Returns data type name (string) for getTypeStr() in FieldInfo class

[febielgiva]

for a object of FieldInfo the getTypeStr() return the field's data type example "Integer/String/int" but I need to check whether the data type is valid or not .

I need to pass the data type's class to the below function
public static boolean isValidType(Class<?> retType) { if (retType.isPrimitive() && retType != void.class) return true; if (Number.class.isAssignableFrom(retType)) return true; if (AbstractCode.class.isAssignableFrom(retType)) return true; if (Boolean.class == retType) return true; if (Character.class == retType) return true; if (String.class == retType) return true; if (Date.class.isAssignableFrom(retType)) return true; if (byte[].class.isAssignableFrom(retType)) return true; if (Enum.class.isAssignableFrom(retType)) return true; return false; }

But what i have is data type name ( string) which is useless . You should store the class type of the data type like the REFLECT Field variables.

[lukehutch]

Field has methods getType() (which is easy to support, I can add that) and getGenericType() (which is not easy to support, the subclasses of Type do not have public constructors).

However, just adding getType() is incomplete. It also can trigger arbitrary class initializer blocks to run, because it requires the ClassLoader to load classes for some types. (This can probably even lead to MadGadget-like security vulnerabilities in some cases.) Therefore, it's not ideal for classpath scanning. But I guess if I write a big caveat about it in the documentation, maybe people can decide to take this risk for themselves!

[lukehutch]

Oh, it gets worse than that: If the type of the field is say String[], then field.getType() returns Class<String[]>, which it is not possible to construct. So even getType() will be broken some of the time.

What you probably need to do is just load the class whose fields you want to probe, and then use regular reflection to test if the fields have the properties you want. The correct way to load classes given a ScanResult is ScanResult#classNamesToClassRefs(List<String> classNames) or ScanResult#classNameToClassRef(String className).

Does that work for you?

[febielgiva]

okay . Got it will try it out. I am working on speed test to compare between reflection and your library for a project in my company. If your lib is efficient we would migrate to it. So please do consider this as request and implement these in future. Thank you .

[lukehutch]

@febielgiva based on a great suggestion by @Alexander--, the exact method you were looking for is now supported: FieldInfo#getType(). (A similar method exists for methods and annotations on fields or methods.) Can you please test the github master version?

[lukehutch]

@febielgiva Added a unit test that seems to indicate this should be working for you too now, so I'll close this. Please still comment if you do test this, to let me know how it goes for you. I'll also release a new version now. Thanks for this report!

[lukehutch]

PS please also let me know what your benchmark scores were for the two libraries, I'm interested in real-world numbers. Thanks.

[febielgiva]

sure I will let know once I tested it . Thank you for implementing those. its would be really useful. Do you have any method similar to invoke() method in reflect and a method to get just the constructor of a class? .

Thank you .

[lukehutch]

You can't invoke a method in this way, because the class hasn't even been loaded (assuming other code hasn't already loaded the calss) -- that's the whole point of ClassInfo, to store info about the class that is read from the classfile directly, without calling the classloader.

The constructors are called <init> in the classfile binary (there is also <clinit> for the static class initializer). But I rename <init> to the name of the class, since the name of the class is the name of the constructor in Java source. So you can just search for the name of the class among the MethodInfo methods to get the constructors. However, I just realized I am using fully-qualified classnames, rather than just the class' "simple name" (leaf name), so the constructor for String will be java.lang.String (which is not correct or helpful). So -- either I should switch this to use just the simple name (String etc.), or maybe I should switch the constructors back to use <init>, since there are still <clinit> blocks anyway, and those don't have a name in Java source. What is your preference?

[lukehutch]

I went ahead and made these changes:

• Constructors are now named <init>.
• I added ClassInfo#getConstructors() and ClassInfo#getMethodsAndConstructors().
• ClassInfo#getMethods() now does not return constructors, to work similarly to java.lang.reflect.

[lukehutch]

(PS although I closed this, I'm still interested in benchmarks, whenever you have a chance to report on that -- no rush, and thanks!)

[lukehutch]

Released in 2.7.0.