C# "dynamic," Part II

re: C# "dynamic," Part II

Keith J. Farmer — Tue, 28 Oct 2008 20:41:30 GMT

Hey, Chris -- do you think you could explain the parameter argument to GetMetaObject, and how/why it's supposed to be used?

re: C# "dynamic," Part II

cburrows — Wed, 29 Oct 2008 00:45:14 GMT

Hi Keith,

The DLR uses expression trees internally to communicate rules and actions that are the result of binding particular operations. The parameter parameter is a ParameterExpression that represents the target object.

I would prefer to spend my time talking about C#, so I'm not going to pursue the details of the DLR implementation much deeper than that. Unfortunately I also don't have any PDC-era pointers for you, although Martin Maly's blog is a great resource (http://blogs.msdn.com/mmaly/). You should install the CTP and debug around. The code I've posted compiles and works with it.

chris

re: C# "dynamic," Part II

kfarmer — Wed, 29 Oct 2008 11:25:45 GMT

Unfortunately, Martin's blog is absent this info. I'll see if I can prod him or his cohorts about it if I remember. It's been a while since I was active in DLR-land (ie, shortly after I joined MSFT).

Back to C# itself, part of why I ask, actually, is in response to the known limitation dynamic method binding in C# not being able to bind extension methods. I'm pondering what sort of workarounds would be possible, if any.

For that matter, I wonder why it's even a limitation? The compiler *could* embed information about what extension methods would have been in-scope at that location during the original compile, and pass that to the late-binding algorithm. Or is this the DLR's limitation? (I remember looking into it a while ago, and almost had it hacked into submission)

Or is it just below the feature bar, like yield foreach? :)

C# "dynamic," Part III

Chris Burrows' Blog — Wed, 29 Oct 2008 15:00:16 GMT

Here are a few good resources that you ought to look at for information about dynamic from PDC: Anders

re: C# "dynamic," Part II

RednaxelaFX — Wed, 29 Oct 2008 22:33:30 GMT

@Keith

My guess that the reason of not being able to bind extension methods is because this piece of info of whether an extension method is in scope of not is not available to the binders at runtime. The default binder for CLR objects can only find out what members a type has, and extension methods are not a part of the type to be extended; there isn't quite a way to pass this piece of info into the DLR in the way it is designed now.

Well, someone might be able to get a smarter binder to do this, of course. If a binder's constructor can take a list of possibly applicable extension methods' MethodInfo (which the C# compiler would know at compile time), then it's possible for the binder to get extension methods working.

re: C# "dynamic," Part II

Keith J. Farmer — Thu, 30 Oct 2008 06:02:54 GMT

@RednaxelaFX

Your second paragraph was the point I was trying to make. Having spelunked through earlier releases about a year ago, I know roughly where the DLR would need that information. It's just a matter of putting it there and, of course, testing it to within an inch of someone else's life.

That, unfortunately, I think would be the biggest blocker in having this capability in C#4.

re: C# "dynamic," Part II

RednaxelaFX — Thu, 30 Oct 2008 09:56:08 GMT

@Keith

Well, I'd like to put it this way:

Think about it, in IronPython, Python types that map onto plain CLR types can have extensions over them; and the same is with IronRuby's Ruby types and all other languages based on the DLR, except...C# and VB I guess. The reason for this is that C# and DLR make extension types in a different way. C# extension methods are just syntactic sugar resolved at compile time and doesn't really carry into runtime, where as IronPython extension methods for Python types register themselves to the binder with attributes. Thus C# extension methods are only available in a certain scope at compile time (which is somehow a good thing, less confusion), while Python extensions affect the whole Python part of the program.

If one wants the same semantics as what IronPython extensions give, he/she can simply implement extension methods simillar to the way IronPython does today -- with the restriction that the target type to be extended has to implement IDynamicObject for custom dynamic lookups, which, probably turns into inheriting from System.Dynamic.DynamicObject, which might not be feasible.

The other way around, the C# compiler will have to generate a list of possibly applicable extension methods in scope, and pass that list into the call payload at the call site, so that C# binder has a chance of picking it up later at runtime. Haven't thought about the implications of this; it might imply further complicated scoping rules, and that'd be a problem.

@Chris

I'd like to ask a question: where's that System.Dynamic.DynamicObject class in Anders' and Jim's talk? Didn't find it in System.Core.dll in the CTP. Tried the IDO impl sample from C# Future and it worked, though.

- RednaxelaFX (Kris Mok)

re: C# "dynamic," Part II

cburrows — Thu, 30 Oct 2008 21:09:00 GMT

Re: the lack of support for extension methods. You're both on the right track. Extension methods are a compile-time feature that would have required us to push some context into the C# Runtime Binder. It wouldn't have been impossible, but consider that extension methods are often used in LINQ scenarios and that there, we have a bigger problem of converting lambdas to "dynamic." So we decided to punt on this for C# 4. All decisions like this are difficult and complicated.

re: C# "dynamic," Part II

cburrows — Thu, 30 Oct 2008 21:10:02 GMT

RednaxelaFX, DynamicObject is not in the CTP. I mentioned that in Part III. Sorry!

re: C# "dynamic," Part II

Tobi — Mon, 03 Nov 2008 22:49:19 GMT

Hey RednaxelaFX - if you want to look at a more detailed example on how to implement IDynamicObject take a look at my post http://saftsack.fs.uni-bayreuth.de/~dun3/archives/first-look-ducktyping-c-4-0-idynamicobject-metaobject/202.html - I don't go into details, but it might help you figure the parameters out.

Cheers,

Tobi

re: C# "dynamic," Part II

RednaxelaFX — Mon, 03 Nov 2008 23:52:10 GMT

@Tobi

Thanks for the link. :-)

I didn't have any problems with understanding IDynamicObject and implementing my own, though...in fact I implement it a few times already. I implemented my first one back in May for my bachelor DP. Later on, the whole IDynamicObject interface changed and got deprecated as the IOldDynamicObject, and a newer version of IDynamicObject appeared, with a clearer MetaObject Protocol.

I'm writing a converter in a different perspective from yours, and I'll see if I can clean the code up a little so that I can write a post about it. Thanks again for the link, didn't think of using the Castle library before.

- RednaxelaFX

Community Convergence XLVII

Charlie Calvert's Community Blog — Tue, 04 Nov 2008 00:11:45 GMT

Welcome to the 47th Community Convergence. We had a very successful trip to PDC this year. In this post

re: C# "dynamic," Part II

Tobi — Tue, 04 Nov 2008 13:19:41 GMT

@RednaxelaFX

Sorry, totally phased out and copied the wrong name. :-D I guess I was thinking about Keith J. Farmer. Sorry. :-D

Anyway, I am looking forward to your post about the converter. I would be glad, if you could drop me a link!

Cheers,

Tobi

C# "dynamic," Part IV

Chris Burrows' Blog — Thu, 06 Nov 2008 20:45:31 GMT

Today, let's geek out about the language design regarding the dynamic type. Type in the language vs.

re: C# "dynamic," Part II

KeithH — Tue, 11 Nov 2008 02:59:30 GMT

I'm messing around with the VS 2010 CTP but I can't find the System.Dynamic namespace. What assembly is this in?

re: C# "dynamic," Part II

RednaxelaFX — Tue, 11 Nov 2008 03:28:10 GMT

@KeithH,

It's not there in the CTP. But you may want to check out the lastest version of DLR, which has this System.Dynamic namespace. You can find it in the source drops on IronPython's CodePlex site.

C# "dynamic," Part V

Chris Burrows' Blog — Tue, 11 Nov 2008 05:04:26 GMT

Let's look at this: dynamic d = null ; object o = d; // not an implicit conversion Last time , I said

re: C# "dynamic," Part II

ignu — Wed, 12 Nov 2008 18:27:42 GMT

I'm having a difficult time grocking MetaObject Call, and getting it to actually return an object.

http://stackoverflow.com/questions/283143/can-i-implement-methodmissing-in-c-4-and-have-it-actually-return-a-value

C# "dynamic," Part VI

Chris Burrows' Blog — Fri, 14 Nov 2008 17:20:58 GMT

We left off last time with this piece of code public class C { public static void M( int i) { } public

C# Dynamic - CSharp's new feature of the coming version 4.0

Journal of Abu Sayed Mohammad Ismail — Mon, 15 Dec 2008 21:01:36 GMT

Very good resources for the coming version... Sam Ng Dynamic in C# Part One Dynamic in C# Part Two Chris

C# “dynamic,” Part VII

Chris Burrows' Blog — Wed, 04 Feb 2009 03:35:48 GMT

It has been a while since I posted anything here, and I have the same excuse everyone else does. Work,