I'm new to Go and also new to the concept of reflection, but should and can the usage of reflect package be avoided in Go? Is there a scenario where reflect is unavoidable?
There are a few problem domains where reflection makes it easier to write reusable libraries:
marshalling/unmarshalling, plenty of examples in the standard library, e.g. encoding/json, encoding/xml
formatting, e.g. text/template, html/template, fmt.Printf.
However there is a price you pay for using reflection:
compile time errors become runtime errors (e.g. fmt.Printf("%d", stringVariable))
performance becomes worse
Very often an alternative solution exists that do not require reflection such as code generation, that is used by marshalling libraries like protobuf or thrift.
I agree with #volker that you should use reflection only when you know that it will simplify already existing code and aware of all downsides.
You should avoid reflection.
Some packages (e.g. fmt) cannot be implemented without reflection as you cannot typeswitch on all existing and upcoming types.
If you are new to Go: Keep away from reflection.
Related
Have a question, never used roslyn before so i'm wondering about maybe experimenting it in a task that i would normally use reflection.
I'm given an external dll, i need to go over some classes in that dll and extract some metadata on them.
Like for example, the class name, property names and types and such.
I would normally use reflection to do it. Should be a super simple task.
But i've been told that this can be achieved using roslyn.
Can it? From what i'm seeing, Roslyn can parse a class but i need to give it the code that represents this calss as text. How would i get the code as text in an already complied code?
Is that even a reasonable scenario to use roslyn? Does it worth the effort?
Thanx!
If all you need is information that's already easily available via reflection, then Roslyn is likely to make it much harder. There's quite a lot of setup required, which can be error-prone and brittle in the face of new releases, in my experience.
I would typically use reflection for anything where the starting point is an assembly. When the starting point is source code, that's when it makes more sense to use Roslyn.
When Roslyn is the right tool for the job, it's amazing - but it doesn't sound like that's the case here.
When you using Roslyn, you have lexical info and symbolic info.
The lexical info won't help you, you must use the symbolic info, for that, you must have compilation and you can create it for compiled code.
With the compilation, you indeed can achieve types info but not runtime info. Anyway, using reflection for this is much straight forward.
When your mission is related to tree transverse or syntax rewriting, Roslyn is perfect, but for metadata info, it's the wrong usage.
It depends on your specific needs but maybe there are other "tools" that more suitable for your task (e.g. cecil or dnlib)
I'm exploring both Go and Entity-Component-Systems. I understand how ECS works, and I'm trying to replicate what seems to be the go-to document of ECS, namely http://cowboyprogramming.com/2007/01/05/evolve-your-heirachy/
For performance, the document recommends to use static arrays of every component type. That is, not arrays of component interfaces (arrays of pointers). The problem with this in Go is circular imports.
I have one package, ecs, which contains the definitions for Entity, Component and System types/interfaces as well as an EntityManager. Another package, ecs/components, contains the various components. Obviously, the ecs/components package depends on ecs. But, to declare arrays of specific components in EntityManager, ecs would depend on ecs/components, therefore creating a circular import.
Is there any way of avoiding this? I am aware that normally a high level system should not depend on lower systems. I'm also want to point out that using an array of pointers is probably fast enough for my purposes, but I'm interested in possible workarounds (for future reference)
Thank you for your help!
For performance, the document recommends to use static arrays of every
component type.
I'm just going to start off saying that I may be blind, but I ctrl+f'd and read that document multiple times and didn't see anything close to that. (Certainly some optimizations could be made this way with regards to things like avoiding cache misses, but I'm dubious it in any way outweighs the clerical overhead).
There's the easy answer to the exact question you asked first, the . import. Any package with an import statement like import . "some/other/package" will treat that package's contents as its own, ignoring circular dependencies. Don't do this.
Unfortunately, without merging the packages, you won't be able to do this (without using interfaces, I mean). Don't fear, though. The article you posted explicitly says this under "implementation details".
Giving each component a common interface means deriving from a base
class with virtual functions. This introduces some additional
overhead. Do not let this turn you against the idea, as the additional
overhead is small, compared to the savings due to simplification of
objects.
It's outright telling you to use interfaces (okay, C++ virtual inheritance, but close enough). It's okay, it's necessary. Especially if you want two slightly different AI components or something, it's a godsend then.
Languages such as Java and PHP support reflection, which allows objects to provide metadata about themselves. Are there any legitimate use cases where you would need to be able to do something like ask an object what methods it has outside of the realm of reverse engineering? Are any of those use cases actually implemented today?
Reflection is used extensively in Java by frameworks which are leveraged at runtime to operate with other code dynamically. Without reflection, all links between code must be done at compile time (statically).
So, for example, any useful plug-in framework (OSGi, JSPF, JPF), leverages Reflection. Any injection framework (Spring, Guice, etc) leverages Reflection.
Any time you want to write a piece of code that will interact with another piece of code without having that piece of code available when compiling, Reflection is the way forward in Java.
However, this is best left to frameworks and should be encapsulated.
There certainly are good use cases. For example, obtaining developer-provided metadata. Java APIs are increasingly using annotations to provide info about methods/fields/classes and their use. Like input validation, binding to data representations... You could use these at compile-time to generate metadata descriptors and use those, but to do it at runtime would require reflection. Even if you used the metadata descriptors, they'd end up containing things like class, method and field names that'd need to be accessed via reflection.
Another use case: dynamic languages. Take Ruby... It allows you to check up-front whether an object would respond to a method name before trying to call that method. Something like that requires reflection.
Or how about when a class or method name must be provided from outside compiled code, like when selecting an implementation of some API. That's just gonna be a bit of text. Looking up what it resolves to comes down to reflection.
Frameworks like Spring or Hibernate make extensive use of reflection to inspect a class and see the annotations.
Frameworks for debugging, serialization, logging, testing...
As a followup to this question about Scala's #BeanProperty generating change events: What would it take to fully implement the behavior that annotating a var field with some custom annotation (#Property, for instance) would generate the code needed to fire property change events? The only way to do it is to write a compiler plugin, right?
More generally: is there a standard way (like in Java) to process annotations in the compiler in Scala?
It may be possible to do this with a proxy, just as you would in Java, possibly involving cglib or similar. A compiler plugin could also do this (as you rightly state), but might be a bit overkill if you're writing anything less than a general-purpose library!
A far better solution would be to manually write the getter and setter methods by hand so that they emit these events, if you're concerned about keeping code clean then these could always be moved into a trait.
For a slightly different approach to the problem, Naftoli Gugenhem has a "reactive" library on GitHub to help with Functional Reactive Programming, arguably a better paradigm than the event-driven model of observable properties.
The ObservableBuffer class is a good place to start looking.
Somebody that I work with and respect once remarked to me that there shouldn't be any need for the use of reflection in application code and that it should only be used in frameworks. He was speaking from a J2EE background and my professional experience of that platform does generally bear that out; although I have written reflective application code using Java once or twice.
My experience of Ruby on Rails is radically different, because Ruby pretty much encourages you to write dynamic code. Much of what Rails gives you simply wouldn't be possible without reflection and metaprogramming and many of the same techniques are equally as applicable and useful to your application code.
Do you agree with the viewpoint that reflection is for frameworks only? I'd be interested to hear your opinions and experiences.
There's the old joke that any sufficiently sophisticated system written in a statically-typed language contains an incomplete, inferior implementation of Lisp.
Since your requirements tend to become more complicated as a project evolves, you often eventually find that the common idioms in statically-typed object systems eventually hit a wall. Sometimes reaching for reflection is the best solution.
I'm happy in dynamically-typed languages like Ruby, and statically-typed languages like C#, but the implicit reflection in Ruby often makes for simpler, easier-to-read code. (Depending on the metaprogramming magic required, sometimes harder to write).
In C#, I've found problems that couldn't be solved without reflection, because of information I didn't have until runtime. One example: When trying to manipulate some third-party code that generated proxies to Silverlight objects running in another process, I had to use reflection to invoke a specific strongly-typed "Generic" version of a method, because the marshalling required the caller to make an assumption about the type of the object in the other process was in order to extract the data we needed from it, and C# doesn't allow the "type" of the generic method invocation to be specified at run time (except with reflection techniques). I guess you could argue our tool was kind of a framework, but I could easily imagine a case in an ordinary application facing a similar problem.
Reflection makes DRY a lot easier. It's certainly possible to write DRY code without reflection, but it's often much more verbose.
If some piece of information is encoded in my program in one way, why wouldn't I use reflection to get at it, if that's the easiest way?
It sounds like he's talking about Java specifically. And in that case, he's just citing a special case of this: in Java, reflection is so wonky it's almost never the easiest way to do something. :-) In other languages like Ruby, as you've seen, it often is.
Reflection is definitely heavily used in frameworks, but when used correctly can help simplify code in applications.
One example I've seen before is using a JDK Proxy of a large interface (20+ methods) to wrap (i.e. delegate to) a specific implementation. Only a couple of methods were overridden using a InvocationHandler, the rest of the methods were invoked via reflection.
Reflection can be useful, but it is slower that doing a regular method call. See this reflection comparison.
Reflection in Java is generally not necessary. It may be the quickest way to solve a certain problem, but I would rather work out the underlying problem that causes you to think it's necessary in app code. I believe this because it frequently pushes errors from compile time to run time, which is always a Bad Thing for large enough software that testing is non-trivial.
I disagree, my application uses reflection to dynamically create providers. I might also use reflection to control logic flow, if the logic is simple and doesn't warrant a more complicated pattern.
In C# I use reflection to grab attributes off Enumeration which help me determine how to display an enumeration to an end user.
I disagree, reflection is very useful in application code and I find myself using it quite often. Most recently, I had to use reflection to load an assembly (in order to investigate its public types) from just the path of the assembly.
Several opinions on this subject are expressed here...
What is reflection and why is it useful?
Use reflection when there is no other way! This is a matter of performance!
If you have looked into .NET performance pitfalls before, it might not surprise you how slow the normal reflection is: a simple test with repeated access to an int property proved to be ~1000 times slower using reflection compared to the direct access to the property (comparing the average of the median 80% of the measured times).
See this: .NET reflection - performance
MSDN has a pretty nice article about When Should You Use Reflection?
If your problem is best solved by using reflection, you should use it.
(Note that the definition of 'best' is something learnt by experience :)
The definition of framework vs. application isn't all that black & white either. Sometimes your app needs a bit of framework to do its job well.
I think the observation that there shouldn't be any need for the use of reflection in application code and that it should only be used in frameworks is more or less true.
On the spectrum of how coupled some piece of code are, code joined by reflection are as loosely coupled as they come.
As such, the code which is doing it's job via reflection can quite happily fulfil it's role in life knowing not-a-thing about the code which is using it.