Fuel
Last updated
Last updated
The core package for Fuel. The documentation outlined here touches most subjects and functions but is not exhaustive.
You can download and install Fuel with Maven and Gradle. The core package has the following dependencies:
You can make requests using functions on Fuel, a FuelManager instance or the string extensions.
The extensions and functions made available by the core package are listed here:
Fuel Method
String extension
Fuel/FuelManager method
Method.GET
"https://httpbin.org/get".httpGet()
Fuel.get("https://httpbin.org/get")
Method.POST
"https://httpbin.org/post".httpPost()
Fuel.post("https://httpbin.org/post")
Method.PUT
"https://httpbin.org/put".httpPut()
Fuel.put("https://httpbin.org/put")
Method.PATCH
"https://httpbin.org/patch".httpPatch()
Fuel.patch("https://httpbin.org/patch")
Method.HEAD
"https://httpbin.org/get".httpHead()
Fuel.head("https://httpbin.org/get")
Method.OPTIONS
not supported
Fuel.request(Method.OPTIONS, "https://httpbin.org/anything")
Method.TRACE
not supported
Fuel.request(Method.TRACE, "https://httpbin.org/anything")
Method.CONNECT
not supported
not supported
PATCH requestsThe default client is HttpClient which is a thin wrapper over java.net.HttpUrlConnection. java.net.HttpUrlConnection does not support a PATCH method. HttpClient converts PATCH requests to a POST request and adds a X-HTTP-Method-Override: PATCH header. While this is a semi-standard industry practice not all APIs are configured to accept this header by default.
Experimental
As of version 1.16.x you can opt-in to forcing a HTTP Method on the java.net.HttpUrlConnnection instance using reflection.
CONNECT requestConnect is not supported by the Java JVM via the regular HTTP clients, and is therefore not supported.
ParametersAll the String extensions listed above, as well as the Fuel and FuelManager calls accept a parameter parameters: Parameters.
URL encoded style for GET and DELETE request
Support x-www-form-urlencoded for PUT, POST and PATCH
Parameters with BodyIf a request already has a body, the parameters are url-encoded instead. You can remove the handling of parameter encoding by removing the default ParameterEncoder request interceptor from your FuelManager.
Parameters with multipart/form-dataThe UploadRequest handles encoding parameters in the body. Therefore by default, parameter encoding is ignored by ParameterEncoder if the content type is multipart/form-data.
Parameters with empty, array, list or null valuesAll requests can have parameters, regardless of the method.
a list is encoded as key[]=value1&key[]=value2&...
an array is encoded as key[]=value1&key[]=value2&...
an empty string value is encoded as key
a null value is removed
Request bodyBodies are formed from generic streams, but there are helpers to set it from values that can be turned into streams. It is important to know that, by default, the streams are NOT read into memory until the Request is sent. However, if you pass in an in-memory value such as a ByteArray or String, Fuel uses RepeatableBody, which are kept into memory until the Request is dereferenced.
When you're using the default Client, bodies are supported for:
POST
PUT
PATCH (actually a POST, as noted above)
DELETE
There are several functions to set a Body for the request. If you are looking for a multipart/form-data upload request, checkout the UploadRequest feature.
application/jsonIf you don't want to set the application/json header, you can use .jsonBody(value: String) extension to automatically do this for you.
StringFileInputStreamlazy source (InputStream)Fuel always reads the body lazily, which means you can also provide a callback that will return a stream. This is also known as a BodyCallback:
The default redirection interceptor only forwards RepeatableBody, and only if the status code is 307 or 308, as per the RFCs. In order to use a RepeatableBody, pass in a String or ByteArray as body, or explicitely set repeatable = true for the fun body(...) call.
NOTE this loads the entire body into memory, and therefore is not suited for large bodies.
HeadersThere are many ways to set, overwrite, remove and append headers. For your convenience, internally used and common header names are attached to the Headers companion and can be accessed (e.g. Headers.CONTENT_TYPE, Headers.ACCEPT, ...).
The most common ones are mentioned here:
HeaderValuesrequest[header]
header: String
Get the current values of the header, after normalisation of the header
request.header(header)
header: String
Get the current values
HeaderValuesrequest[header] = values
header: String, values: Collection<*>
Set the values of the header, overriding what's there, after normalisation of the header
request[header] = value
header: String, value: Any
Set the value of the header, overriding what's there, after normalisation of the header
request.header(map)
map: Map<String, Any>
Replace the headers with the map provided
request.header(pair, pair, ...)
vararg pairs: Pair<String, Any>
Replace the headers with the pairs provided
request.header(header, values)
header: String, values: Collection<*>
Replace the header with the provided values
request.header(header, value)
header: String, value: Any
Replace the header with the provided value
request.header(header, value, value, ...)
header: String, vararg values: Any
Replace the header with the provided values
HeaderValuesrequest.appendHeader(pair, pair, ...)
vararg pairs: Pair<String, Any>
Append each pair, using the key as header name and value as header content
request.appendHeader(header, value)
header: String, value: Any
Appends the value to the header or sets it if there was none yet
request.appendHeader(header, value, value, ...)
header: String, vararg values: Any
Appends the value to the header or sets it if there was none yet
Note that headers which by the RFC may only have one value are always overwritten, such as Content-Type.
FuelManager base headers vs. Request headersThe baseHeaders set through a FuelManager are only applied to a Request if that request does not have that specific header set yet. There is no appending logic. If you set a header it will overwrite the base value.
Client headers vs. Request headersAny Client can add, remove or transform HeaderValues before it sends the Request or after it receives the Response. The default Client for example sets TE values.
HeaderValues values are ListEven though some headers can only be set once (and will overwrite even when you try to append), the internal structure is always a list. Before a Request is made, the default Client collapses the multiple values, if allowed by the RFCs, into a single header value delimited by a separator for that header. Headers that can only be set once will use the last value by default and ignore earlier set values.
Authentication can be added to a Request using the .authentication() feature. By default, authentication is passed on when using the default redirectResponseInterceptor (which is enabled by default), unless it is redirecting to a different host. You can remove this behaviour by implementing your own redirection logic.
When you call
.authentication(), a few extra functions are available. If you call a regular function (e.g..header()) the extra functions are no longer available, but you can safely call.authentication()again without losing any previous calls.
Basic authentication
Bearer authentication
Any authentication using a header
Any request supports Progress callbacks when uploading or downloading a body; the Connection header does not support progress (which is the only thing that is sent if there are no bodies). You can have as many progress handlers of each type as you like.
Often the progress of a download will be shown as notification. Since Android Nougat (API 24), only 10 notification updates per second per app are allowed. Anything more than that will result in a log error like E/NotificationService: Package enqueue rate is 10.062265. Shedding events. package=....
It is the responsibility of the library user – not of fuel – to remedy this limit. A simple solution could look like this:
Not all source Body or Response Body report their total size. If the size is not known, the current size will be reported. This means that you will constantly get an increasing amount of totalBytes that equals readBytes.
multipart/form-data (UploadRequest)Fuel supports multipart uploads using the .upload() feature. You can turn any Request into a upload request by calling .upload() or call .upload(method = Method.POST) directly onto Fuel / FuelManager.
When you call
.upload(), a few extra functions are available. If you call a regular function (e.g..header()) the extra functions are no longer available, but you can safely call.upload()again without losing any previous calls.
request.add { }
varargs dataparts: (Request) -> DataPart
Add one or multiple DataParts lazily
request.add()
varargs dataparts: DataPart
Add one or multiple DataParts
request.progress(handler)
hander: ProgressCallback
Add a requestProgress handler
DataPart from FileIn order to add DataParts that are sources from a File, you can use FileDataPart, which takes a file: File. There are some sane defaults for the field name name: String, and remote file name filename: String, as well as the Content-Type and Content-Disposition fields, but you can override them.
In order to receive a list of files, for example in the field files, use the array notation:
Sending multiple files in a single datapart is not supported as it's deprecated by the multipart/form-data RFCs, but to simulate this behaviour, give the same name to multiple parts.
You can use the convenience constructors FileDataPart.from(directory: , filename: , ...args) to create a FileDataPart from String arguments.
DataPart from inline contentSometimes you have some content inline that you want to turn into a DataPart. You can do this with InlineDataPart:
A filename is not mandatory and is empty by default; the contentType is text/plain by default.
DataPart from InputStream (formely Blob)You can also add dataparts from arbitrary InputStreams, which you can do using BlobDataPart:
If you don't set the contentLength to a positive integer, your entire Request Content-Length will be undeterminable and the default HttpClient will switch to chunked streaming mode with an arbitrary stream buffer size.
Simply don't call add. The parameters are encoded as parts!
As mentioned before, you can use Fuel both synchronously and a-synchronously, with support for coroutines.
By default, there are three response functions to get a request synchronously:
response()
none
ResponseResultOf<ByteArray>
responseString(charset)
charset: Charset
ResponseResultOf<String>
responseObject(deserializer)
deserializer: Deserializer<U>
ResponseResultOf<U>
The default charset is UTF-8. If you want to implement your own deserializers, scroll down to advanced usage.
Add a handler to a blocking function, to make it asynchronous:
response() { handler }
handler: Handler
CancellableRequest
responseString(charset) { handler }
charset: Charset, handler: Handler
CancellableRequest
responseObject(deserializer) { handler }
deserializer: Deserializer, handler: Handler
CancellableRequest
The default charset is UTF-8. If you want to implement your own deserializers, scroll down to advanced usage.
The core package has limited support for coroutines:
await(deserializer)
deserializer: Deserializer<U>
U
awaitResult(deserializer)
deserializer: Deserializer<U>
Result<U, FuelError>
awaitResponse(deserializer)
deserializer: Deserializer<U>
ResponseOf<U>
awaitResponseResult(deserializer)
deserializer: Deserializer<U>
ResponseResultOf<U>
When using other packages such as fuel-coroutines, more response/await functions are available.
The ResponseResultOf<U> type is a Triple of the Request, Response and a Result<U, FuelError>
The ResponseOf<U> type is a Triple of the Request, Response and a U; errors are thrown
The Result<U, FuelError> type is a non-throwing wrapper around U
The U type doesn't wrap anything; errors are thrown
When defining a handler, you can use one of the following for all responseXXX functions that accept a Handler:
Handler<T>
2
1
calls success with an instance of T or failure on errors
ResponseHandler<T>
2
3
calls success with Request, Response and an instance of T, or failure or errors
ResultHandler<T>
1
1
invokes the function with Result<T, FuelError>
ResponseResultHandler<T>
1
3
invokes the function with Request Response and Result<T, FuelError>
This means that you can either choose to unwrap the Result yourself using a ResultHandler or ResponseResultHandler, or define dedicated callbacks in case of success or failure.
Result<T, FuelError>Result is a functional style data structure that represents data that contains result of Success or Failure but not both. It represents the result of an action that can be success (with result) or error.
Working with result is easy:
You can call [fold] and define a tranformation function for both cases that results in the same return type,
[destructure] as (data, error) = result because it is just a data class or
use when checking whether it is Result.Success or Result.Failure
File or OutputStream)Fuel supports downloading the request Body to a file using the .download() feature. You can turn any Request into a download request by calling .download() or call .download(method = Method.GET) directly onto Fuel / FuelManager.
When you call
.download(), a few extra functions are available. If you call a regular function (e.g..header()) the extra functions are no longer available, but you can safely call.download()again without losing any previous calls.
request.fileDestination { }
(Response, Request) -> File
Set the destination file callback where to store the data
request.streamDestination { }
(Response, Request) -> Pair<OutputStream, () -> InputStream>
Set the destination file callback where to store the data
request.progress(handler)
hander: ProgressCallback
Add a responseProgress handler
The stream variant expects your callback to provide a Pair with both the OutputStream to write too, as well as a callback that gives an InputStream, or raises an error.
The OutputStream is always closed after the body has been written. Make sure you wrap whatever functionality you need on top of the stream and don't rely on the stream to remain open.
The () -> InputStream replaces the body after the current body has been written to the OutputStream. It is used to make sure you can also retrieve the body via the response / await method results. If you don't want the body to be readable after downloading it, you have to do two things:
use an EmptyDeserializer with await(deserializer) or one of the response(deserializer) variants
provide an InputStream callback that throws or returns an empty InputStream.
For downloading larger files it is good to use request.streamDestination { } instead of request.fileDestination { }
RequestThe response functions called with a handler are async and return a CancellableRequest. These requests expose a few extra functions that can be used to control the Future that should resolve a response:
If you can't get hold of the CancellableRequest because, for example, you are adding this logic in an Interceptor, a generic Queue, or a ProgressCallback, you can call tryCancel() which returns true if it was cancelled and false otherwise. At this moment blocking requests can not be cancelled.
baseHeaders is to manage common HTTP header pairs in format of Map<String, String>.
The base headers are only applied if the request does not have those headers set.
Headers can be added to a request via various methods including
By default, all subsequent calls overwrite earlier calls, but you may use the appendHeader variant to append values to existing values.
In earlier versions (1.x.y), a mapOf overwrote, and varargs pair did not, but this was confusing. In 2.0, this issue has been fixed and improved so it works as expected.
Some of the HTTP headers are defined under Headers.Companion and can be used instead of literal strings. This is an encouraged way to configure your header in 2.x.y.
Fuel provides built-in support for response deserialization. Here is how one might want to use Fuel together with Gson
There are 4 methods to support response deserialization depending on your needs (also depending on JSON parsing library of your choice), and you are required to implement only one of them.
Another example may be parsing a website that is not UTF-8. By default, Fuel serializes text as UTF-8, we need to define our deserializer as such
Use singleton FuelManager.instance to manage global configurations.
Create separate managers using FuelManager()
basePath is used to manage common root path. Great usage is for your static API endpoint.
baseParams is used to manage common key=value query param, which will be automatically included in all of your subsequent requests in format of Parameters (Any is converted to String by toString() method)
client is a raw HTTP client driver. Generally, it is responsible to make Request into Response. Default is HttpClient which is a thin wrapper over java.net.HttpUrlConnection. You could use any httpClient of your choice by conforming to client protocol, and set back to FuelManager.instance to kick off the effect.
keyStore is configurable by user. By default it is null.
socketFactory can be supplied by user. If keyStore is not null, socketFactory will be derived from it.
hostnameVerifier is configurable by user. By default, it uses HttpsURLConnection.getDefaultHostnameVerifier().
requestInterceptors responseInterceptors is a side-effect to add to Request and/or Response objects.
For example, one might wanna print cUrlString style for every request that hits server in DEBUG mode.
Another example is that you might wanna add data into your Database, you can achieve that with providing responseInterceptors such as
In order to organize better your network stack FuelRouting interface allows you to easily setup a Router design pattern.
