I have some java object and I want to expose all methods of it to custom javascript object. So, seems it would be reasonable to loop over all java object members, via reflection, and generate jsni code like: $wnd.myjso.func1 = function() { #package.MyJavaObject::func1()} and so on... Afterwards generated jsni must be executed to create javascript object on browser's window. How to implement it?
As you cannot use reflection at run-time, you have to do it at compile-time, using a GWT Generator.
That being said, you don't need to write such a generator by yourself, someone (now a Googler, member of the GWT team) has already done so: http://code.google.com/p/gwt-exporter/
Related
I have the following entity in GWT
#JsType(namespace = "my.entities")
public class MyEntity {
private Set<String> texts;
public Set<String> getTexts(){
if(this.texts==null)
this.texts=new LinkedHashSet<String>();
return this.texts;
}
public void setTexts(Set<String> texts){
this.texts=texts;
}
}
When I call myEntityVar.getTexts() in Javascript the returned object is a HashSet. It seems like jsinterop converts the java implementation of HashSet to JavaScript. But how can I create a new HashSet in JavaScript in order to use myEntityVar.setTexts(texts)? I tried an array for the "texts" param, but that doesn't work. So I somehow need to use HashSet in JavaScript.
However, I can't figure out, where to find it.
Any idea?
The short answer is that you can't - but then again, you also can't create a plain HashSet in JS either!
The reason that this works at all is that you've enabled -generateJsInteropExports, and while there is a JsInterop annotation on your MyEntity type, there is also one on java.util.Set (and a few other base JRE types). This allows for your code to return emulated java Sets without issue - any method which is compatible with running in JS is supported.
There are a few downsides:
Compiled size increases somewhat, since even if you don't use a method, it must be compiled in to your app this way, just in case JS uses it
Some methods are not supported - JS doesn't really have method overloading, so toArray() and toArray(T[]) look like the same method in JS. GWT solves this by not supporting the second method at all. (java.util.List has the same issue with remove(T) and remove(int), but it resolves it by renaming remove(int) to removeAtIndex(int) when compiled.)
If you never return these types, you'll probably want to disable this feature entirely - the -excludeJsInteropExports and -includeJsInteropExports flags to the compiler let you control what is exported.
To answer your question more directly, you have a few options that I can see:
Allow the setTexts method to be passed something else from JS, such as a JsArrayLike so that you could let users pass in a plain JS array of strings, or even a JS Set. You could go further and accept Object, and then type-check it to see what was passed in. You could even leave the Set override so it could be called from your own Java if necessary, but mark it as #JsIgnore so that GWT doesn't break when you attempt to export overloaded methods.
Create a factory method somewhere to create the Set implementation you would like your JS users to use. Since add and remove are supported, the calling JS code can build up the set before passing it in. Something like this:
#JsMethod(namespace = "my.Util")
public static <T> LinkedHashSet<T> createSet() {
return new LinkedHashSet<>();
}
Now they can call my.Util.createSet(), append items, and then pass it to your setTexts method.
Scenario: I have a GWT web application running within a JavaFX WebView/WebEngine. I am able to pass Strings from GWT to JavaScript to JavaFX without any issues.
Problem: When passing an array of custom objects like Data[] in the same fashion, the result on the JavaFX side is null.
An example of what Data looks like:
public class Data extends Serializable
{
char[] name;
int code;
short bar;
}
Here's the code to send the data to JavaScript:
public static native void doNativeStuff(String id, Data[] data) /*-{
$wnd.javaInterface.doStuff(id, data);
}-*/;
I've verified in the debugger that the Java object being passed in is populated with data and looks good.
Now on the JavaFX side, I have the following code to add the javaInterface to the page:
JSObject win = (JSObject) engine.executeScript("window");
win.setMember("javaInterface", new JavaInterface());
I know that this works because I'm using it for other methods that pass only Strings and they work great.
public void doStuff(String id, Data[] data)
{
// Right here, id == "validId" and data == null
if (data != null)
{
... do what is needed ...
}
}
Note that the Data object is defined and accessible on both sides.
From the GWT documentation:
Incoming Java type How it appears to JavaScript code
Java array opaque value that can only be passed back into Java code
I'm not touching it in JavaScript at all and I'm only passing it through from Java->JavaScript->Java, but the final step appears to be what is failing.
I've spent the last few hours scouring Stack Overflow, Google, GWT groups, gwtproject.org, etc. But most all of the examples only show a single argument being passed through and almost none of them show a Java Array being used.
I'd much rather just pass the object through rather than going to->from JSON, but I did give that a try out of desperation. I tried to use GSON but it doesn't work on the GWT client side. I tried to use the GWT AutoBean Framework but my Data object isn't a valid bean (I think because of no default constructor) and I cannot change that at this time.
I'm not using any Long or long values.
I've seen examples like this:
#com.google.gwt.examples.JSNIExample::staticFoo(Ljava/lang/String;)(s);
But from what I can tell that's just for going from JavaScript to GWT over JSNI. I'm trying to go the other way. I also couldn't find an example of this for multiple arguments.
I'm sure that there is just a minor tweak here that I'm missing, but I haven't been able to figure it out just yet. Please let me know if you see something that I'm missing here.
opaque value that can only be passed back into Java code
I think this means you cannot pass Java array into JavaScript code.
Agree with jat. I used to provide support for the similar needs and I had to serialize the objects myself.
And you can pass multiple arguments like this (types of arguments are given just for example):
private native void doJSAction(MyClass handler)/*-{
// do smth in JS
// then call external non-static method
handler.#com.myclient.helper.MyClass::doMyAction(Lcom/google/gwt/core/client/JavaScriptObject;Ljava/lang/String;Lcom/myclient/helper/MyClass;II)(jsNativeSmth, myString, handler, intA, intB);
}-*/;
where doMyAction is something like the following:
void doMyAction(JavaScriptObject jsObject, String s, MyClass instance, int a, int b)
I haven't played with JavaFX, but since it runs in a different VM and knows nothing about the GWT DevMode protocol (for example, a Java object is wrapped in a JS object that basically makes RPC calls to manipulate it), I am pretty sure you are going to have to serialize everything between GWT and JavaFX as Strings and primitives.
Hi I am trying to learn JAVA deeply and so I am digging into the JDK source code in the following lines:
URL url = new URL("http://www.google.com");
URLConnection tmpConn = url.openConnection();
I attached the source code and set the breakpoint at the second line and stepped into the code. I can see the code flow is: URL.openConnection() -> sun.net.www.protocol.http.Handler.openConnection()
I have two questions about this
First In URL.openConnection() the code is:
public URLConnection openConnection() throws java.io.IOException {
return handler.openConnection(this);
}
handler is an object of URLStreamHandler, define as blow
transient URLStreamHandler handler;
But URLStreamHandler is a abstract class and method openConnection() is not implement in it so when handler calls this method, it should go to find a subclass who implement this method, right? But there are a lot classes who implement this methods in sun.net.www.protocol (like http.Hanlder, ftp.Handler ) How should the code know which "openConnection" method it should call? In this example, this handler.openConnection() will go into http.Handler and it is correct. (if I set the url as ftp://www.google.com, it will go into ftp.Handler) I cannot understand the mechanism.
second. I have attached the source code so I can step into the JDK and see the variables but for many classes like sun.net.www.protocol.http.Handler, there are not source code in src.zip. I googled this class and there is source code online I can get but why they did not put it (and many other classes) in the src.zip? Where can I find a comprehensive version of source code?
Thanks!
First the easy part:
... I googled this class and there is source code online I can get but why they did not put it (and many other classes) in the src.zip?
Two reasons:
In the old days when the Java code base was proprietary, this was treated as secret-ish ... and not included in the src.zip. When they relicensed Java 6 under the GPL, they didn't bother to change this. (Don't know why. Ask Oracle.)
Because any code in the sun.* tree is officially "an implementation detail subject to change without notice". If they provided the code directly, it helps customers to ignore that advice. That could lead to more friction / bad press when customer code breaks as a result on an unannounced change to sun.* code.
Where can I find a comprehensive version of source code?
You can find it in the OpenJDK 6 / 7 / 8 repositories and associated download bundles:
http://hg.openjdk.java.net/jdk6/jdk6 - http://download.java.net/openjdk/jdk6/
http://hg.openjdk.java.net/jdk7/jdk7 - http://download.java.net/openjdk/jdk7/
http://hg.openjdk.java.net/jdk8/jdk8
Now for the part about "learning Java deeply".
First, I think you are probably going about this learning in a "suboptimal" fashion. Rather than reading the Java class library, I think you should be reading books on java and design patterns and writing code for yourself.
To the specifics:
But URLStreamHandler is a abstract class and method openConnection() is not implement in it so when handler calls this method, it should go to find a subclass who implement this method, right?
At the point that the handler calls than method, it is calling it on an instance of the subclass. So finding the right method is handled by the JVM ... just like any other polymorphic dispatch.
The tricky part is how you got the instance of the sun.net.www.protocol.* handler class. And that happens something like this:
When a URL object is created, it calls getURLStreamHandler(protocol) to obtain a handler instance.
The code for this method looks to see if the handler instance for the protocol already exists and returns that if it does.
Otherwise, it sees if a protocol handler factory exists, and if it does it uses that to create the handler instance. (The protocol handler factory object can be set by an application.)
Otherwise, searches a configurable list of Java packages to find a class whose FQN is package + "." + protocol + "." + "Handler", loads it, and uses reflection to create an instance. (Configuration is via a System property.)
The reference to handler is stored in the URL's handler field, and the URL construction continues.
So, later on, when you call openConnection() on the URL object, the method uses the Handler instance that is specific to the protocol of the URL to create the connection object.
The purpose of this complicated process is to support URL connections for an open-ended set of protocols, to allow applications to provide handlers for new protocols, and to substitute their own handlers for existing protocols, both statically and dynamically. (And the code is more complicated than I've described above because it has to cope with multiple threads.)
This is making use of a number of design patterns (Caches, Adapters, Factory Objects, and so on) together with Java specific stuff such as the system properties and reflection. But if you haven't read about and understood those design patterns, etcetera, you are unlikely to recognize them, and as a result you are likely to find the code totally bamboozling. Hence my advice above: learn the basics first!!
Take a look at URL.java. openConnection uses the URLStreamHandler that was previously set in the URL object itself.
The constructor calls getURLStreamHandler, which generates a class name dynamically and loads, and the instantiates, the appropriate class with the class loader.
But URLStreamHandler is a abstract class and method openConnection()
is not implement in it so when handler calls this method, it should go
to find a subclass who implement this method, right?
It has to be declared or abstract or implemented in URLStreamHandler. If you then give an instance of a class that extends URLStreamHandler with type URLStreamHandler and call the openConnection() method, it will call the one you have overriden in the instance of the class that extends URLStreamHandler if any, if none it will try to call the one in URLStreamHandler if implemented and else it will probably throw an exception or something.
Is there any way to create a fake from a System.Type object in FakeItEasy? Similar to:
var instance = A.Fake(type);
I try to write a fake container for AutoFac that automatically return fakes for all resolved types. I have looked in the code for FakeItEasy and all methods that support this is behind internal classes but I have found the interface IFakeObjectContainer that looks pretty interesting, but the implementations still need registration of objects that is the thing that I want to come around.
As of FakeItEasy 2.1.0 (but do consider upgrading to the latest release for more features and better bugfixes), you can create a fake from a Type like so:
using FakeItEasy.Sdk;
…
object fake = Create.Fake(type);
If you must use an earlier release, you could use some reflection based approach to create a method info for the A.Fake() method. (since it's about auto mocking this shouldn't be a problem really).
This is best done using a registration handler. You should look into how AutofacContrib.Moq implements its MoqRegistrationHandler. You'll see that it is actually using the generic method MockRepository.Create to make fake instances. Creating a similar handler for FakeItEasy should be quite simple.
Looks like the class that is generated for PlaceHistoryMapper is hard-coded to use AbstractPlaceHistoryMapper as the super class.
So, I am trying to work around this by trying to replace this AbstractPlaceHistoryMapper with a custom mapper of mine using deferred binding . I am using the following rule in my *.gwt.xml:
<replace-with class="com.google.gwt.place.impl.AbstractPlaceHistoryMapper">
<when-type-is class="com.test.sampleapp.CustomPlaceHistoryMapper" />
</replace-with>
But for some reason the replace does not seem to be happening. CustomPlaceHistoryMapper is not getting kicked in and the generated class still uses AbstractPlaceHistoryMapper.
Any thoughts/pointers as to what might be resulting this behavior are much appreciated.
Note: I have also posted this on the GWT group but haven't received an answer so far.
To make the deferred binding work a class must be created with GWT.create(). However, AbstractPlaceHistoryMapper is only used as an extended class. So it will never be created via GWT.create, but always by instantiation the subclass. And therefor deferred binding won't work in this case. If you want a complete different implementation you have to implement a custom PlaceHistoryMapper, and manage the known tokens yourself. This also means you can't use the History annotations either.
As a side note the classnames in your rule should be swapped. But for the end result this doesn't matter, since it won't work in the first place.
It is absolutely possible to have custom history tokens (eg. #mail or #mail/bla instead of only #mail:inbox) using the out-of-the-box Place-related classes that GWT (2.0) provides.
Instead of replacing AbstractPlaceHistoryMapper you could instantiate the default PlaceHistoryMapper passing in it's constructor your implementation of PlaceHistoryMapper<T> or PlaceHistoryMapperWithFactory<T>.
eg.:
final PlaceHistoryHandler placeHistoryHandler = new PlaceHistoryHandler(new CustomHistoryMapper());
You will be able then to map tokens as you wish.
I personally recommend you to use an unique PlaceTokenizer in you mapper custom implementation so that I dont have to have an inner PlaceTokenizer class in each of your Places.
Hope that helps. Feel free to ask any doubts.