The String Literal Returns

re: The String Literal Returns

Zach Zolton — Wed, 08 Dec 2004 22:33:00 GMT

Mr. Lippman,

I am a mid-twenties programmer working and doing grad school. I wasn't there when you guys were deciding about C++, but in a way I wish I was! I love C++, but at my work I'm using .NET languages.

I really like the uncluttered C# style, and the .NET Framework. I've also had to use VB.NET, which I think is an abomination. VB.NET tried to keep syntactically compatible with VB 6, while adding language "features" for .NET coding. They only created an verbose Frankenstein.

I hope that doesn't that doesn't become analogous to CLI/C++. (I couldn't even get into old Managed C++) There are a lot of great independent C++ libraries that are what help identify the C++ community (of which I'd like to be included).

So, if at the end of this all, I can compile non-Microsoft C++ libraries, and use the .NET Framework - both with ease - I'll declare CLI/C++ a winner... I'm waiting; I want to try and reserve my judgement.

Best of Luck,

Zach Z

re: The String Literal Returns

Fritz Onion — Wed, 08 Dec 2004 23:44:00 GMT

Stan - I just want to know why you're throwing me as an exception? What have I ever done to deserve this?? :) :)

-Fritz

PS - this post alone has convinced me to spend some serious quality time with this next release of managed C++

re: The String Literal Returns

stan lippman — Thu, 09 Dec 2004 05:39:00 GMT

hi, fritz. gee, it's nice to hear from you. i hope you are well :-) let us know what you think!

re: The String Literal Returns

Nish — Thu, 09 Dec 2004 07:00:00 GMT

Stan

I believe that there are limits to which C# (or VB) can go to implement every IL-supported feature in .NET. But since the nature of C++ allows it to be the thinnest wrapper possible over assembler (or IL in this case), it'd be wonderful if C++/CLI allows us to implement every IL-supported feature there is. That alone should entice native C++ coders into moving to C++/CLI rather than C#/VB for .NET programming.

re: The String Literal Returns

coops — Thu, 09 Dec 2004 10:25:00 GMT

i would like to say im sooooo looking forward to checking out the new C++/CLI. Ive been following both you and herbs entries for some time and think your doing excellent work..

re: The String Literal Returns

RBischoff — Fri, 10 Dec 2004 16:16:00 GMT

Hi Stan,
First off I just want to thank you and your team at Microsoft for the hard work that has taking place on C++/CLI.

I actually love C++, but I was doing all development in C#.NET and I didn't particularly like the syntax of Managed C++. However, I am really excited about C++/CLI, it seems to now be a first class citizen in the .NET realm.

C++/CLI (IMHO) is a very nice langauge to program .NET, and it will offer the developer power/control with .NET coding. I will be doing all my .NET coding in C++/CLI instead of C#.

Oh yeah, C++/CLI is also a great langauge for unmanaged code, and the best langauge for mixing <un>managed code.

Thanks again to the C++/CLI team.

re: The String Literal Returns

Ben Chan — Mon, 13 Dec 2004 13:14:00 GMT

I used to be a win32 C++ programmer and a big C++ fans.However, no matter how much I love C++,the STL and templates.It cannot beat the productively of C# and java in high level application development.
I have no idea why other ppl is so hostile in C++/CLI.The native C++ sucks, I hate long compliation and linking time. I hate the pre-compiled header stuff and tons of complicated compiler settings.I hate I have to maintain both .h and .ccp file.I hate my template code is not 100% portable to other compiler.I suffered enough.
I love programming and also love to become a productive developer.I don't care about the standard stuff, I don't give a dxxn it is MS/SUN who design the next generation of C++.As long as it is good,I'll use it.
If C++ not getting upgrade, I bet I won't have any chance to use C++ forever.I love C++ and I don't want to see it die slowly.I hate hearing ppl keep attacking C++/CLI without doing any constructive suggestion.Some of the ppl fight against it because C++/CLI is build by MS, some are simply religion matters.
For whatever reason, I just feel sad about those irrational ppl and I think the current C++/CLI's implementation is cool as hell.For me,Stan lippman and his team is saviour to C++ and I would love to say thanks to you and your team's effort.C++/CLI is the only MS product I'll go for.

re: The String Literal Returns

Glen — Tue, 14 Dec 2004 23:06:00 GMT

Stan

I have long hated Microsoft for turning the endless possibilities of C++ programming, into the endless monotony of unproductive error prone COM programming (and poor design objectives), that Windows development had grown to typify. Microsoft I charge you with no less than nearly killing the Windows C++ community that you created.

But that was yesterday! I am truly astounded and excited that that you and the C++ team have finally heard our call, and have taken up the enormous challenge of returning our great language back to us. The new C++ 2005 appears to be all that I have been hoping for. Though I use C# now and have a lot of respect for it, I relish the chance to ditch it! If C++ can return to being the language of endless possibilities, then I can’t wait to get back to using it on the world’s most popular programming platform.

You and your team appear to be doing a fantastic and thorough job. Keep up the good work.

re: The String Literal Returns

Kate Gregory — Wed, 15 Dec 2004 16:49:00 GMT

I have been spending the last new months showing C++/CLI to people wherever and whenever I can, and I keep hearing exclamations of delight, followed by "I'm switching." Sometimes it's back from C#, sometimes it's they're finally leaving MFC, whatever. Their hearts were with C++ but they couldn't use it, and now they feel a sense of freedom and relief that they can again. I really like being the bearer of this news. Thanks to Stan, Herb, and the rest of the team for creating the news for me to bear.

re: The String Literal Returns

Chris Wundram — Wed, 15 Dec 2004 17:47:00 GMT

I thought the decorated string literals used in MC++ made perfect sense. If you were expecting a String^ you should use S"fritz", and if you are expecting a const wchar *, you should use L"fritz". It doesn't seem like much to ask, that people specify what kind of string it is in places where it would be ambigious.

I really like what has been done with the new C++/CLI stuff, but just like in MC++, I would expect any C++ code to do the same thing wheather it was compiled for win32 or for CLR.

re: The String Literal Returns

Mark Mullin — Wed, 15 Dec 2004 19:41:00 GMT

Geez, Stan - I've got to read your blog more often so I don't just post complaints on it. But here goes....

I can see the reason for C++/CLI as a waypoint for legacy code, and I can certainly see a limited implementation of C++ if you just want the familiar C++ syntax (that said. C# is not exactly off on another planet). And I can see mixed code, and all the other things you've talked about.

What I can't see, in fact what I think is dangerous, is to make C++/CLI try and 'act' too much like a C++ native. What I mean by this is illustrated by the example you give. While I also wouldn't (errr, shouldn't :-) ) throw numbers and strings as exceptions, I don't believe that the CLI should bypass it's own native types when effectively throwing one of it's own native types.

Specifically, in this example I'd argue that the developers you refer to recognize the interplay between CLI and native and do something real fun, along the lines of MI - try and throw to the CLI string^ catchers - not caught ? then try the const char *. Admittedly, this is a can of worms when you start looking at a series of nested catchers, but we sorta figured that out for multiple inheritence, and this problem is roughly similar, specifically -- whether this search resolves breadth or depth wise (search local exception catchers for any acceptable type or search hierarchy for best acceptable type

In any case, I think that there are going to be two modes of C++ use given the success of this project - people like your friend and myself burn millions upon millions of machine cycles. .net is fast, John Carmack proved that beyond a shadow of a doubt, but 85% of realtime ain't realtime. At other times, yeah it's nice to have a 'safer' environment, although I personally lean towards a real OO language in that case (Smalltalk :-)) BUT THE TWO ARE NOT SYNONYMOUS, and trying to make them so is a receipe for disaster! In short, C++/CLI should propagate towards C++ native, it should not do things as in your exception example, because this looses the most vital distinction of all - am I managed code for managed problems, or am I some performance or bust hairball. There's three areas here, (native, managed, and moving between the two) and it's that last one that needs to be recognized as a real situation, not something to be swept under the covers. If my managed code threw a string, then I think that implies I prefer a string^ over a const char*, but that I also expect a const char* to catch if there is no 'earlier' valid string^ catch

Anyway, my 2c. Keep fighting the good fight, happy holidays, etc

re: The String Literal Returns

Herb Sutter — Thu, 16 Dec 2004 18:42:00 GMT

Thanks, Mark.

>>There's three areas here, (native, managed, and moving between the two) and it's that last one that needs to be recognized as a real situation, not something to be swept under the covers.<<

Sure. I hope that we do that.

To recap, we convert a string literal to String^ implicitly in those cases where we can be sure it’s the right thing to do, usually because we know the immediate target, and throw “literal” isn’t such a case.

You then requested:

>>If my managed code threw a string, then I think that implies I prefer a string^ over a const char*, but that I also expect a const char* to catch if there is no 'earlier' valid string^ catch<<

It sounds like you want to additionally make catch clauses for either String^ or const char* implicitly be able to catch an exception of either type. This seems pretty subtle, likely problematic, and probably unwise.

Example off the top of my head: Say f1 calls f2 calls f3 which throws “literal” – if f2 catches String^ only and f1 catches const char*, do you really want f2 catching the “literal” as a String^ instead of letting f1 catch it as its exact type? And what if f1 and f3 already exist and are working, and then someone later writes f2?

I am uncomfortable with that path because it would explicitly blur the distinction between two distinct types. That’s a very different scenario from the convenience scenario we are supporting with the “literal”-to-String^ convenience conversion which basically knows the right type and saves you writing a cast.

Fortunately, we're talking about how we support a construct that programmers shouldn’t be writing in the first place. :-)

Thanks,

Herb

re: The String Literal Returns

Mark Mullin — Fri, 17 Dec 2004 20:03:00 GMT

Thanks for the response Herb (& Stan, implicitly)

I agree with your point, that's why I alluded to it being like MI - then again, it's that relation that makes this 'interesting' :-)

I agree you've been bad if you do this particular example in the first place, but, then again, I've been bad before and will certainly be again.

But......... C++/ILM is going to exist in a world of nice managed types, and most likely not so nice unmanaged types. No big deal, IntPtr is already my special friend in C#. But^2 - I know whenever I look at my C# code and see IntPtr, I'm looking at something nasty from the outside. I can easily write C# code that deals with things coming across the p/invoke boundary and from within the managed app without standing on my head to figure out what's what.

I think, because of the greater overlap between C++/ILM and C++/native, it might be worth considering the 'network' of type relations - A string^ is a const char* is a char* is a void*.... being just one example of the trip from the warm safe womb of managed code to the raw metal of the machine.

It's been a long time since I've been involved in language law, but I think there's an issue here which is a product of the fact that C++ will have two very distinct operating environments after you are done, and that there is an excellent chance those environments will be mixed in a single program, via p/invoke at the very least. I think there's a good case for amping up type conversion operations, and I think that Stan's original example showed that.

As a concrete example, consider this. F1 is a high level control function, and calls two functions - F2M (managed) and F2U (unmanaged). F2M throws a string^ and returns a string^, F2U throws a char* and returns a char* (yeah, I'd use stl).
Now assume that F1 implements Process(string^ myData) and Process(char* myData)

Now consider
try {
f1.Process(virtualInF2(myData))
}
catch(???)

If I've got a clearly defined pattern of type propagation (yes, I admit what I'm asking for would actually be a real bear, but anything worth having......) then I've got a world where like goes to like, or follows clear rules of conversion. In the case above, given you could solve tbe ?multiple type/MI? issue, I'd have a much more powerful system. This would be far more powerful than just shoehorning C++ into .net, it would be C++ that could allow developers to stand simulataneously (and as safely as possible) in managed and native worlds at once.

I don't deny this is difficult (that's why I said network and not dag) - I also believe you folk have something truly new under the sun here, and I think you need to recognize it's needs. I don't like the idea of just tossing the managed sides typing information and racing for the lowest common denominator in C++. After all, if I wanted that, I'd sed all my types to void* and use native :-)

Regards
MMM

re: The String Literal Returns

Jason Li — Sat, 18 Dec 2004 05:23:00 GMT

I like the current string literal's design. It is reasonable & convenient. And I think if there is a amphibolous statement like the following:

throw "fritz";

it is OK for C++/CLI compiler to just give an error/warning info, since there is no context to decide the type of "fritz". The programmers are responsible for changing it to a clear statement like the following:

throw (String^)"abc";

And the analogy of instinct vs. a decision is very wonderful.

As for those who are unhappy with C++/CLI, I think the problem is Microsoft, not C++/CLI itself. If C++/CLI can only be used on Microsoft platform in future, it will be rebuked by people continuously & emotionally.

If Microsoft has an open mind on CLI, let more organizations & companies involved in CLI's development, C++/CLI will gain great success. As a CLI language, C++/CLI has already succeeded. But C++/CLI's popularity is dependent on CLI's popularity, C++/CLI's final success is dependent on CLI's success.

Yes, Microsoft has submitted CLI as ISO standard, but that's not enough. After all, the current fact is that only Microsoft has implemented an industry-level CLI (Mono is too young.)

To tell the truth, C++/CLI's design is already good enough, and the current C++ team is the most respected & excellent team in Microsoft's history. By the way, in China, many programmers tell me that they love C++/CLI, they will come back to C++/CLI after having given C++ up for several years.

Thanks for you & your team's great work!

re: The String Literal Returns

stan lippman — Sat, 18 Dec 2004 22:00:00 GMT

i don't see how you can generalize like this on such a fringe inconsequential oddity. however, i appreciate your taking the time to articulate your position. i hope we've given it full consideration. that's all we can do in a debate like this. thank you.

that said, (a) i still stand by my initial evaluation in my post, and (b) i am closing
off this thread. no further comments on it
will appear.

thanks to mark and herb for engaging in a thoughtful debate. everyone's positions should be clear at this point.

have a good holiday ... i will see everyone next year.

re: The String Literal Returns

objectref — Thu, 30 Dec 2004 13:12:00 GMT

I don't see the usefulness of "C# mean nothing to me"...
C# is .net oriented at the first place and in my opinion, a beautifully designed language.

So, all the work your team has done, although very good but still it will try to accomplish the right way something that C# was designed to accomplish 4 years ago, and it surelly has succeeded on it.

I also do not see, if .net is the future, why someone will dump C# and go back to C++.

And don't tell me about the somewhat less instruction set that C++ compliler will produce over the C# compiler because, alas, we live in 3GHz processors, and productivity is in demand nowadays...

I now C++ is your favorite one, but as i see the evolving of other languages, i think that they put C++ on the corner, release by release...

lippman_rss2lj: The String Literal Returns

lippman_rss2lj: The String Literal Returns — Tue, 05 Sep 2006 13:54:17 GMT

PingBack from http://www.livejournal.com/users/lippman_rss2lj/8151.html

The String Literal Returns by lippman_rss2lj () | LjSEEK.COM

The String Literal Returns by lippman_rss2lj () | LjSEEK.COM — Tue, 05 Sep 2006 13:54:20 GMT

PingBack from http://www.ljseek.com/the-string-literal-returns_12697914.html

Anothr feed track -Stan Lippman's BLog

anothr user — Thu, 05 Apr 2007 04:31:48 GMT

One new subscriber from Anothr Alerts

Stan Lippman's BLog : The String Literal Returns

Relationship Compatibility — Sat, 07 Jun 2008 21:42:44 GMT

In the last entry, I celebrated what I felt was an elegant solution to the problem of the string literal in the context of overload function resolution. But it turns out there is another area in which the string literal proves problematic. Who would hav

Stan Lippman s BLog The String Literal Returns | Paid Surveys

Stan Lippman s BLog The String Literal Returns | Paid Surveys — Sat, 30 May 2009 03:42:17 GMT

PingBack from http://paidsurveyshub.info/story.php?title=stan-lippman-s-blog-the-string-literal-returns

Stan Lippman s BLog The String Literal Returns | Outdoor Ceiling Fans

Stan Lippman s BLog The String Literal Returns | Outdoor Ceiling Fans — Sun, 31 May 2009 14:18:19 GMT

PingBack from http://outdoorceilingfansite.info/story.php?id=4375

Stan Lippman s BLog The String Literal Returns | Outdoor Ceiling Fans

Stan Lippman s BLog The String Literal Returns | Outdoor Ceiling Fans — Sun, 31 May 2009 19:54:29 GMT

PingBack from http://outdoorceilingfansite.info/story.php?id=22011

Stan Lippman s BLog The String Literal Returns | Toe Nail Fungus

Stan Lippman s BLog The String Literal Returns | Toe Nail Fungus — Tue, 09 Jun 2009 09:04:43 GMT

PingBack from http://toenailfungusite.info/story.php?id=4987

Stan Lippman s BLog The String Literal Returns | Weak Bladder

Stan Lippman s BLog The String Literal Returns | Weak Bladder — Tue, 09 Jun 2009 21:45:14 GMT

PingBack from http://weakbladder.info/story.php?id=4914

Stan Lippman s BLog The String Literal Returns | Outdoor Decor

Stan Lippman s BLog The String Literal Returns | Outdoor Decor — Fri, 19 Jun 2009 07:13:32 GMT

PingBack from http://outdoordecoration.info/story.php?id=1566