Eclipse's JDT compiler provide an interface INameEnvironment which defines method findType(...) enable you to do cascade compilation. Curiously I would like to know if there are any means to do it using standard JDK compiler toolkit?
Note, the scenario is a template engine which do in memory compilation for template file generated classes which have inter-dependencies, and it cannot forecast the order you encountered a template file, thus Foo might needs to be compiled first before it's parent Bar compiled already, therefore you need a mechanism to do cascade compilation, meaning during compilation of Foo you need to generate another source Bar and compile it first in order to continue Foo's compilation: some code like the follows:
private NameEnvironmentAnswer findType(final String name) {
try {
if (!name.contains(TemplateClass.CN_SUFFIX)) {
return findStandType(name);
}
char[] fileName = name.toCharArray();
TemplateClass templateClass = classCache.getByClassName(name);
// TemplateClass exists
if (templateClass != null) {
if (templateClass.javaByteCode != null) {
ClassFileReader classFileReader = new ClassFileReader(templateClass.javaByteCode, fileName, true);
return new NameEnvironmentAnswer(classFileReader, null);
}
// Cascade compilation
ICompilationUnit compilationUnit = new CompilationUnit(name);
return new NameEnvironmentAnswer(compilationUnit, null);
}
// So it's a standard class
return findStandType(name);
} catch (ClassFormatException e) {
// Something very very bad
throw new RuntimeException(e);
}
}
Based on our comment conversation, I think the answer is a clear: no, you can't do that with the JDK compiler. It does give you a hook when it requests the package, but not the specific class dependency.
About as close as you can get so far as I know is:
Here's a nice article with code though it needs to be adapted to handle in memory classes. Specifically the issue you're describing is handled by the JavaFileManager.list(...) method. You have to return back JavaFileObjects here that you have cached in memory. You'll most likely need to create a subclass of ForwardingJavaFileManager as described in the article--though modified to handle the cached classes you are working with.
You can use that to compile something. If it returns errors, use regex to find out what is missing. After generating/compiling the code for the missing thing, retry compiling the original code.
NOTE: It does ask for the FQN of the dependent class as the packageName argument in ForwardingFileManager.list(...) at some point. I haven't tried to return the class at that point. It might not work because the package would mismatch, but maybe it would.
Try reading through this HelloWorld example to see if it solves your problem. Without posting the code, it's hard to say what your specific issue is.
Related
There is dsl.DefaultExpander and dsl.DefaultExpanderResolver. Design and javadoc gives an idea I can add my custom expander to the flow. But I didn't found real possibility to do that. Do you see how could I achieve DSL to parsed with my custom expander?
Well, in KnowledgeBuilderImpl (the class used to compile DRL/DSL) I see the following method:
public DefaultExpander getDslExpander() {
DefaultExpander expander = new DefaultExpander();
if (this.dslFiles == null || this.dslFiles.isEmpty()) {
return null;
}
for (DSLMappingFile file : this.dslFiles) {
expander.addDSLMapping(file.getMapping());
}
return expander;
}
As you can see, DefaultExpander is hardcoded in there, so I guess there is no way to specify any other implementation.
Same thing happens in DrlParser.getDefaultResolver() where a DefaultExpanderResolver is always instantiated no matters what.
In my opinion, the only way to go is to pre-generate your DRL (outside Drools) based on your own semantics and then to feed the generated DRL to Drools.
Hope it helps,
Update: Oracle has confirmed this as a bug.
Summary: Certain custom BeanInfos and PropertyDescriptors that work in JDK 1.6 fail in JDK 1.7, and some only fail after Garbage Collection has run and cleared certain SoftReferences.
Edit: This will also break the ExtendedBeanInfo in Spring 3.1 as noted at the bottom of the post.
Edit: If you invoke sections 7.1 or 8.3 of the JavaBeans spec, explain
exactly where those parts of the spec require anything. The
language is not imperative or normative in those sections. The
language in those sections is that of examples, which are at best
ambiguous as a specification. Furthermore, the BeanInfo API
specifically allows one to change the default behavior, and it is
clearly broken in the second example below.
The Java Beans specification looks for default setter methods with a void return type, but it allows customization of the getter and setter methods through a java.beans.PropertyDescriptor. The simplest way to use it has been to specify the names of the getter and setter.
new PropertyDescriptor("foo", MyClass.class, "getFoo", "setFoo");
This has worked in JDK 1.5 and JDK 1.6 to specify the setter name even when its return type is not void as in the test case below:
import java.beans.IntrospectionException;
import java.beans.PropertyDescriptor;
import org.testng.annotations.*;
/**
* Shows what has worked up until JDK 1.7.
*/
public class PropertyDescriptorTest
{
private int i;
public int getI() { return i; }
// A setter that my people call "fluent".
public PropertyDescriptorTest setI(final int i) { this.i = i; return this; }
#Test
public void fluentBeans() throws IntrospectionException
{
// This throws an exception only in JDK 1.7.
final PropertyDescriptor pd = new PropertyDescriptor("i",
PropertyDescriptorTest.class, "getI", "setI");
assert pd.getReadMethod() != null;
assert pd.getWriteMethod() != null;
}
}
The example of custom BeanInfos, which allow the programmatic control of PropertyDescriptors in the Java Beans specification all use void return types for their setters, but nothing in the specification indicates that those examples are normative, and now the behavior of this low-level utility has changed in the new Java classes, which happens to have broken some code on which I am working.
There are numerous changes in in the java.beans package between JDK 1.6 and 1.7, but the one that causes this test to fail appears to be in this diff:
## -240,11 +289,16 ##
}
if (writeMethodName == null) {
- writeMethodName = "set" + getBaseName();
+ writeMethodName = Introspector.SET_PREFIX + getBaseName();
}
- writeMethod = Introspector.findMethod(cls, writeMethodName, 1,
- (type == null) ? null : new Class[] { type });
+ Class[] args = (type == null) ? null : new Class[] { type };
+ writeMethod = Introspector.findMethod(cls, writeMethodName, 1, args);
+ if (writeMethod != null) {
+ if (!writeMethod.getReturnType().equals(void.class)) {
+ writeMethod = null;
+ }
+ }
try {
setWriteMethod(writeMethod);
} catch (IntrospectionException ex) {
Instead of simply accepting the method with the correct name and parameters, the PropertyDescriptor is now also checking the return type to see whether it is null, so the fluent setter no longer gets used. The PropertyDescriptor throws an IntrospectionException in this case: "Method not found: setI".
However, the problem is much more insidious than the simple test above. Another way to specify the getter and setter methods in the PropertyDescriptor for a custom BeanInfo is to use the actual Method objects:
#Test
public void fluentBeansByMethod()
throws IntrospectionException, NoSuchMethodException
{
final Method readMethod = PropertyDescriptorTest.class.getMethod("getI");
final Method writeMethod = PropertyDescriptorTest.class.getMethod("setI",
Integer.TYPE);
final PropertyDescriptor pd = new PropertyDescriptor("i", readMethod,
writeMethod);
assert pd.getReadMethod() != null;
assert pd.getWriteMethod() != null;
}
Now the above code will pass a unit test in both 1.6 and in 1.7, but the code will begin to fail at some point in time during the life of the JVM instance owing to the very same change that causes the first example to fail immediately. In the second example the only indication that anything has gone wrong comes when trying to use the custom PropertyDescriptor. The setter is null, and most utility code takes that to mean that the property is read-only.
The code in the diff is inside PropertyDescriptor.getWriteMethod(). It executes when the SoftReference holding the actual setter Method is empty. This code is invoked by the PropertyDescriptor constructor in the first example that takes the accessor method names above because initially there is no Method saved in the SoftReferences holding the actual getter and setter.
In the second example the read method and write method are stored in SoftReference objects in the PropertyDescriptor by the constructor, and at first these will contain references to the readMethod and writeMethod getter and setter Methods given to the constructor. If at some point those Soft references are cleared as the Garbage Collector is allowed to do (and it will do), then the getWriteMethod() code will see that the SoftReference gives back null, and it will try to discover the setter. This time, using the same code path inside PropertyDescriptor that causes the first example to fail in JDK 1.7, it will set the write Method to null because the return type is not void. (The return type is not part of a Java method signature.)
Having the behavior change like this over time when using a custom BeanInfo can be extremely confusing. Trying to duplicate the conditions that cause the Garbage Collector to clear those particular SoftReferences is also tedious (though maybe some instrumenting mocking may help.)
The Spring ExtendedBeanInfo class has tests similar to those above. Here is an actual Spring 3.1.1 unit test from ExtendedBeanInfoTest that will pass in unit test mode, but the code being tested will fail in the post-GC insidious mode::
#Test
public void nonStandardWriteMethodOnly() throws IntrospectionException {
#SuppressWarnings("unused") class C {
public C setFoo(String foo) { return this; }
}
BeanInfo bi = Introspector.getBeanInfo(C.class);
ExtendedBeanInfo ebi = new ExtendedBeanInfo(bi);
assertThat(hasReadMethodForProperty(bi, "foo"), is(false));
assertThat(hasWriteMethodForProperty(bi, "foo"), is(false));
assertThat(hasReadMethodForProperty(ebi, "foo"), is(false));
assertThat(hasWriteMethodForProperty(ebi, "foo"), is(true));
}
One suggestion is that we can keep the current code working with the non-void setters by preventing the setter methods from being only softly reachable. That seems like it would work, but that is rather a hack around the changed behavior in JDK 1.7.
Q: Is there some definitive specification stating that non-void setters should be anathema? I've found nothing, and I currently consider this a bug in the JDK 1.7 libraries.
Am I wrong, and why?
Looks like the specification hasn't changed (it requires void setter) but the implementation has been updated to only allow void setters.
Specification:
http://www.oracle.com/technetwork/java/javase/documentation/spec-136004.html
More specifically see section 7.1 (accessor methods) and 8.3 (design patterns for simple properties)
See some of the later answers in this stackoverflow question:
Does Java bean's setter permit return this?
Section 8.2 specifies:
However, within Java Beans the use of method and type names that match design patterns is entirely optional. If a programmer is prepared to explicitly specify their properties, methods, and events using the BeanInfo interface then they can call their methods and types whatever they like. However, these methods and types will still have to match the required type signatures, as this is essential to their operation.
(emphasis added)
Also, I beleive the method signatures shown in 7.1 and 8.3 are, in fact, normative. They are examples only in the sense that they use "foo" as an example property name.
I would also opt to say disallowing non-void setters is an error. It simply makes fluent programming impossible. Thats why it needs to be be changed.
Since I found Spring 3.1.1 ExtendedBeanInfo unit tests that expect the code not to be broken, and because having the behavior change after garbage collection is obviously a bug, I shall answer this and note the Java bug numbers. The bugs are still not visible on the external database of Java bugs, but I hope that they will become visible at some point:
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7172854 (Oracle closed this as a duplicate of the bug below since they have the same cause despite different manifestations.)
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7172865
(The bugs were submitted on May 30, 2012.)
As of June 20, 2012, the bugs are visible in the external database via the links above.
I currently use Com4j to talk to iTunes from my Java app, unfortunately it does not work with 64bit Java and looks like it never will, so Im trying to use an alternative called Jacob instead.
Both libraries provide a tool to generate Java classes from a DLL, and the resultant classes are very similar and its been straightforward to change most of the code but Im failing on how to find subtypes
IITPlaylist object = itunes.createFolder(TextLabel.SAVE_ITUNES_PLAYLIST_FOLDER.getMsg());
IITUserPlaylist playlistFolder = object.queryInterface(IITUserPlaylist.class);
Both libraries have created IITPlaylist and IITUSerPlaylist classes but only com4j provides the queryInterface class, and no IITUserPlaylist is not actually a subclass of IITPlaylist.
Also com4j provides an is method, but jacob does not
if (next.is(IITFileOrCDTrack.class))
Anyone know how to resolve these issues ?
EDIT:
Made some progress but still not got it working, there is a QueryInterface method that takes the guid of the class (include the curly brackets) , I found the guid by looking at the jacobgenlog.txt file which is created when you run jacobgen on the iTunes executable
This then returns another Dispatch object that is meant to relate to the subclass, however the simple cast Ive done is invalid, whats the mising step ?
private static final String USER_PLAYLIST_GUID = "{0A504DED-A0B5-465A-8A94-50E20D7DF692}";
IITPlaylist object = itunes.createFolder(TextLabel.SAVE_ITUNES_PLAYLIST_FOLDER.getMsg());
IITUserPlaylist playlistFolder = (IITUserPlaylist)object.QueryInterface(USER_PLAYLIST_GUID);
The is() functionality is replaced by checking the kind
IITTrack next = tracks.getItem(i);
if(next.getKind().equals(ITTrackKind.ITTrackKindFile))
A spanner in the works is that jacobgen getKind() methods are invalid Java because they try to return a new interface, and of course you cannot instantiate an interface, so I had to modify them as follows
ITPlayListKind goes from
public interface ITPlaylistKind extends __MIDL___MIDL_itf_iTunesCOMInterface_0001_0081_0001 {
}
to
public enum ITPlaylistKind {
ITPlaylistKindUnknown,
ITPlaylistKindLibrary,
ITPlaylistKindUser,
ITPlaylistKindCD,
ITPlaylistKindDevice,
ITPlaylistKindRadioTuner;
}
Within IITUserPlaylist
public ITPlaylistKind getKind() {
return new ITPlaylistKind(Dispatch.get(this, "Kind").toDispatch());
}
to
public ITPlaylistKind getKind() {
return ITPlaylistKind.values()[Dispatch.get(this, "Kind").getInt()];
}
this wasnt an original idea by me, I got the idea from http://dot-totally.co.uk/software/itunescon/ which appears to be a modified version of the iTunes classes created by jacobgen, I didnt find it added that much and decided to stick with the jacobgen generated classes.
As soon as I set a bounty I work out the answer for myself.
Simply just use the constructor
IITPlaylist object = itunes.createFolder
(TextLabel.SAVE_ITUNES_PLAYLIST_FOLDER.getMsg());
IITUserPlaylist playlistFolder = new IITUserPlayList(object);
The QueryInterface and GUID sctrings re not required.
I was also having a problem working out how to add a track to a playlist, but you just need to
construct a Variant from the track ( I dont have to do this anywhere else)
IITTrack next = itunes.getLibraryPlaylist().getTracks()
.getItemByPersistentID(persistentId.getHighBit(),
persistentId.getLowBit());
playlist.addTrack(new Variant(nextTrack));
How can I change what a method is doing in Java ?
I mean, I am trying to use annotations to make the following code
#Anno1(Argument = "Option1")
public class TestClass
{
#Anno2
public void test()
{
}
}
Into
public class TestClass
{
private static StaticReference z;
public void test()
{
z.invokeToAll();
}
}
This is a very simplified example of what I am trying to do. Anno1 will have many possible combinations, but this is not my problem so far. My problem is how to add code to method test()
I am looking for a more generic solution if possible. Eg. A way to add every kind of code in the method (not just a way to .invokeToAll())
So far I am using import javax.annotation.processing.*; and I have the following code, but I don't know how to go on from there
private void processMethodAnnotations(RoundEnvironment env)
{
for (Element e : env.getElementsAnnotatedWith(Anno2.class))
{
//If it is a valid annotation over a method
if (e.getKind() == ElementKind.METHOD)
{
//What to do here :S
}else
{
processingEnv.getMessager().printMessage(Diagnostic.Kind.WARNING,"Not a method!", e);
}
}
}
I have found something about Java Reflection but I have not found any source to help me with what I am doing.
Obviously I extends AbstractProcessor in my code
I have found this tutorial (http://www.zdnetasia.com/writing-and-processing-custom-annotations-part-3-39362483.htm) But this concerns creating a new class, not just changing a method. and the javax.lang.model.elements do not provide any way of editing that element (which in my case represents a Method).
I hope my question is clear and inline with the rules. If not please comment and I will clarify. Thanks.
Annotation processing is the wrong way to go for you, from Wikipedia:
When Java source code is compiled,
annotations can be processed by
compiler plug-ins called annotation
processors. Processors can produce
informational messages or create
additional Java source files or
resources, which in turn may be
compiled and processed, but annotation
processors cannot modify the annotated
code itself.
People suggested to you the right way - AOP. Specifically, you can use AspectJ. "Quick result" way is (if you use Eclipse):
Install AJDT (AspectJ Development Tools)
Create an AspectJ project and add there your classes and annotations
Create Aspect:
public aspect Processor {
private StaticReference z;
pointcut generic()
// intercept execution of method named test, annotated with #Anno1
// from any class type, annotated with #Anno2
: execution(#Anno2 * (#Anno1 *).test())
// method takes no arguments
&& args ();
// here you have written what you want the method to actually do
void around () : generic() {
z.invokeToAll();
}
}
now you can execute a test and you will see that it works ;) AJDT compiles code for you automatically, so do not need any manual work to do, hope that's what you called "magic" ;)
UPDATE:
if your code in the test() method depends on the Anno1 annotation value, then inside aspect you can get class annotation for which it is executed this way:
void around () : generic() {
Annotation[] classAnnotations = thisJoinPoint.getThis().getClass().getAnnotations();
String ArgumentValue = null;
for ( Annotation annotation : classAnnotations ) {
if ( annotation instanceof Anno1 ) {
ArgumentValue = ((Anno1) annotation).Argument();
break;
}
}
if ( ArgumentValue != null && ArgumentValue.equals("Option1")) {
z.invokeToAll();
}
}
where thisJoinPoint is a special reference variable.
UPDATE2:
if you want to add System.out.println( this ) in your aspect, you need to write there System.out.println( thisJoinPoint.getThis() ), just tested and it works. thisJoinPoint.getThis() returns you "this" but not exactly; in fact this is Object variable and if you want to get any propery you need either to cast or to use reflection. And thisJoinPoint.getThis() does not provide access to private properties.
Well, now seems that your question is answered, but if I missed anything, or you get additional question/problems with this way - feel free to ask ;)
It's perfectly possible to do what you ask, although there is a caveat: relying on private compiler APIs. Sounds scary, but it isn't really (compiler implementations tend to be stable).
There's a paper that explains the procedure: The Hacker's Guide to Javac.
Notably, this is used by Project Lombok to provide automatic getter/setter generation (amongst other things). The following article explains how it does it, basically re-iterating what is said the aforementioned paper.
Well, you might see if the following boilerplate code will be useful:
public void magic(Object bean, String[] args) throws Exception {
for (Method method : bean.getClass().getDeclaredMethods()) {
if (method.isAnnotationPresent(Anno2.class)) {
// Invoke the original method
method.invoke(bean, args);
// Invoke your 'z' method
StaticReference.invokeAll();
}
}
}
As an alternative your might employ aspect oriented programming, for instance you have the AspectJ project.
I'm not sure at all if it is even possible to change the source or byte code via annotations. From what your describing it looks as if aspect oriented programming could provide a solution to your problem.
Your annotations are pretty similiar to the pointcut concept (they mark a location where code needs to be inserted) and the inserted code is close the advice concept.
Another approach would be parsing the java source file into an abstract syntax tree, modify this AST and serialize to a java compiler input.
If your class extends a suitable interface, you could wrap it in a DynamicProxy, which delegates all calls to the original methods, except the call to test.
There's something very unsatisfactory about this code:
/*
Given a command string in which the first 8 characters are the command name
padded on the right with whitespace, construct the appropriate kind of
Command object.
*/
public class CommandFactory {
public Command getCommand(String cmd) {
cmdName = cmd.subString(0,8).trim();
if(cmdName.equals("START")) {
return new StartCommand(cmd);
}
if(cmdName.equals("END")) {
return new EndCommand(cmd);
}
// ... more commands in more if blocks here
// else it's a bad command.
return new InvalidCommand(cmd);
}
}
I'm unrepentant about the multiple exit points - the structure is clear. But I'm not happy about the series of near-identical if statements. I've considered making a Map of Strings to Commands:
commandMap = new HashMap();
commandMap.put("START",StartCommand.class);
// ... etc.
... then using Reflection to make instances of the appropriate class looked up from the Map. However while conceptually elegant, this involves a fair amount of Reflection code that whoever inherits this code might not appreciate - although that cost might be offset by the benefits. All the lines hardcoding values into the commandMap smell almost as bad as the if block.
Even better would be if the factory's constructor could scan the classpath for subclasses of Command, query them for String representations, and automatically add them them to its repertoire.
So - how should I go about refactoring this?
I guess some of the frameworks out there give me this kind of thing for free. Let's assume I'm not in a position to migrate this stuff into such a framework.
How about the following code:
public enum CommandFactory {
START {
#Override
Command create(String cmd) {
return new StartCommand(cmd);
}
},
END {
#Override
Command create(String cmd) {
return new EndCommand(cmd);
}
};
abstract Command create(String cmd);
public static Command getCommand(String cmd) {
String cmdName = cmd.substring(0, 8).trim();
CommandFactory factory;
try {
factory = valueOf(cmdName);
}
catch (IllegalArgumentException e) {
return new InvalidCommand(cmd);
}
return factory.create(cmd);
}
}
The valueOf(String) of the enum is used to find the correct factory method. If the factory doesn't exist it will throw an IllegalArgumentException. We can use this as a signal to create the InvalidCommand object.
An extra benefit is that if you can make the method create(String cmd) public if you would also make this way of constructing a Command object compile time checked available to the rest of your code. You could then use CommandFactory.START.create(String cmd) to create a Command object.
The last benefit is that you can easily create a list of all available command in your Javadoc documentation.
Your map of strings to commands I think is good. You could even factor out the string command name to the constructor (i.e. shouldn't StartCommand know that its command is "START"?) If you could do this, instantiation of your command objects is much simpler:
Class c = commandMap.get(cmdName);
if (c != null)
return c.newInstance();
else
throw new IllegalArgumentException(cmdName + " is not as valid command");
Another option is to create an enum of all your commands with links to the classes (assume all your command objects implement CommandInterface):
public enum Command
{
START(StartCommand.class),
END(EndCommand.class);
private Class<? extends CommandInterface> mappedClass;
private Command(Class<? extends CommandInterface> c) { mappedClass = c; }
public CommandInterface getInstance()
{
return mappedClass.newInstance();
}
}
since the toString of an enum is its name, you can use EnumSet to locate the right object and get the class from within.
With the exception of the
cmd.subString(0,8).trim();
part, this doesn't look too bad to me. You could go with the Map and use reflection, but, depending on how often you add/change commands, this might not buy you much.
You should probably document why you only want the first 8 characters, or maybe change the protocol so it's easier to figure out which part of that string is the command (e.g. put a marker like ':' or ';' after the command key-word).
Its not directly an answer to your question, but why don't you throw an InvalidCommandException (or something similar), rather then returning an object of type InvalidCommand?
Unless there is a reason they can't be I always try to make my command implementations stateless. If that's the case you can add a method boolean identifier(String id) method to your command interface which would tell whether this instance could be used for the given string identifier. Then your factory could look something like this (note: I did not compile or test this):
public class CommandFactory {
private static List<Command> commands = new ArrayList<Command>();
public static void registerCommand(Command cmd) {
commands.add(cmd);
}
public Command getCommand(String cmd) {
for(Command instance : commands) {
if(instance.identifier(cmd)) {
return cmd;
}
}
throw new CommandNotRegisteredException(cmd);
}
}
I like your idea, but if you want to avoid reflection you could add instead instances to the HashMap:
commandMap = new HashMap();
commandMap.put("START",new StartCommand());
Whenever you need a command, you just clone it:
command = ((Command) commandMap.get(cmdName)).clone();
And afterwards, you set the command string:
command.setCommandString(cmdName);
But using clone() doesn't sound as elegant as using reflection :(
Taking a Convetion vs Configuration approach and using reflection to scan for available Command objects and loading them into your map would be the way to go. You then have the ability to expose new Commands without a recompile of the factory.
Another approach to dynamically finding the class to load, would be to omit the explicit map, and just try to build the class name from the command string. A title case and concatenate algorithm could turn "START" -> "com.mypackage.commands.StartCommand", and just use reflection to try to instantiate it. Fail somehow (InvalidCommand instance or an Exception of your own) if you can't find the class.
Then you add commands just by adding one object and start using it.
One option would be for each command type to have its own factory. This gives you two advantages:
1) Your generic factory wouldn't call new. So each command type could in future return an object of a different class according to the arguments following the space padding in the string.
2) In your HashMap scheme, you could avoid reflection by, for each command class, mapping to an object implementing a SpecialisedCommandFactory interface, instead of mapping to the class itself. This object in practice would probably be a singleton, but need not be specified as such. Your generic getCommand then calls the specialised getCommand.
That said, factory proliferation can get out of hand, and the code you have is the simplest thing that does the job. Personally I'd probably leave it as it is: you can compare command lists in source and spec without non-local considerations like what might have previously called CommandFactory.registerCommand, or what classes have been discovered through reflection. It's not confusing. It's very unlikely to be slow for less than a thousand commands. The only problem is that you can't add new command types without modifying the factory. But the modification you'd make is simple and repetitive, and if you forget to make it you get an obvious error for command lines containing the new type, so it's not onerous.
Having this repetitive object creation code all hidden in the factory is not so bad. If it has to be done somewhere, at least it's all here, so I'd not worry about it too much.
If you really want to do something about it, maybe go for the Map, but configure it from a properties file, and build the map from that props file.
Without going the classpath discovery route (about which I don't know), you'll always be modifying 2 places: writing a class, and then adding a mapping somewhere (factory, map init, or properties file).
Thinking about this, You could create little instantiation classes, like:
class CreateStartCommands implements CommandCreator {
public bool is_fitting_commandstring(String identifier) {
return identifier == "START"
}
public Startcommand create_instance(cmd) {
return StartCommand(cmd);
}
}
Of course, this adds a whole bunch if tiny classes that can't do much more than say "yes, thats start, give me that" or "nope, don't like that", however, you can now rework the factory to contain a list of those CommandCreators and just ask each of it: "you like this command?" and return the result of create_instance of the first accepting CommandCreator. Of course it now looks kind of akward to extract the first 8 characters outside of the CommandCreator, so I would rework that so you pass the entire command string into the CommandCreator.
I think I applied some "Replace switch with polymorphism"-Refactoring here, in case anyone wonders about that.
I'd go for the map and creation via reflection. If scanning the class path is too slow, you can always add a custom annotation to the class, have an annotation processor running at compile time and store all class names in the jar metadata.
Then, the only mistake you can do is forgetting the annotation.
I did something like this a while ago, using maven and APT.
The way I do it is to not have a generic Factory method.
I like to use Domain Objects as my command objects. Since I use Spring MVC this is a great approach since the DataBinder.setAllowedFields method allows me a great deal of flexibility to use a single domain object for several different forms.
To get a command object, I have a static factory method on the Domain object class. For example, in the member class I'd have methods like -
public static Member getCommandObjectForRegistration();
public static Member getCommandObjectForChangePassword();
And so on.
I'm not sure that this is a great approach, I never saw it suggested anywhere and kind of just came up with it on my own b/c I like the idea of keeping things like this in one place. If anybody sees any reason to object please let me know in the comments...
I would suggest avoiding reflection if at all possible. It is somewhat evil.
You can make your code more concise by using the ternary operator:
return
cmdName.equals("START") ? new StartCommand (cmd) :
cmdName.equals("END" ) ? new EndCommand (cmd) :
new InvalidCommand(cmd);
You could introduce an enum. Making each enum constant a factory is verbose and also has some runtime memory cost. But you can eaily lookup an enum and then use that with == or switch.
import xx.example.Command.*;
Command command = Command.valueOf(commandStr);
return
command == START ? new StartCommand (commandLine) :
command == END ? new EndCommand (commandLine) :
new InvalidCommand(commandLine);
Go with your gut, and reflect. However, in this solution, your Command interface is now assumed to have the setCommandString(String s) method accessible, so that newInstance is easily useable. Also, commandMap is any map with String keys (cmd) to Command class instances that they correspond to.
public class CommandFactory {
public Command getCommand(String cmd) {
if(cmd == null) {
return new InvalidCommand(cmd);
}
Class commandClass = (Class) commandMap.get(cmd);
if(commandClass == null) {
return new InvalidCommand(cmd);
}
try {
Command newCommand = (Command) commandClass.newInstance();
newCommand.setCommandString(cmd);
return newCommand;
}
catch(Exception e) {
return new InvalidCommand(cmd);
}
}
Hmm, browsing, and only just came across this. Can I still comment?
IMHO there's nothing wrong with the original if/else block code. This is simple, and simplicity must always be our first call in design (http://c2.com/cgi/wiki?DoTheSimplestThingThatCouldPossiblyWork)
This seems esp true as all the solutions offered are much less self documenting than the original code...I mean shouldn't we write our code for reading rather than translation...
At the very least, your command should have a getCommandString() -- where StartCommand overrides to return "START". Then you can just register or discover the classes.
+1 on the reflection suggestion, it will give you a more sane structure in your class.
Actually you could do the following (if you haven't thought about it already)
create methods corresponding to the String you'd be expecting as an argument to your getCommand() factory method, then all you have to do is reflect and invoke() these methods and return the correct object.