Swagger UI. Combining it with a tool like DefaultApplication.jl can help open a browser tab directly from Julia.
DefaultApplication.open(OpenAPI.swagger_editor("/my/openapi/spec.json"))
To stop the Swagger Editor container, use OpenAPI.stop_swagger_editor
.
OpenAPI.stop_swagger_editor(;
use_sudo::Bool=false # whether to use sudo while invoking docker
)
Not all OpenAPI features are supported yet, e.g.:
not
- inheritance and polymorphism
- some of the JSON schema keywords
- some subtler data types
- native representaion of some of the string formats, e.g. uuid, url
- read-only and write-only properties
- better enum support
- authentication schemes
deepObject
s in query parameters
There could be more unsupported features than what is listed above.