The primary difference in Unirest 3 is that the org.json dependency has been replaced by a clean-room implementation of the it's interface using Google Gson as the engine.
This was done due to conflicts with the org.json license which requires that "The Software shall be used for Good, not Evil.". While many people would rightly view this as silly and unenforceable by law, many organizations such as Eclipse, Debian, and Apache will not allow using it.
Several reasons:
- It has not been maintained in several years and no longer matches the org.json signatures.
- It causes classpath conflicts which many projects forbid.
- We would like Unirest to be able to expand beyond org.json and offer more advanced native features like object mapping.
- Gson is closest in spirit and method signature to org.json and was deemed quicker to adopt.
- It's small, mature and a single dependency.
- It would conflict less in other projects than Jackson would which is both more popular and far more complex.
Implementation was done without looking at the internals of the org.json classes. This was accomplished by writing extensive unit tests in order to document behavior and method signatures and then simply changing the test to use this projects own classes as well as Google Gson.
- The namespace is now
kong.unirest.json
- For the most part kong.unirest.json honors all public interfaces and behavior of
JSONArray
,JSONObject
, andJSONPointer
. - The utility classes in org.json have NOT been implemented as they are not required for Unirest's use case. So things like XML-to-JSON, and CSV-to-JSON have not been implemented.
- Custom indenting with
.toString(int spaces)
does not honor the indent factor and always uses 2 spaces. Waiting on google/gson#1280 for a fix. - There are some slight differences in the details of some error messages.
All main classes are now in the kong.unirest
package. Classes related to the underlying Apache Http client that powers unirest are kept in kong.unirest.apache
This project doesn't have many files, it really doesn't need anything more complicated than that.
- Java 8: Java 8 is now required for Unirest due to extensive lambda support.
.asBinary()
and.getRawResponse()
methods have been removed. These have been replaced by consumer methods which allow you to read the InputStream directly and not a copy. (seeHttpRequest::thenConsume(Consumer<RawResponse> consumer)
- Removal of all Apache classes in the non-config interfaces. These have ben replaced by Unirest native interfaces. Typically these interfaces are very similar to the older Apache classes and so updating shouldn't be a problem.
Previous versions of unirest had configuration split across several different places.
// Sometimes it was on Unirest
Unirest.setTimeouts(5000, 10000);
Unirest.setConcurrency(25, 10);
//Sometimes it was on Options
Options.setOption(HTTPCLIENT, client);
Often you could do it in both places with different impacts on the lifecycle of the client. All configuration has now been centralized in Unirest.config()
Unirest config allows easy access to build a configuration just like you would build a request:
Unirest.config()
.socketTimeout(500)
.connectTimeout(1000)
.concurrency(10, 5)
.proxy(new Proxy("https://proxy"))
.setDefaultHeader("Accept", "application/json")
.followRedirects(false)
.enableCookieManagement(false)
.addInterceptor(new MyCustomInterceptor());
Changing Unirest's config should ideally be done once, or rarely. There are several background threads spawned by both Unirest itself and Apache HttpAsyncClient. Once Unirest has been activated configuration options that are involved in creating the client cannot be changed without an explicit shutdown or reset.
Unirest.config()
.reset()
.connectTimeout(5000)
You can set your own custom Apache HttpClient and HttpAsyncClient. Note that Unirest settings like timeouts or interceptors are not applied to custom clients.
Unirest.config()
.httpClient(myClient)
.asyncClient(myAsyncClient)
As usual, Unirest maintains a primary single instance. Sometimes you might want different configurations for different systems. You might also want an instance rather than a static context for testing purposes.
// this returns the same instance used by Unirest.get("http://somewhere/")
UnirestInstance unirest = Unirest.primaryInstance();
// It can be configured and used just like the static context
unirest.config().connectTimeout(5000);
String result = unirest.get("http://foo").asString().getBody();
// You can also get a whole new instance
UnirestInstance unirest = Unirest.spawnInstance();
WARNING! If you get a new instance of unirest YOU are responsible for shutting it down when the JVM shuts down. It is not tracked or shut down by Unirest.shutDown();