Why does GWT use code generators rather than annotation processors? - java

GWT uses code generators with a GWT.create() command to generate implementations. This provides a declarative way of doing things in GWT but I don't understand why it's not done with annotation processors like in most java.this articleis what original made me think of it. It seems like it would simplify the build and let you see your compile time errors and be much more compatible with regular java. This is a side question but am I the only one that thinks GWT should support reflection. I've heard that this makes the code slower but still. Isn't it worth it in some cases?

GWT was created before annotations even existed in the Java language/platform (Java 1.4), this is why.
New developments tend to lean towards annotation processing though, and it will eventually be supported as part of a SuperDevMode recompile (you currently have to run your annotation processor and have SDM pick the –modified– generated sources); one example is Dagger 2.
I however don't think this approach is applicable to all cases where a generator is currently used in GWT.
As for GWT.create(), it'll still be needed/useful for the replace with cases of deferred bindings.
As for reflection, it's not about "slowness", it's about the metadata that would need to be embedded into the generated JS code to support runtime reflection. Given that everything is known at compile-time though, there's hardly any need for reflection at runtime; you can generally use compile-time generated code as a replacement for runtime reflection. So no, it's not worth it.

Related

A Java compiler plugin can change the AST of a compiled program. Is that a hack or is it officially supported?

In Java, a compiler plugin is made by subclassing com.sun.source.util.Plugin (in Java 11). Java compiler plugins use classes that inherit from Tree,
https://docs.oracle.com/en/java/javase/11/docs/api/jdk.compiler/com/sun/source/tree/Tree.html
These classes seem not to allow changes in the AST. However, some plugins, as those described in
https://bkushigian.github.io/2018/06/03/optimizing-javac-plugin.html
and
https://www.baeldung.com/java-build-compiler-plugin
do change the AST. The question is: is that a hack?
Not a hack. The examples are old/misleading (anything using Method.setAccessible(true) is always a hack) but the com.sun.tools.javac.tree.TreeScanner visitor and TreeMaker factory are as official as the javac API is ever going to get.
From their javadoc:
This is NOT part of any supported API. If you write code that depends
on this, you do so at your own risk. This code and its internal
interfaces are subject to change or deletion without notice.
This is their freedom to innovate without the compiler becoming a backwards-compatible mess like the main JLS/JDK sometimes is. You can understand why they would need this statement and I do not consider it a hack but it a warning about the amount of future maintenance you may need over time. IMHO sub-classing TreeScanner is quite accommodating of future language changes and I would not expect fundamental incompatibilities as new language features are rolled out.

Alternatives to Java bytecode instrumentation

I'm starting a project that will have to instrument java applications for coverage purposes (definition-usage of variables, etc). It has to add trace statements and some logic to the application and maybe remove statements.
I have searched for ways of instrument Java code and what I always find is about bytecode instrumentation.
My question is: It's the only way to instrument Java applications? There is any other way to do that? What are the advantages of bytecode instrumentation over the others?
I'll probably use the bytecode solution, but I want to know what are the problems with the other approaches (if any) to decide precisely.
Thanks!
The other method close to changing bytecode is using AOP (Aspect Oriented Programming).
The main library is AspectJ which also mostly defines the area.
The third option that might be interesting (since you are just starting out with the program) is using Spring.
It means you will have to learn a bit about IOC (inversion of control) but it basically means that instead of creating your objects yourself you let spring do it for you, and it has it advantages because when spring is incharge of the creation it can add all sorts of things in the creation process without you having to really declare it all yourself in aspectj.
In terms of complexity I would probably rate it:
spring (easiest)
aspectj
bytecode instrumentation (hardest)
but it's exactly the other way around when talking about capabilities (power). for example doing something like substracting code is only possible using the last one (I think)
You should definitely check out AspectJ
From what you describe you will be able to do what you want with it.
Doing bytecode instrumentation yourself is absolutely possible but it much more complicated.
I think you should check out AsepctJ first and got back to do bytecode instrumentation yourself as last resort.
See my paper on building coverage tools using program transformation engines. This approach has the advantage that it can be used on arbitrary programming languages. In addition, it sees the source code the way the programmer sees it, not as compiled byte codes (as generics get more complex, and get ground into finer byte codes, it gets harder to understand that source code by inspecting the byte code).
It is perhaps worth noting that program transformation generalizes aspect-oriented programming.

Complete metaprogramming framework for Java?

I'm interested in metaprogramming (i.e. programs that help programmers do tedious programming tasks). I'm looking for a tool which has the following properties:
usable both at compile time and runtime;
inspects program structure;
can add new classes, methods or fields and make them visible to Java compiler;
can change behavior of methods;
Java-based (well, Java is most popular programming language according to some rankings);
good integration with IDEs and build tools like Ant, Gradle or Maven;
actively maintained project;
easy to use and extend;
There are some solutions for this, like:
reflection
AspectJ
Annotation Processing Tool
bytecode manipulation (CGLIB, Javassist, java.lang.instrument)
Eclipse JDT
Project Lombok
Groovy, JRuby, Scala
But unfortunately none of them meets all the criteria above. Is there any complete metaprogramming solution for Java?
There's JackPot, which is Java based but I don't think gets a lot a current attention. Has ASTs and symbol tables AFAIK. You can probably extend it; I doubt anybody will stop (or help) you.
There's the Java-based compiler APIs for the Sun, er, Oracle java compiler. They're likely actively maintained, but I don't think you can modify source code and regenerate it. Certainly has symbol tables; dunno about trees. Probably pretty hard to extend; you have to keep up with the compiler guys, not the other way round.
There is ANTLR, which has a Java implementation and a Java parser that will build ASTs. I don't think it has full symbol tables, so doing serious code analysis/revision is likely to be hard. ANTLR is certainly actively maintained, and nobody will object to you enhancing the Java grammar with symbol tables. Just know that will take you about 6 months for Java 1.6 if that's all you do. (That's how long it took our internal [smart] guy to do it for DMS, starting with symbol table support for 1.4).
Not in Java, and not easily integrated into IDEs, but capable of carrying massive analysis and transformation on Java code is our DMS Software Reengineering Toolkit with its Java Front End.
DMS is generic compiler machinery: parsing, AST building, symbol table machinery, flow analysis machinery, with that additional bonuses of source-to-source transformations and generic prettyprinting of ASTs back to legal text including retention of comments. It offers a set of APIs supporting these services, and additional tools for defining grammars and langauge-dependent flow analyzers.
The Java Front End gives crucial detail (using those APIs) to DMS to allow it process Java: a grammar/parser, full symbol table construction for Java 1.4-1.6 (with 1.7 due momentarily), as well as some control and data flow analysis (to be extended over time because this stuff is so useful).
By using the services provided by DMS and the Java Front end, one can reasonably contemplate building arbitrary Java anlaysis and transformation tools. (This makes the tool a "complete" metaprogramming tool, in that it can inspect any language structure, or change any language structure, as opposed to say template metaprogramming or reflection). We believe this to be much more effective than ad hoc tools because you don't have to build the infrastructure, the infrastructure provided is robust and handles cases you don't have the energy to implement, and it is designed to support such tasks. YMMV.
DMS/Java Front end have been used to construct a variety of Java tools: test coverage, profilers, dead code elimination, clone detection on scale, JavaDoc with hyperlinked source-code, fast XML parser/generators, etc.
Yes, its actively maintained; undergoing continuous enhancement since the first version in 1998.
There's a Java metaprogramming framework that is part of Tapestry IOC, it's called Plastic. It munges class bytecodes using custom classloaders, I haven't tried it yet but it looks like it gives a simple interface that still enables the programmer to make powerful metaprogramming changes.
Check out the Meta Programming System:
http://www.jetbrains.com/mps/
It has great IDE support and is used quite frequently by the smart folks at JetBrains.
Check out Spring Roo.

What are the risks with Project Lombok?

I'm coming up with performance goals for the new year, and I thought I'd be fun to put a goal to reduce the size of the code base, especially boilerplate. One action I've come up with to address this is to use Project Lombok to make beans as short as they should be. But I have a habit of overlooking downsides to new software and approaches, so I'm relying on the Stack Overflow community: Can anyone tell me why Lombok is a bad idea?
A limitation of Lombok is the fact that it is closely tied to the java compiler. Since the annotation processor API only allows creation of new files during the compilation (and not the modification of the existing files) lombok uses that API as a entry point to modify the java compiler. Unfortunately these modifications of the compiler make heavy usage of non-public APIs. Using lombok may be a good idea but you must be aware that upgrading your compiler may break your code. The probability is low but I always feel uncomfortable using non-public APIs.
In my opinion source code in "Java+Lombok" is not Java source code anymore. I see it as something similar Borland company made many years ago in their Borland C++ Builder IDE for VCL - they introduced "properties" in C++ code effectively introducing some kind of a new programming language that wasn't C++ anymore (not C++ in sense of standard of C++ language). Sources using "Java+Lombok" are not valid sources in sense of Java language specification. Moreover I think annotations were not designed to influence language semantic.
A major downside is IDE support. Since Lombok is not actually a language change, and since your IDE only understands java, you will need an IDE that supports Lombok to get things working right. As of now, that's only Eclipse that includes Eclipse and IntelliJ. If you use eclipse that might be ok, but remember that you are making a decision for future developers as well.
I'd suggest you consider moving some of your code into a less ceremonial language such as groovy. We've had success moving some of our business logic and models into groovy and it works really smoothly.
One potential disadvantage to something like Lombok is that with the setters/getters "missing", source tools may not "recognize" aspects of the resulting object that give it "bean" qualities, since those qualities only manifest in the compiled class.
Another disadvantage is that it's Yet Another piece of "black magic" within the tool chain. Fortunately, it seems to be a rather benign piece (I have not used it), and the fact that it happens at compile time rather than runtime is actually a blessing (IMHO). But, you're not going to be able to reuse or share your code without the project since it's adding artifacts to your code base. So, while the compiled class file may be a "POJO", I'd argue that your source code is NOT a POJO.
Neither of these are crippling downsides, rather just aspects to be aware of looking forward.
As pointed out by user #Jcs in another answer, i would like to add more.
In our project we, are using mapstruct which is used to generate mapper classes, before the code is compiled, using mvn generate-sources command, this is done at process phase using maven processor plugin.
project lombok adds the bytecode for the getter/setter in class file at compile phase.
since process phase is executed before the compile, it finds that there is no getter/setter available in class.
There are some workarounds available to execute compile phase more than one.
See this git hub ticket for more details.
Note : I am using STS ide by Spring and it is supported by lombok :)
It's a third party library, and there are developers who don't know it well.
IDE should support annotations processing (there are plugins for IDEA and Eclipse).
As was mentioned above, your code will be without getters/setters. It leads to sonar/checkstyle violations.
In my opinion, The most obvious risk with Project Lombok is that when you decide to use lombok, you also decide that everyone else who deals with your code uses lombok. This is a true statement for all libraries, but Lombok is special in that it is a build-time dependency and your IDE needs plugins to figure out what is going on. That means anyone who has reason to touch your code ex. someone trying to debug weird behavior, etc.) needs to know how to set it up and how it works. That can be a little frustrating.
To add to other responses.
The main reason to not use it is a new record keyword added as experimental feature in Java 14. Java 16 brings records out of preview which will make project Lombok obsolete in most cases.
Since Java 14 one is able able to write:
record Book(String title, String author, String isbn);
which gives automatically access to the constructor, getters/setter, hashCode, equals and toString methods without any annotations.

Is static metaprogramming possible in Java?

I am a fan of static metaprogramming in C++. I know Java now has generics. Does this mean that static metaprogramming (i.e., compile-time program execution) is possible in Java? If so, can anyone recommend any good resources where one can learn more about it?
No, this is not possible. Generics are not as powerful as templates. For instance, a template argument can be a user-defined type, a primitive type, or a value; but a generic template argument can only be Object or a subtype thereof.
Edit: This is an old answer; since 2011 we have Java 7, which has Annotations that can be used for such trickery.
The short answer
This question is nearly more than 10 years old, but I am still missing one answer to this. And this is: yes, but not because of generics and note quite the same as C++.
As of Java 6, we have the pluggable annotation processing api. Static metaprogramming is (as you already stated in your question)
compile-time program execution
If you know about metaprogramming, then you also know that this is not really true, but for the sake of simplicity, we will use this. Please look here if you want to learn more about metaprogramming in general.
The pluggable annotation processing api is called by the compiler, right after the .java files are read but before the compiler writes the byte-code to the .class files. (I had one source for this, but i cannot find it anymore.. maybe someone can help me out here?).
It allows you, to do logic at compile time with pure java-code. However, the world you are coding in is quite different. Not specifically bad or anything, just different. The classes you are analyzing do not yet exist and you are working on meta data of the classes. But the compiler is run in a JVM, which means you can also create classes and program normally. But furthermore, you can analyze generics, because our annotation processor is called before type erasure.
The main gist about static metaprogramming in java is, that you provide meta-data (in form of annotations) and the processor will be able to find all annotated classes to process them. On (more easy) example can be found on Baeldung, where an easy example is formed. In my opinion, this is quite a good source for getting started. If you understand this, try to google yourself. There are multiple good sources out there, to much to list here. Also take a look at Google AutoService, which utilizes an annotation processor, to take away your hassle of creating and maintaining the service files. If you want to create classes, i recommend looking at JavaPoet.
Sadly though, this API does not allow us, to manipulate source code. But if you really want to, you should take a look at Project Lombok. They do it, but it is not supported.
Why is this important (Further reading for the interested ones among you)
TL;DR: It is quite baffling to me, why we don't use static metaprogramming as much as dynamic, because it has many many advantages.
Most developers see "Dynamic and Static" and immediately jump to the conclusion that dynamic is better. Nothing wrong with that, static has a lot of negative connotations for developers. But in this case (and specifically for java) this is the exact other way around.
Dynamic metaprogramming requires reflections, which has some major drawbacks. There are quite a lot of them. In short: Performance, Security, and Design.
Static metaprogramming (i.e. Annotation Processing) allows us to intersect the compiler, which already does most of the things we try to accomplish with reflections. We can also create classes in this process, which are again passed to the annotation processors. You then can (for example) generate classes, which do what normally had to be done using reflections. Further more, we can implement a "fail fast" system, because we can inform the compiler about errors, warnings and such.
To conclude and compare as much as possible: let us imagine Spring. Spring tries to find all Component annotated classes at runtime (which we could simplify by using service files at compile time), then generates certain proxy classes (which we already could have done at compile time) and resolves bean dependencies (which, again, we already could have done at compile time). Jake Whartons talk about Dagger2, in which he explains why they switched to static metaprogramming. I still don't understand why the big players like Spring don't use it.
This post is to short to fully explain those differences and why static would be more powerful. If you want, i am currently working on a presentation for this. If you are interested and speak German (sorry about that), you can have a look at my website. There you find a presentation, which tries to explain the differences in 45 minutes. Only the slides though.
Take a look at Clojure. It's a LISP with Macros (meta-programming) that runs on the JVM and is very interoperable with Java.
What do you exactly mean by "static metaprogramming"? Yes, C++ template metaprogramming is impossible in Java, but it offers other methods, much more powerful than those from C++:
reflection
aspect-oriented programming (#AspectJ)
bytecode manipulation (Javassist, ObjectWeb ASM, Java agents)
code generation (Annotation Processing Tool, template engines like Velocity)
Abstract Syntax Tree manipulations (APIs provided by popular IDEs)
possibility to run Java compiler and use compiled code even at runtime
There's no best method: each of those methods has its strengths and weaknesses.
Due to flexibility of JVM, all of those methods in Java can be used both at compilation time and runtime.
No. Even more, generic types are erased to their upper bound by the compiler, so you cannot create a new instance of a generic type T at runtime.
The best way to do metaprogamming in Java is to circumvent the type erasure and hand in the Class<T> object of your type T. Still, this is only a hack.
If you need powerful compile-time logic for Java, one way to do that is with some kind of code generation. Since, as other posters have pointed out, the Java language doesn't provide any features suitable for doing compile-time logic, this may be your best option (iff you really do have a need for compile-time logic). Once you have exhausted the other possibilities and you are sure you want to do code-generation, you might be interested in my open source project Rjava, available at:
http://www.github.com/blak3mill3r
It is a Java code generation library written in Ruby, which I wrote in order to generate Google Web Toolkit interfaces for Ruby on Rails applications automatically. It has proved quite handy for that.
As a warning, it can be very difficult to debug Rjava code, Rjava doesn't do much checking, it just assumes you know what you're doing. That's pretty much the state of static metaprogramming anyway. I'd say it's significantly easier to debug than anything non-trivial done with C++ TMP, and it is possible to use it for the same kinds of things.
Anyway, if you were considering writing a program which outputs Java source code, stop right now and check out Rjava. It might not do what you want yet, but it's MIT licensed, so feel free to improve it, deep fry it, or sell it to your grandma. I'd be glad to have other devs who are experienced with generic programming to comment on the design.
Lombok offers a weak form of compile time metaprogramming. However, the technique they use is completely general.
See Java code transform at compile time for a related discussion
You can use a metaprogramming library for Java such as Spoon: https://github.com/INRIA/spoon/
No, generics in Java is purely a way to avoid casting of Object.
In a very reduced sense, maybe?
http://michid.wordpress.com/2008/08/13/type-safe-builder-pattern-in-java/

Categories

Resources