RestEase is a little type-safe REST API client library for .NET Framework 4.5 and higher and .NET Core (.NET Platform Standard 1.1, PCL profiles 7, 32, 44, 111, 151, excluding iOS), which aims to make interacting with remote REST endpoints easy, without adding unnecessary compexity.
Almost every aspect of RestEase can be overridden and customized, leading to a large level of flexibility.
To use it, you define an interface which represents the endpoint you wish to communicate with (more on that in a bit), where methods on that interface correspond to requests that can be made on it. RestEase will then generate an implementation of that interface for you, and by calling the methods you defined, the appropriate requests will be made.
RestEase is built on top of HttpClient and is deliberately a "leaky abstraction": it is easy to gain access to the full capabilities of HttpClient, giving you control and flexibility, when you need it.
RestEase is heavily inspired by Paul Betts' Refit, which in turn is inspired by Retrofit.
- Installation
- Quick Start
- Request Types
- Return Types
- Query Parameters
- Path Placeholders
- Body Content
- Response Status Codes
- Cancelling Requests
- Headers
- Controlling Serialization and Deserialization
- Controlling the Requests
- Customizing RestEase
- Interface Accessibility
- Using Generic Interfaces
- Interface Inheritance
- Advanced Functionality Using Extension Methods
- FAQs
- Comparison to Refit
RestEase is available on NuGet.
Either open the package console and type:
PM> Install-Package RestEase
Or right-click your project -> Manage NuGet Packages... -> Online -> search for RestEase in the top right.
I also publish symbols on SymbolSource, so you can use the NuGet package but still have access to RestEase's source when debugging. If you haven't yet set up Visual Studio to use SymbolSource, do that now:
- Go to Tools -> Options -> Debugger -> General.
- Uncheck "Enable Just My Code (Managed only)".
- Uncheck "Enable .NET Framework source stepping". Yes, it is misleading, but if you don't, then Visual Studio will ignore your custom server order (see further on) and only use it's own servers.
- Check "Enable source server support".
- Uncheck "Require source files to exactly match the original version"
- Go to Tools -> Options -> Debugger -> Symbols.
- Select a folder for the local symbol/source cache. You may experience silent failures in getting symbols if it doesn't exist or is read-only for some reason.
- Add
http://srv.symbolsource.org/pdb/Publicunder "Symbol file (.pdb) locations".
To start, first create an public interface which represents the endpoint you wish to make requests to. Please note that it does have to be public, or you must add RestEase as a friend assembly, see Interface Accessibility below.
using System;
using System.Threading.Tasks;
using Newtonsoft.Json;
using RestEase;
namespace RestEaseSampleApplication
{
// We receive a JSON response, so define a class to deserialize the json into
public class User
{
public string Name { get; set; }
public string Blog { get; set; }
// This is deserialized using Json.NET, so use attributes as necessary
[JsonProperty("created_at")]
public DateTime CreatedAt { get; set; }
}
// Define an interface representing the API
// GitHub requires a User-Agent header, so specify one
[Header("User-Agent", "RestEase")]
public interface IGitHubApi
{
// The [Get] attribute marks this method as a GET request
// The "users" is a relative path the a base URL, which we'll provide later
// "{userId}" is a placeholder in the URL: the value from the "userId" method parameter is used
[Get("users/{userId}")]
Task<User> GetUserAsync([Path] string userId);
}
public class Program
{
public static void Main(string[] args)
{
// Create an implementation of that interface
// We'll pass in the base URL for the API
IGitHubApi api = RestClient.For<IGitHubApi>("https://api.github.com");
// Now we can simply call methods on it
// Normally you'd await the request, but this is a console app
User user = api.GetUserAsync("canton7").Result;
Console.WriteLine($"Name: {user.Name}. Blog: {user.Blog}. CreatedAt: {user.CreatedAt}");
Console.ReadLine();
}
}
}See the [Get("path")] attribute used above?
That's how you mark that method as being a GET request.
There are a number of other attributes you can use here - in fact, there's one for each type of request: [Get("path")], [Post("path")], [Put("path")], [Delete("path")], [Head("path")], [Options("path")], [Trace("path)")], [Patch("path")].
Use whichever one you need to.
The argument to [Get] (or [Post], or whatever) is typically a relative path, and will be relative to the base uri that you provide to RestClient.For<T>.
(You can specify an absolute path here if you need to, in which case the base uri will be ignored).
Your interface methods may return one of the following types:
Task: This method does not return any data, but the task will complete when the request has completedTask<T>(whereTis not one of the types listed below): This method will deserialize the response into an object of typeT, using Json.NET (or a custom deserializer, see Controlling Serialization and Deserialization below).Task<string>: This method returns the raw response, as a stringTask<HttpResponseMessage>: This method returns the rawHttpResponseMessageresulting from the request. It does not do any deserialiationTask<Response<T>>: This method returns aResponse<T>. AResponse<T>contains both the deserialied response (of typeT), but also theHttpResponseMessage. Use this when you want to have both the deserialized response, and access to things like the response headers
Non-async methods are not supported (use .Wait() or .Result as appropriate if you do want to make your request synchronous).
If you return a Task<HttpResponseMessage>, then HttpCompletionOption.ResponseHeadersRead is used, so that you can choose whether or not the response body should be fetched (or report its download progress, etc).
If however you return a Task<T>, Task<string>, or Task<Response<T>>, then HttpCompletionOption.ResponseContentRead is used, meaning that any CancellationToken that you pass will cancel the body download.
If you return a Task, then the response body isn't fetched, unless an ApiException is thrown.
It is very common to want to include query parameters in your request (e.g. /foo?key=value), and RestEase makes this easy.
The most basic type of query parameter is a constant - the value never changes. For these, simply put the query parameter as part of the URL:
public interface IGitHubApi
{
[Get("users/list?sort=desc")]
Task<List<User>> GetUsersAsync();
}Any parameters to a method which are:
- Decorated with the
[Query]attribute, or - Not decorated at all
will be interpreted as query parameters.
The name of the parameter will be used as the key, unless an argument is passed to [Query("key")], in which case that will be used instead.
For example:
public interface IGitHubApi
{
[Get("user")]
Task<User> FetchUserAsync(int userid);
// Is the same as:
[Get("user")]
Task<User> FetchUserAsync([Query] int userid);
// Is the same as:
// (Note the casing of the parameter name)
[Get("user")]
Task<User> FetchUserAsync([Query("userid")] int userId);
}
IGithubApi api = RestClient.For<IGithubApi>("http://api.github.com");
// Requests http://api.github.com/user?userId=3
await api.FetchUserAsync(3);You can have duplicate keys if you want:
public interface ISomeApi
{
[Get("search")]
Task<SearchResult> SearchAsync([Query("filter")] string filter1, [Query("filter")] string filter2);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://somenedpoint.com/search?filter=foo&filter=bar
await api.SearchAsync("foo", "bar");You can also have an array of query parameters:
public interface ISomeApi
{
// You can use IEnumerable<T>, or any type which implements IEnumerable<T>
[Get("search")]
Task<SearchResult> SearchAsync([Query("filter")] IEnumerable<string> filters);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://api.exapmle.com/search?filter=foo&filter=bar&filter=baz
await api.SearchAsync(new[] { "foo", "bar", "baz" });If you specify a key that is null, i.e. [Query(null)], then the name of the key is not used, and the value is inserted into the query string.
If you specify a key that is emptystring, then then query key will be left empty.
public interface ISomeApi
{
[Get("foo")]
Task FooAsync([Query(null)] string nullParam, [Query("")] string emptyParam);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://api.example.com/foo?onitsown&=nokey
await api.FooAsync("onitsown", "nokey");By default, query parameter values will be serialized by calling ToString() on them.
This means that the primitive types most often used as query parameters - string, int, etc - are serialized correctly.
However, you can also specify a string format to use.
If the type implements IFormattable, then your string format is passed to the IFormattable.ToString() method.
(null is passed for the IFormatProvider argument, meaning that the current thread's culture is used if applicable).
You specify a format using the Format property of the [Query] attribute, for example:
public interface ISomeApi
{
[Get("foo")]
Task FooAsync([Query(Format = "X2")] int param);
}
ISomeApi = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://api.example.com/foo?param=FE
await api.FooAsync(254);If you use a custom serializer, then the format is passed to that serializer, and you can use it as you like.
Sometimes calling ToString() is not enough: some APIs require that you send e.g. JSON as a query parameter.
In this case, you can mark the parameter for custom serialization using QuerySerializationMethod.Serialized, and further control it by using a custom serializer.
For example:
public class SearchParams
{
public string Term { get; set; }
public string Mode { get; set; }
}
public interface ISomeApi
{
[Get("search")]
Task<SearchResult> SearchAsync([Query(QuerySerializationMethod.Serialized)] SearchParams param);
}
ISomeApi = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://api.example.com/search?params={"Term": "foo", "Mode": "basic"}
await api.SearchAsync(new SearchParams() { Term = "foo", Mode = "basic" });You can also specify the default serialization method for an entire api by specifying [SerializationMethods(Query = QuerySerializationMethod.Serialized)] on the interface, or for all parameters in a given method by specifying it on the method, for example:
[SerializationMethods(Query = QuerySerializationMethods.Serialized)]
public interface ISomeApi
{
[Get("search")]
[SerializationMethods(Query = QuerySerializationMethod.ToString)]
Task<SearchResult> SearchWithToStringAsync([Query] SearchParams param);
[Get("search")]
Task<SearchResult> SearchWithSerializedAsync([Query] SearchParams param);
}Sometimes you have a load of query parameters, or they're generated dynamically, etc. In this case, you may want to supply a dictionary of query parameters, rather than specifying a load of method parameters.
To facilitate this, you may decorate one or more method parameters with [QueryMap].
The parameter type must be an IDictionary<TKey, TValue>.
Query maps are handled the same way as other query parameters: serialization, handling of enumerables, null values, etc, behave the same.
You can control whether values are serialized using a custom serializer or ToString() using e.g. [QueryMap(QuerySerializationMethod.Serialized)].
For example:
public interface ISomeApi
{
[Get("search")]
// I've used IDictionary<string, string[]> here, but you can use whatever type parameters you like,
// or any type which implements IDictionary<TKey, TValue>
Task<SearchResult> SearchBlogPostsAsync([QueryMap] IDictionary<string, string[]> filters);
}
var api = RestClient.For<ISomeApi>("http://api.example.com");
var filters = new Dictionary<string, string[]>()
{
{ "title", new[] { "bobby" } },
{ "tag", new[] { "c#", "programming" } }
};
// Requests http://api.example.com/search?title=bobby&tag=c%23&tag=programming
var searchResults = await api.SearchBlogPostsAsync(filters);In rare cases, you may have generated a query string by other means, and want to give this to RestEase.
To do this, provide a single parameter decorated with [RawQueryString].
This parameter can be of any type, and .ToString() will be called on it to turn it into a string.
Its value will be prepended, verbatim, to the query string: you are responsible for any escaping.
It must not begin or end with "&" or "?".
For example:
public interface ISomeApi
{
[Get("search")]
Task<SearchResult> SearchAsync([RawQueryString] string customFilter);
}
var api = RestClient.For<ISomeApi>("http://api.example.com");
var filter = "filter=foo"
var searchResults = await api.SearchAsync(filter);Sometimes you also want to be able to control some parts of the path itself, rather than just the query parameters.
Path parameters are the most common means of controlling a part of the path.
This is done using placeholders in the path, and corresponding method parameters decorated with [Path].
For example:
public interface ISomeApi
{
[Get("user/{userId}")]
Task<User> FetchUserAsync([Path] string userId);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://api.example.com/user/fred
await api.FetchUserAsync("fred");As with [Query], the name of the placeholder to substitute is determined by the name of the parameter.
If you want to override this, you can pass an argument to [Path("placeholder")], e.g.:
public interface ISomeApi
{
[Get("user/{userId}")]
Task<User> FetchUserAsync([Path("userId")] string idOfTheUser);
}Every placeholder must have a corresponding parameter, and every parameter must relate to a placeholder.
As with [Query], path parameter values will be serialized by calling ToString() on them.
This means that the primitive types most often used as query parameters - string, int, etc - are serialized correctly.
However, you can also specify a string format to use.
If the type implements IFormattable, then your string format is passed to the IFormattable.ToString() method.
(null is passed for the IFormatProvider argument, meaning that the current thread's culture is used if applicable).
You specify a format using the Format property of the [Path] attribute, for example:
public interface ISomeApi
{
[Get("foo/{bar}")]
Task FooAsync([Path("bar", Format = "D2")] int param);
}
ISomeApi = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://api.example.com/foo/01
await api.FooAsync(1);By default, path parameters are URL-encoded, which means things like / are escaped.
If you don't want this, for example you want to specify a literal section of the URL, this can be disabled using the UrlEncode property of the [Path] attribute, for example:
public interface ISomeApi
{
[Get("foo/{bar}")]
Task FooAsync([Path(UrlEncode = false)] string bar);
}
ISomeApi = RestClient.For<ISomeApi>("http://api.example.com");
// Requests http://api.example.com/foo/bar/baz
await api.FooAsync("bar/baz");Sometimes you've got a placeholder which is present in all (or most) of the paths on the interface, for example an account ID.
In this case, you can specify a [Path] property.
These work in the same way as path parameters, but they're on the level of the entire API.
Properties must have both a getter and a setter.
If the placeholder of the path property isn't given (i.e. you use [Path] instead of [Path("placeholder")]), then the name of the property will be used.
Unlike with path parameters, you don't need to have the placeholder present in every path. If you have both a path parameter and a path property with the same name, the path parameter is used.
For example:
public interface ISomeApi
{
[Path("accountId")]
int AccountId { get; set; }
[Get("{accountId}/profile")]
Task<Profile> GetProfileAsync();
[Delete("{accountId})]
Task DeleteAsync([Path("accountId")] int accountId);
}
var api = RestClient.For<ISomeApi>("http://api.example.com/user");
api.AccountId = 3;
// Requests /user/3/profile
var profile = await api.GetProfileAsync();
// Requests /user/4/profile
await api.DeleteAsync(4);As with Path Parameters, you can specify a string format to use if the value implements IFormattable.
For example:
public interface ISomeApi
{
[Path("accountId", Format = "N")]
Guid AccountId { get; set; }
[Get("{accountId}/profile")]
Task<Profile> GetProfileAsync()
}
var api = RestClient.For<ISomeApi>("http://api.example.com/user");
api.AccountId = someGuid;
// Requests e.g. /user/00000000000000000000000000000000 /profile
var profile = await api.GetProfileAsync();As with path parameters, you can disable URL encoding for path properties.
For example:
public interface ISomeApi
{
[Path("pathPart", UrlEncode = false)]
Guid PathPart { get; set; }
[Get("{pathPart}/profile")]
Task GetAsync();
}
var api = RestClient.For<ISomeApi>("http://api.example.com");
api.PathPart = "users/abc";
// Requests http://api.example.com/users/abc/profile
await api.GetAsync();If you're sending a request with a body, you can specify that one of the parameters to your method contains the body you want to send, using the [Body] attribute.
public interface ISomeApi
{
[Post("users/new")]
Task CreateUserAsync([Body] User user);
}Exactly how this will be serialized depends on the type of parameters:
- If the type is
Stream, then the content will be streamed viaStreamContent. - If the type is
String, then the string will be used directly as the content (usingStringContent). - If the parameter has the attribute
[Body(BodySerializationMethod.UrlEncoded)], then the content will be URL-encoded (see below). - If the type is a
HttpContent(or one of its subclasses), then it will be used directly. This is useful for advanced scenarios - Otherwise, the parameter will be serialized as JSON (by default, or you can customize this if you want, see Controlling Serialization and Deserialization).
For APIs which take form posts (i.e. serialized as application/x-www-form-urlencoded), initialize the [Body] attribute with BodySerializationMethod.UrlEncoded.
This parameter must implement IDictionary or IDictionary<TKey, TValue>.
If any of the values implement IEnumerable, then they will be serialized as an array of values.
For example:
public interface IMeasurementProtocolApi
{
[Post("collect")]
Task CollectAsync([Body(BodySerializationMethod.UrlEncoded)] Dictionary<string, object> data);
}
var data = new Dictionary<string, object> {
{"v", 1},
{"tids", new[] { "UA-1234-5", "UA-1234-6" },
{"cid", new Guid("d1e9ea6b-2e8b-4699-93e0-0bcbd26c206c")},
{"t", "event"},
};
// Serialized as: v=1&tids=UA-1234-5&tids=UA-1234-6&cid=d1e9ea6b-2e8b-4699-93e0-0bcbd26c206c&t=event
await api.CollectAsync(data);You can also control the default body serialization method for an entire API by specifying [SerializationMethods(BodySerializationMthod.UrlEncoded)] on the interface itself:
[SerializationMethods(BodySerializationMethod.UrlEncoded)]
public interface ISomeApi
{
[Post("collect")]
Task CollectAsync([Body] Dictionary<string, object> data);
}By default, any response status code which does not indicate success (as indicated by HttpResponseMessage.IsSuccessStatusCode) will cause an ApiException to be thrown.
This is usually what you want (you don't want to try and parse the result of a failed request), but sometimes you're expecting failure.
In this case, you can apply [AllowAnyStatusCode] to you method, or indeed to the whole interface, to suppress this behaviour. If you do this, then you probably want to make your method return either a HttpResponseMessage or a Response<T> (see Return Types) so you can examine the response code yourself.
For example:
public interface ISomeApi
{
[Get("users/{userId}")]
[AllowAnyStatusCode]
Task<Response<User>> FetchUserThatMayNotExistAsync([Path] int userId);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
var response = await api.FetchUserThatMayNotExistAsync(3);
if (response.ResponseMessage.StatusCode == HttpStatusCode.NotFound)
{
// User wasn't found
}
else
{
var user = response.GetContent();
// ...
}If you want to be able to cancel a request, pass a CancellationToken as one of the method paramters.
public interface ISomeApi
{
[Get("very-large-response")]
Task<LargeResponse> GetVeryLargeResponseAsync(CancellationToken cancellationToken);
}Specifying headers is actually a surprisingly large topic, and can be done in several ways, depending on the precise behaviour you want.
If you want to have a header that applies to every single request, and whose value is fixed, use a constant interface headers.
These are specified as [Header("Name", "Value")] attributes on the interface.
For example:
[Header("User-Agent", "RestEase")]
[Header("Cache-Control", "no-cache")]
public interface IGitHubApi
{
[Get("users")]
Task<List<User>> GetUsersAsync();
}If you want to have a header that applies to every single request, and whose value is variable, then use a variable interface header.
These are specifed using properties, using a [Header("Name")] attribute on that property.
For example:
public interface ISomeApi
{
[Header("X-API-Key")]
string ApiKey { get; set; }
[Get("users/{userId}")]
Task<User> FetchUserId([Path] string userId);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com")
api.ApiKey = "The-API-KEY-value";
// ...For nullable property types, you can also specify a default (which will be used when the property is null):
public interface ISomeApi
{
[Header("X-API-Key", "None")]
string ApiKey { get; set; }
[Get("users/{userId}")]
Task<User> FetchUserId([Path] string userId);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com")
// "X-API-Key: None"
var user = await api.FetchUserAsync("bob");If you want to have a header which only applies to a particular method, and whose value never changes, then use a constant method header. Like constant interface headers, these are defined in their entirety using an attribute. However, instead of applying the attribute to the interface, you apply it to the method.
public interface IGitHubApi
{
[Header("User-Agent", "RestEase")]
[Header("Cache-Control", "no-cache")]
[Get("users")]
Task<List<User>> GetUsersAsync();
// This method doesn't have any headers applied
[Get("users/{userId}")]
Task<User> GetUserAsync([Path] string userId);
}Finally, you can have headers which only apply to a single method and whose values are variable.
These consist of a [Header("Name")] attribute applied to a method parameter.
public interface ISomeApi
{
[Get("users")]
Task<List<User>> GetUsersAsync([Header("Authorization")] string authorization);
}You've probably noticed that there are 4 places you can define a header: on the interface, as a property, on a method, and as a parameter (or, Constant Interface Headers, Variable Interface Headers, Constant Method Headers, and Variable Method Headers, respectively). There are rules specifying how headers from different places are merged.
Constant and Variable Interface headers are merged, as are Constant and Variable Method headers. That is, if a header is supplied both as an attribute on the interface, and as a property, that header will have multiple values.
Method headers will replace Interface headers. If you have the same header on a method and on the interface, then the header on the method will replace the one on the interface.
Another rule is that a header with a value of null will not be added, but can still replace a previously-defined header of the same name.
Example time:
[Header("X-InterfaceOnly", "InterfaceValue")]
[Header("X-InterfaceAndParamater", "InterfaceValue")]
[Header("X-InterfaceAndMethod", "InterfaceValue"]
[Header("X-InterfaceAndParameter", "InterfaceValue"]
[Header("X-InterfaceAndMethod-ToBeRemoved", "InterfaceValue")]
public interface ISomeApi
{
[Header("X-ParameterOnly")]
string ParameterOnlyHeader { get; set; }
[Header("X-InterfaceAndParameter")]
string InterfaceAndParameterHeader { get; set; }
[Header("X-ParameterAndMethod")]
string ParameterAndMethodHeader { get; set; }
[Get("url")]
[Header("X-MethodOnly", "MethodValue")]
[Header("X-MethodAndParameter", "MethodValue")]
[Header("X-ParameterAndMethod", "MethodValue")]
[Header("X-InterfaceAndMethod-ToBeRemoved", null)]
Task DoSomethingAsync(
[Header("X-ParameterOnly")] string parameterOnly,
[Header("X-MethodAndParameter")] string methodAndParameter,
[Header("X-InterfaceAndParameter")] string interfaceAndParameter
);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
api.ParameterOnlyHeader = "ParameterValue";
api.InterfaceAndParameterHeader = "ParameterValue";
api.ParameterAndMethodHeader = "ParameterValue";
await api.DoSomethingAsync("ParameterValue", "ParameterValue", "ParameterValue");
// Has the following headers:
// X-InterfaceOnly: InterfaceValue
// X-InterfaceAndParameter: InterfaceValue, ParameterValue
// X-InterfaceAndMethod: MethodValue
// X-InterfaceAndParameter: ParameterValue
// X-ParameterAndMethod: MethodValue
// X-MethodOnly: MethodValue
// X-MethodAndParameter: MethodValue, ParameterValue
// X-ParameterAndMethod-ToBeRemoved isn't set, because it was removedBy default, RestEase will use Json.NET to deserialize responses, and serialize request bodies and query parameters.
However, you can change this, either by specifying custom JsonSerializerSettings, or by providing your own serializers / deserializers
If you want to specify your own JsonSerializerSettings, you can do this by constructing a new RestClient, assigning JsonSerializerSettings, then calling For<T>() to obtain an implementation of your interface, for example:
var settings = new JsonSerializerSettings()
{
ContractResolver = new CamelCasePropertyNamesContractResolver(),
Converters = { new StringEnumConverter() }
};
var api = new RestClient("http://api.example.com")
{
JsonSerializerSettings = settings
}.For<ISomeApi>();You can completely customize how requests are serialized, and responses deserialized, by providing your own serializer/deserializer implementations:
- To control how responses are deserialized, implement
IResponseDeserializer - To control how request bodies are serialized, implement
IRequestBodySerializer - To control how request query parameters are serialized, implement
IRequestQueryParamSerializer
You can, of course, provide a custom implementation of only one of these, or all of them, or any number in between.
This class has a single method, which is called whenever a response is received which needs deserializing.
It is passed the HttpResponseMessage (so you can read headers, etc, if you want) and its string content which has already been asynchronously read.
For an example, see JsonResponseDeserializer.
To tell RestEase to use it, you must create a new RestClient, assign its ResponseDeserializer property, then call For<T>() to get an implementation of your interface.
// This API returns XML
public class XmlResponseDeserializer : IResponseDeserializer
{
public T Deserialize<T>(string content, HttpResponseMessage response)
{
// Consider caching generated XmlSerializers
var serializer = new XmlSerializer(typeof(T));
using (var stringReader = new StringReader(content))
{
return (T)serializer.Deserialize(stringReader);
}
}
}
// ...
var api = new RestClient("http://api.example.com")
{
ResponseDeserializer = new XmlResponseDeserializer()
}.For<ISomeApi>();This class has a single method, which is called whenever a request body requires serialization (i.e. is decorated with [Body(BodySerializationMethod.Serialized)]).
It returns any HttpContent subclass you like, although StringContent is likely to be a common choice.
When writing an IRequestBodySerializer's SerializeBody implementation, you may choose to provide some default headers, such as Content-Type.
These will be overidden by any [Header] attributes.
For an example, see JsonRequestBodySerializer.
To tell RestEase to use it, you must create a new RestClient, assign its RequestBodySerializer property, then call For<T>() to get an implementation of your interface.
For example:
public class XmlRequestBodySerializer : IRequestBodySerializer
{
public HttpContent SerializeBody<T>(T body)
{
if (body == null)
return null;
// Consider caching generated XmlSerializers
var serializer = new XmlSerializer(typeof(T));
using (var stringWriter = new StringWriter())
{
serializer.Serialize(stringWriter, body);
var content = new StringContent(stringWriter.ToString());
// Set the default Content-Type header to application/xml
content.Headers.ContentType.MediaType = "application/xml";
return content;
}
}
}
// ...
var api = new RestClient("http://api.example.com")
{
RequestBodySerializer = new XmlRequestBodySerializer()
}.For<ISomeApi>();This class has two methods: one is called whenever a scalar query parameter requires serialization (i.e. is decorated with [Query(QuerySerializationMethod.Serialized)]); the other is called whenever a collection of query parameters (that is, the query parameter has type IEnumerable<T> for some T) requires serialization.
Both of these methods want you to return an IEnumerable<KeyValuePair<string, string>>, where each key corresponds to the name of a query name/value pair, and each value corresponds to the value.
For example:
return new[]
{
new KeyValuePair<string, string>("foo", "bar"),
new KeyValuePair<string, string>("foo", "baz"),
new KeyValuePair<string, string>("yay", "woo")
}
// Will get serialized to '...?foo=bar&foo=baz&yay=woo'It is unlikely that you will return more than one KeyValuePair from the method which serializes scalar query parameters, but the flexibility is there.
For an example, see JsonRequestQueryParamSerializer.
To tell RestEase to use it, you must create a new RestClient, assign its RequestQueryParamSerializer property, then call For<T>() to get an implementation of your interface.
For example:
// It's highly unlikely that you'll get an API which requires xml-encoded query
// parameters, but for the sake of an example:
public class XmlRequestQueryParamSerializer : IRequestQueryParamSerializer
{
public IEnumerable<KeyValuePair<string, string>> SerializeQueryParam<T>(string name, T value, RequestQueryParamSerializerInfo info)
{
if (value == null)
yield break;
// Consider caching generated XmlSerializers
var serializer = new XmlSerializer(typeof(T));
using (var stringWriter = new StringWriter())
{
serializer.Serialize(stringWriter, value);
yield return new KeyValuePair<string, string>(name, stringWriter.ToString()));
}
}
public IEnumerable<KeyValuePair<string, string>> SerializeQueryCollectionParam<T>(string name, IEnumerable<T> values, RequestQueryParamSerializerInfo info)
{
if (values == null)
yield break;
// Consider caching generated XmlSerializers
var serializer = new XmlSerializer(typeof(T));
foreach (var value in values)
{
if (value != null)
{
using (var stringWriter = new StringWriter())
{
serializer.Serialize(stringWriter, value);
yield return new KeyValuePair<string, string>(name, stringWriter.ToString()));
}
}
}
}
}
var api = new RestClient("http://api.example.com")
{
RequestQueryParamSerializer = new XmlRequestQueryParamSerializer()
}.For<ISomeApi>();If you specified a Format property on the [Query] attribute, this will be available as info.Format.
By default, this is null.
RestEase provides two ways for you to manipulate how exactly requests are made, before you need to resort to Customizing RestEase.
The first is a RestClient.For<T> overload which lets you specify a delegate which is invoked whenever a request is made.
This allows you to inspect and alter the request in any way you want: changing the content, changing the headers, make your own requests in the meantime, etc.
For example, if you need to refresh an oAuth access token occasionally (using the ADAL library as an example):
public interface IMyRestService
{
[Get("getPublicInfo")]
Task<Foobar> SomePublicMethodAsync();
[Get("secretStuff")]
[Header("Authorization", "Bearer")]
Task<Location> GetLocationOfRebelBaseAsync();
}
AuthenticationContext context = new AuthenticationContext(...);
IGitHubApi api = RestClient.For<IGitHubApi>("http://api.github.com", async (request, cancellationToken) =>
{
// See if the request has an authorize header
var auth = request.Headers.Authorization;
if (auth != null)
{
// The AquireTokenAsync call will prompt with a UI if necessary
// Or otherwise silently use a refresh token to return a valid access token
var token = await context.AcquireTokenAsync("http://my.service.uri/app", "clientId", new Uri("callback://complete")).ConfigureAwait(false);
request.Headers.Authorization = new AuthenticationHeaderValue(auth.Scheme, token);
}
});The second is a RestClient.For<T> overload which lets you specify a custom HttpClient to use.
This lets you customize the HttpClient, e.g. to set the request timeout.
It also lets you specify a custom HttpMessageHandler subclass, which allows you to control all sorts of things.
For example, if you wanted to 1) adjust the request timeout, and 2) allow invalid certificates (although the same approach would apply if you wanted to customize how certificates are validated), you could do something like this. Note that WebRequestHandler is a HttpMessageHandler subclass which allows you to specify things like ServerCertificateValidationCallback.
public class CustomHttpClientHandler : WebRequestHandler
{
// Let's log all of our requests!
private static readonly Logger logger = LogManager.GetCurrentClassLogger();
public CustomHttpClientHandler()
{
// Allow any cert, valid or invalid
this.ServerCertificateValidationCallback = (sender, certificate, chain, sslPolicyErrors) => true;
}
protected override async Task<HttpResponseMessage> SendAsync(HttpRequestMessage request, CancellationToken cancellationToken)
{
if (logger.IsTraceEnabled)
{
var response = await base.SendAsync(request, cancellationToken);
logger.Trace((await response.Content.ReadAsStringAsync()).Trim());
return response;
}
else
{
return await base.SendAsync(request, cancellationToken);
}
}
}
var httpClient = new HttpClient(new CustomHttpClientHandler())
{
BaseAddress = new Uri("https://secure-api.example.com"),
Timeout = TimeSpan.FromSeconds(3), // Very slow to respond, this server
};
ISomeApi api = RestClient.For<ISomeApi>(httpClient);You've already seen how to specify custom Serializers and Deserializers, and control requests.
RestEase has been written in a way which makes it very easy to customize exactly how it works. In order to describe this, I'm first going to have to outline its architecture.
Given an API like:
public interface ISomeApi
{
[Get("users/{userId}")]
Task GetUserAsync([Path] string userId);
}Calling RestClient.For<ISomeApi>(...) will cause a class like this to be generated:
namespace RestEase.AutoGenerated
{
public class ISomeApi
{
private readonly IRequester requester;
public ISomeApi(IRequester requester)
{
this.requester = requester;
}
public Task GetUserAsync(string userId)
{
var requestInfo = new RequestInfo(HttpMethod.Get, "users/{userId}");
requestInfo.AddPathParameter<string>("userId", userId);
return this.requester.RequestVoidAsync(requestInfo);
}
}
}Now, you cannot customize what this generated class looks like, but you can see it doesn't actually do very much: it just builds up a RequestInfo object, then sends it off to the IRequester (which does all of the hard work).
What you can do however is to provide your own IRequester implementation, and pass that to an appropriate overload of RestClient.For<T>.
In fact, the default implementation of IRequester, Requester, has been carefully written so that it's easy to extend: each little bit of functionality is broken out into its own virtual method, so it's easy to replace just the behaviour you need.
Have a read through Requester, figure out what you want to change, subclass it, and provide an instance of that subclass to RestClient.For<T>.
Since RestEase generates an interface implementation in a separate assembly, the interface ideally needs to be public.
If you don't want to do this, you'll need to mark RestEase as being a 'friend' assembly, which allows RestEase to see your internal types.
Add the following line to your AssemblyInfo.cs:
[assembly: InternalsVisibleTo(RestEase.RestClient.FactoryAssemblyName)]
When using something like ASP.NET Web API, it's a fairly common pattern to have a whole stack of CRUD REST services. RestEase supports these, allowing you to define a single API interface with a generic type:
public interface IReallyExcitingCrudApi<T, TKey>
{
[Post("")]
Task<T> Create([Body] T paylod);
[Get("")]
Task<List<T>> ReadAll();
[Get("{key}")]
Task<T> ReadOne(TKey key);
[Put("{key}")]
Task Update(TKey key, [Body]T payload);
[Delete("{key}")]
Task Delete(TKey key);
}Which can be used like this:
// The "/users" part here is kind of important if you want it to work for more
// than one type (unless you have a different domain for each type)
var api = RestClient.For<IReallyExcitingCrudApi<User, string>>("http://api.example.com/users"); You're allowed to use interface inheritance to share common properties and methods between different APIs.
However, you are not allowed to put any attributes ([Header] or [AllowAnyStatusCode]) onto the child interfaces being inherited, just onto the parent-most interface.
For example:
public interface IAuthenticatedEndpoint
{
[Header("X-Api-Token")]
string ApiToken { get; set; }
[Header("X-Api-Username")]
string ApiUsername { get; set; }
[Path("userId")]
int UserId { get; set; }
}
public interface IDevicesEndpoint : IAuthenticatedEndpoint
{
[Get("/devices")]
Task<IList<Device>> GetAllDevices([QueryMap] IDictionary<string, string> filters);
}
public interface IUsersEndpoint : IAuthenticatedEndpoint
{
[Get("/user/{userId}")]
Task<User> FetchUserAsync();
}If your interface implements IDisposable, then RestEase will generate a Dispose() method which disposes the underlying HttpClient.
Do this if you want to be able to dispose the HttpClient.
Sometimes you'll have cases where you want to do something that's more complex than can be achieved using RestEase alone (e.g. uploading multipart form data), but you still want to provide a nice interface to consumers. One option is to write extension methods on your interface. There are two ways of doing this.
The easiest thing to do is to put a method on your interface which won't be called by your code, but which can be wrapped by your extension method. This approach is unit testable.
public interface ISomeApi
{
// Method which is only used by SomeApiExtensions
[Post("upload")]
Task UploadAsync(HttpContent content);
}
public static class SomeApiExtensions
{
public static Task UploadAsync(this ISomeApi api, byte[] imageData, string token)
{
var content = new MultipartFormDataContent();
var imageContent = new ByteArrayContent(imageData);
imageContent.Headers.ContentType = new MediaTypeHeaderValue("image/jpeg");
content.Add(imageContent);
var tokenContent = new FormUrlEncodedContent(new[]
{
new KeyValuePair<string, string>("token", token)
});
content.Add(tokenContent);
return api.UploadAsync(content);
}
}Alternatively, you can put a property of type IRequester on your interface, then write an extension method which uses the IRequester.
Note that this approach is not unit testable.
public interface ISomeApi
{
IRequester Requester { get; }
}
public static class SomeApiExtensions
{
public static Task UploadAsync(this ISomeApi api, byte[] imageData, string token)
{
var content = new MultipartFormDataContent();
var imageContent = new ByteArrayContent(imageData);
imageContent.Headers.ContentType = new MediaTypeHeaderValue("image/jpeg");
content.Add(imageContent);
var tokenContent = new FormUrlEncodedContent(new[]
{
new KeyValuePair<string, string>("token", token)
});
content.Add(tokenContent);
var requestInfo = new RequestInfo(HttpMethod.Post, "upload");
requestInfo.SetBodyParameterInfo(BodySerializationMethod.Default, content);
return api.Requester.RequestVoidAsync(requestInfo);
}
}IRequester and RequestInfo are not documented in this README.
Read the doc comments.
Something like this...
public interface ISomeApi
{
[Header("Authorization")]
AuthenticationHeaderValue Authorization { get; set; }
[Get("foo")]
Task DoSomethingAsync();
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
var value = Convert.ToBase64String(Encoding.ASCII.GetBytes("username:password1234"));
api.Authorization = new AuthenticationHeaderValue("Basic", value);
await api.DoSomethingAsync();Sometimes your API responses will contain absolute URLs, for example a "next page" link. Therefore you'll want a way to request a resource using an absolute URL which overrides the base URL you specified.
Thankfully this is easy: if you give an absolute URL to e.g. [Get("http://api.example.com/foo")], then the base URL will be ignored.
You will also need to disable URL encoding.
public interface ISomeApi
{
[Get("users")]
Task<UsersResponse> FetchUsersAsync();
[Get("{url}")]
Task<UsersResponse> FetchUsersByUrlAsync([Path(UrlEncode = false)] string url);
}
ISomeApi api = RestClient.For<ISomeApi("http://api.example.com");
var firstPage = await api.FetchUsersAsync();
// Actually put decent logic here...
var secondPage = await api.FetchUsersByUrlAsync(firstPage.NextPage);Occasionally you get an API which can return both JSON and XML (apparently...). In this case, you'll want to auto-detect what sort of response you got, and deserialize with an appropriate deserializer.
To do this, use a custom deserializer, which can do this detection.
public class HybridResponseDeserializer : IResponseDeserializer
{
private T DeserializeXml<T>(string content)
{
// Consider caching generated XmlSerializers
var serializer = new XmlSerializer(typeof(T));
using (var stringReader = new StringReader(content))
{
return (T)serializer.Deserialize(stringReader);
}
}
private T DeserializeJson<T>(string content)
{
return JsonConvert.Deserialize<T>(content);
}
public T Deserialize<T>(string content, HttpResponseMessage response)
{
switch (response.Content.Headers.ContentType.MediaType)
{
case "application/json":
return this.DeserializeJson<T>(content);
case "application/xml":
return this.DeserializeXml<T>(content);
}
throw new ArgumentException("Response was not JSON or XML");
}
}
var api = RestClient.For<ISomeApi>("http://api.example.com", new HybridResponseDeserializer());Yes. It is safe to create implementations of interfaces from multiple threads at the same time (and to create multiple implementations of the same interface), and it is safe to use an implementation from multiple threads at the same time.
Let's assume you want to upload a file (from a stream), setting its name and content-type manually (skip these bits of not). There are a couple of ways of doing this, depending on your needs:
public interface ISomeApi
{
[Header("Content-Disposition", "form-data; filename=\"somefile.txt\"")]
[Header("Content-Type", "text/plain")]
[Post("upload")]
Task UploadFileVersionOneAsync([Body] Stream file);
[Post("upload")]
// You can use strings instead of strongly-typed header values, if you want
Task UploadFileVersionTwoAsync(
[Header("Content-Disposition")] ContentDispositionHeaderValue contentDisposition,
[Header("Content-Type")] MediaTypeHeaderValue contentType,
[Body] Stream file);
[Post("upload")]
Task UploadFileVersionThreeAsync([Body] HttpContent content);
}
ISomeApi api = RestClient.For<ISomeApi>("http://api.example.com");
// Version one (constant headers)
using (var fileStream = File.OpenRead("somefile.txt"))
{
await api.UploadFileVersionOneAsync(fileStream);
}
// Version two (variable headers)
using (var fileStream = File.OpenRead("somefile.txt"))
{
var contentDisposition = new ContentDispositionHeaderValue("form-header") { FileName = "\"somefile.txt\"" };
var contentType = new MediaTypeHeaderValue("text/plain");
await api.UploadFileVersionTwoAsync(contentDisposition, contentType, fileStream);
}
// Version three (precise control over HttpContent)
using (var fileStream = File.OpenRead("somefile.txt"))
{
var fileContent = new StreamContent(fileStream);
fileContent.Headers.ContentDisposition = new ContentDispositionHeaderValue("form-header") { FileName = "\"somefile.txt\"" };
fileContent.Headers.ContentType = new MediaTypeHeaderValue("text/plain");
await api.UploadFileVersionThreeAsync(fileContent);
}Obviously, set the headers you need - don't just copy me blindly.
You can use extension methods to make this more palatable for consumers.
RestEase is very heavily inspired by Paul Betts' Refit. Refit is a fantastic library, and in my opinion does a lot of things very right. It was the first C# REST client library that I actually enjoyed using.
I wrote RestEase for two reasons: 1) there were a couple of things about Refit which I didn't like, and 2) I thought it would be fun.
Here's a brief summary of pros/cons, compared to Refit:
- No autogenerated
RefitStubs.cs - Supports
CancellationTokensfor Task-based methods - Supports method overloading
- Supports properties to define common headers and path values
- Better support for API calls which are expected to fail:
[AllowAnyStatusCode]andResponse<T> - Easier to customize:
- Can specify custom response deserializer
- Can specify custom request serializer
- Can customize almost every aspect of setting up and creating the request (through implementing
IRequester)
- Supports
[QueryMap] - Supports custom query parameter serialization
- Supports arrays of query parameters (and body parameters when serializing a body parameter as UrlEncoded)
- Supports
IDictionary<TKey, TValue>as well asIDictionarytypes when serializing a body parameter as UrlEncoded. This allows e.g.ExpandoObjectto be used here - Supports formatting
IFormattablepath and query params - Allows
HttpClientto be disposed
- Interfaces need to be public, or you need to add
[assembly: InternalsVisibleTo(RestClient.FactoryAssemblyName)]to yourAssemblyInfo.cs - No
IObservablesupport - Slightly more work done at runtime (but not very much more)