www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Unclear about the benefits of D over C++ and Java

reply Dibyendu Majumdar <d.majumdar gmail.com> writes:
Hi,

I am a long time observer of D but have never done any coding in 
D. I have considered using D several times but have ended up 
either going with C++ or Java.

I feel that for someone like me it is not clear what significant 
advantages D has over C++ or Java to make a switch. I cannot 
afford to experiment with a language unless I know for certain 
the benefits of using it.

For instance I would like to know:

* Do D compilers generate more efficient code compared to C++ 
compilers? Are there performance benchmarks comparing D 
implementation against C and C++?

* Can D generics do everything C++ templates can (are they Turing 
complete for instance) - and what additional capabilities does D 
have that would make a significant difference to the way 
templates are used in C++?

* How does D's performance compare with Java when it comes to 
concurrent applications or garbage collection?

* How stable is D? Can one use it without fear that the next 
version will significantly change the language and therefore one 
would have to rewrite?

I feel that the D website could do more to explain why someone 
should move from C++ or Java to D - and this has to be explained 
in terms of real world benefits in terms of:

* Performance
* Quality of implementation

For someone like me - cool language features don't count that 
much unless they lead to real world benefits.

My post here was triggered after watching a video by Andrei where 
he talked about how D generics are different from C++ templates - 
I think stuff like that needs to be explained in the front page 
of D website.

Regards
Dibyendu

http://ravilang.org
http://simpledbm.org
Jan 02 2016
next sibling parent Rikki Cattermole <alphaglosined gmail.com> writes:
On 03/01/16 12:31 AM, Dibyendu Majumdar wrote:
 Hi,

 I am a long time observer of D but have never done any coding in D. I
 have considered using D several times but have ended up either going
 with C++ or Java.

 I feel that for someone like me it is not clear what significant
 advantages D has over C++ or Java to make a switch. I cannot afford to
 experiment with a language unless I know for certain the benefits of
 using it.

 For instance I would like to know:

 * Do D compilers generate more efficient code compared to C++ compilers?
 Are there performance benchmarks comparing D implementation against C
 and C++?

 * Can D generics do everything C++ templates can (are they Turing
 complete for instance) - and what additional capabilities does D have
 that would make a significant difference to the way templates are used
 in C++?

 * How does D's performance compare with Java when it comes to concurrent
 applications or garbage collection?

 * How stable is D? Can one use it without fear that the next version
 will significantly change the language and therefore one would have to
 rewrite?

 I feel that the D website could do more to explain why someone should
 move from C++ or Java to D - and this has to be explained in terms of
 real world benefits in terms of:

 * Performance
 * Quality of implementation

 For someone like me - cool language features don't count that much
 unless they lead to real world benefits.

 My post here was triggered after watching a video by Andrei where he
 talked about how D generics are different from C++ templates - I think
 stuff like that needs to be explained in the front page of D website.

 Regards
 Dibyendu

 http://ravilang.org
 http://simpledbm.org
Okay so lets start with code efficiency. We can beat in some cases and match C++ code generation using gdc or ldc. Which use GCC and LLVM backends. About D templates. Anything can be executed at compile time can be used in templates. You just can't interact with non D code or globals. Same with e.g. static assert. Because of this D's meta-programming capabilities beat almost all other languages (Lisp in general is the only set of language that we don't). Okay, now about general performance and the GC. If you want performant apps you can't use the GC. It's as simple as that. The Java and .net crowd like to say otherwise but you just can't. No GC will work as well for memory management quite as like a developer can. In terms of language changes, you shouldn't need to worry about it too much. Most of the time these days you can go between versions without any changes if you're not pushing the language too much. It's not like what it was a few years ago.
Jan 02 2016
prev sibling next sibling parent Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
wrote:
 I feel that for someone like me it is not clear what 
 significant advantages D has over C++ or Java to make a switch. 
 I cannot afford to experiment with a language unless I know for 
 certain the benefits of using it.
If you are truly proficient in C++ then you probably are better off with C++ at this point than using languages like D, Rust or Swift, at least for long term commercial projects.
 * Do D compilers generate more efficient code compared to C++ 
 compilers?
No. The best D compilers use backends written for C/C++ (llvm/gcc).
 * Can D generics do everything C++ templates can (are they 
 Turing complete for instance) - and what additional 
 capabilities does D have that would make a significant 
 difference to the way templates are used in C++?
Certain things are more tedious in C++, like adding methods based on template parameters. Compile time reflection is also more tedious in C++. Yet again, certain other things look syntactically better in C++ and other things look better syntactically in D.
 * How does D's performance compare with Java when it comes to 
 concurrent applications or garbage collection?
Java has much better garbage collectors. Concurrency frameworks have been more thoroughly tested in Java and C++.
 * How stable is D? Can one use it without fear that the next 
 version will significantly change the language and therefore 
 one would have to rewrite?
Java and C++ are more mature and have been more stable.
Jan 02 2016
prev sibling next sibling parent reply Dibyendu Majumdar <d.majumdar gmail.com> writes:
On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
wrote:
 I am a long time observer of D but have never done any coding 
 in D. I have considered using D several times but have ended up 
 either going with C++ or Java.

 I feel that for someone like me it is not clear what 
 significant advantages D has over C++ or Java to make a switch. 
 I cannot afford to experiment with a language unless I know for 
 certain the benefits of using it.

 My post here was triggered after watching a video by Andrei 
 where he talked about how D generics are different from C++ 
 templates - I think stuff like that needs to be explained in 
 the front page of D website.
To clarify - the reason for my post here is to raise that the benefits of D over C++ or Java should be explained right up-front on D website in real world terms (i.e. with examples and benchmarks) to help people like me. Regards Dibyendu
Jan 02 2016
parent Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Saturday, 2 January 2016 at 12:01:20 UTC, Dibyendu Majumdar 
wrote:
 To clarify - the reason for my post here is to raise that the 
 benefits of D over C++ or Java should be explained right 
 up-front on D website in real world terms (i.e. with examples 
 and benchmarks) to help people like me.
Ok, that is not something I can help you with, but would you trust information provided by the language creators on performance and language feature trade offs? That said: I really don't think you should consider languages like D and Rust as replacements for Java. If you use Java, you usually have dependencies on frameworks/infrastructure and also a strong preference for garbage collection. Go is closer to being a replacement for small Java projects. D's garbage collector is comparable to the C++ Boehm collector (I think Inkscape uses that). D is semantically a close cousin of C++, but the standard libraries are different, although inspired by C++ libraries and has adopted C++ terminology. And the template programming semantics are slightly different. Advanced C++ meta programming is more painful than in other languages because it came about as an accident, and even simple things can be demanding and verbose to implement. D's approach to template programming is pretty straight forward, nothing spectacular and with some syntactical annoyances, but without the really weird C++ SFINAE and inheritance constructs that C++ programmers have to deal with. D has also implemented some of the features that has been proposed for C++, but not adopted in C++ yet. Some areas where C++ is better: - number of available template libraries - does not have modular arithmetics for signed integers - somewhat more powerful lambdas - more flexible operator overloading - Objective-C++ integration - better platform support (e.g. asm.js) - C++17 brings stackless concurrency Overall, C++ is more verbose than D and that means that C++ isn't a good language for testing out ideas. D is more suitable for that.
Jan 02 2016
prev sibling next sibling parent reply Guillaume Piolat <first.last gmail.com> writes:
On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
wrote:
 Hi,

 I am a long time observer of D but have never done any coding 
 in D. I have considered using D several times but have ended up 
 either going with C++ or Java.

 I feel that for someone like me it is not clear what 
 significant advantages D has over C++ or Java to make a switch. 
 I cannot afford to experiment with a language unless I know for 
 certain the benefits of using it.

 For instance I would like to know:

 * Do D compilers generate more efficient code compared to C++ 
 compilers? Are there performance benchmarks comparing D 
 implementation against C and C++?
Same efficiency if using LDC or GDC, divided by 2 if you use DMD. But I've found coding in D is more productive (I'd say 2x on small projects vs C++, correct me if you disagree) and you'll be more pleased with the results so you'll have more time and energy to optimize. C++ has some overhead to create a program so it may lead you to tool less.
 * Can D generics do everything C++ templates can (are they 
 Turing complete for instance) - and what additional 
 capabilities does D have that would make a significant 
 difference to the way templates are used in C++?
The difference is mainly in community, most are very confident with meta-programming, it's not up to single individual in a team to explain static polymorphism to everyone. To put bluntly, I've never seen a C++ programmer in a team use std::enable_if in years of C++, but every D programmer has written heaps of "static if". D lets you use static or runtime polymorphism as easily, in C++ it's quite a learning and often people think they don't need it. That said, it's not all community, there is tangible improvements. Syntax for templates, alias this over multiple inheritance, string mixin over preprocessor macros, static foreach, sane operator overloading and so on.
 * How does D's performance compare with Java when it comes to 
 concurrent applications or garbage collection?
Java have more concurrency primitives and faster garbage collection. OTOH D runtime is smaller.
 * How stable is D? Can one use it without fear that the next 
 version will significantly change the language and therefore 
 one would have to rewrite?
This won't be a problem. Breaking language changes are rare, and are easy to fix.
 I feel that the D website could do more to explain why someone 
 should move from C++ or Java to D - and this has to be 
 explained in terms of real world benefits in terms of:

 * Performance
 * Quality of implementation
Can't speak with anyone else, but for me the real reasons are: - it's fun - D says "yes" to everything and is not a Big Idea language - I'm happier with the results, to the point I want to maintain them - C++ puts you in learning-mode forever. D also does, but without you taking notice :) - I'm hooked and going back to C++ feels horrible, like for many people here. We need a D rehab center! At the language level D improves things in all directions: http://p0nce.github.io/d-idioms/#How-does-D-improve-on-C++17?
 For someone like me - cool language features don't count that 
 much unless they lead to real world benefits.
Me too, I care more about shared library support and codegen for example than any language feature. Yet in some cases they do matter. For example default initialization brings invisible productivity gains you only see when going back to C++.
Jan 02 2016
parent reply Dibyendu Majumdar <d.majumdar gmail.com> writes:
Hi,

I am looking to choose between D, Swift and Rust for a project 
that I am currently coding in C++. So far D seems the alternative 
but I guess I won't know until I try out a few things.

Is this the right forum to ask questions or should questions be 
sent to another forum?

Regards
Dibyendu
Jan 03 2016
next sibling parent tsbockman <thomas.bockman gmail.com> writes:
On Sunday, 3 January 2016 at 15:38:18 UTC, Dibyendu Majumdar 
wrote:
 Hi,

 I am looking to choose between D, Swift and Rust for a project 
 that I am currently coding in C++. So far D seems the 
 alternative but I guess I won't know until I try out a few 
 things.

 Is this the right forum to ask questions or should questions be 
 sent to another forum?

 Regards
 Dibyendu
Specific questions about how to program in D belong in the "Learn" forum: http://forum.dlang.org/group/learn
Jan 03 2016
prev sibling parent reply Shannon <refaQtor gmail.com> writes:
On Sunday, 3 January 2016 at 15:38:18 UTC, Dibyendu Majumdar 
wrote:
 I am looking to choose between D, Swift and Rust for a project 
 that I am currently coding in C++. So far D seems the 
 alternative but I guess I won't know until I try out a few 
 things.
Why I now choose D, even for commercial jobs... C++ had been my primary language (personally and professionally) for 20 years, and it was effective and enjoyable for me. But, I was naive. I didn't realize that what made it enjoyable for me, is that I always used effective and complete frameworks/toolsets (IBM VisualAgeC++ on OS/2, BeOS standard API, Qt on everything) with effective base classes and data structures that all worked together. When forced out of my comfortable space, by a project a couple years ago that allowed no dependencies on Qt, I discovered the grief of C++. So, time to give D a try. I'd watched it through the tumultuous years, and saw that it had settled down. I like the pedegree of D - a pragmatic approach from a creator of a commercial C++ compiler. Also, I was encouraged to see Alexander jump in to contribute to D (even before the D foundation announcement), as he's been tirelessly working to improve C++ over the years. I, like you, wanted to "try out a few things" before committing to D on commercial effort. I found myself fantastically productive (more than C++, even with Qt) using rdmd to begin script-like and refactor as it grew. Easy, like Python that produces fast compiled single executables that are easy to deploy (even to folks without a Python development environment installed), especially with the 'import' feature to embed external resources. It is so easy to keep unit tests current and running, and so easy to see testing code coverage, that I felt always safe refactoring. I appreciate the ability to mark any scope safe, and have the compiler keep me honest. And I found myself using more "esoteric" features that were a pain in C++. I loved the concept of meta-programming in C++, but always found myself using features in my framework - which had taken the great pains to provide the necessary functionality. I found implementing threads and concurrency in D every bit as simple and safe as Qt. Admittedly, GUI stuff in D is a mixed bag. I've settled on creating D console applications that transmits json data, via stdin/stdout, directly into Qt base model classes so the rest of the Qt framework works seamlessly. I quite like the solution, and I'm actually glad there was no "standard" gui interface that bound me. But, gtk and sdl bindings seemed to work easy enough for simple things, and dlanggui is looking promising. As for future commercial efforts, I'll continue with D. I feel safe that straight C is relatively easy to integrate. Having access to proven libs like sqlite, redis, libgit2, nanomsg makes little incentive to ever feel like replacing them. Even many of the interesting Python libs are just linked to a standard C lib underneath. Anyway, I'm just more productive at producing robust, fast code in D. Best of luck in your efforts.
Jan 03 2016
parent Wyatt <wyatt.epp gmail.com> writes:
On Sunday, 3 January 2016 at 18:39:21 UTC, Shannon wrote:
 On Sunday, 3 January 2016 at 15:38:18 UTC, Dibyendu Majumdar 
 wrote:
 I am looking to choose between D, Swift and Rust for a project 
 that I am currently coding in C++. So far D seems the 
 alternative but I guess I won't know until I try out a few 
 things.
Why I now choose D, even for commercial jobs...
Nice story; mind if we put it on the user narratives wiki page? -Wyatt
Jan 05 2016
prev sibling next sibling parent reply JerryR <jerryr gmail.com> writes:
On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
wrote:
 ...
 * How stable is D? Can one use it without fear that the next 
 version will significantly change the language and therefore 
 one would have to rewrite?
 ...
I recently discover a problem with breakage when trying to compile using a newer version of DMD: http://forum.dlang.org/thread/affaugpppegxuxyoilea forum.dlang.org Well I don't know how thing are now in fact, but the above wasn't the first time. And to be honest I already get it that using languages like D, Rust and Go which "still" are growing is a risk when talking about non-breakage. J.
Jan 02 2016
parent tsbockman <thomas.bockman gmail.com> writes:
On Saturday, 2 January 2016 at 19:53:04 UTC, JerryR wrote:
 On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
 wrote:
 ...
 * How stable is D? Can one use it without fear that the next 
 version will significantly change the language and therefore 
 one would have to rewrite?
 ...
I recently discover a problem with breakage when trying to compile using a newer version of DMD: http://forum.dlang.org/thread/affaugpppegxuxyoilea forum.dlang.org Well I don't know how thing are now in fact, but the above wasn't the first time. And to be honest I already get it that using languages like D, Rust and Go which "still" are growing is a risk when talking about non-breakage. J.
To clarify: many projects will fail to compile when a new version comes out. However, you will NOT have to "rewrite" the project. Rather, a few lines across the entire project might have to be updated.
Jan 02 2016
prev sibling next sibling parent reply israel <tl12000 live.com> writes:
On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
wrote:
 Hi,

 I am a long time observer of D but have never done any coding 
 in D. I have considered using D several times but have ended up 
 either going with C++ or Java.

 I feel that for someone like me it is not clear what 
 significant advantages D has over C++ or Java to make a switch. 
 I cannot afford to experiment with a language unless I know for 
 certain the benefits of using it.
You know, from the last 2 years that i have followed D, i have come to the conclusion that: If you enjoy programming, use D. If its just your job or because its the standard, stay with C/C++. because not once have i heard someone say programming in C++ is enjoyable. But maybe its just me, maybe im too young to know the olden days of the 80s/90s where C++ was a godsend compared to C. Or was it? I dont know, i wasnt alive...
Jan 02 2016
next sibling parent rsw0x <anonymous anonymous.com> writes:
On Sunday, 3 January 2016 at 04:25:55 UTC, israel wrote:
 On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
 wrote:
 Hi,

 I am a long time observer of D but have never done any coding 
 in D. I have considered using D several times but have ended 
 up either going with C++ or Java.

 I feel that for someone like me it is not clear what 
 significant advantages D has over C++ or Java to make a 
 switch. I cannot afford to experiment with a language unless I 
 know for certain the benefits of using it.
You know, from the last 2 years that i have followed D, i have come to the conclusion that: If you enjoy programming, use D. If its just your job or because its the standard, stay with C/C++. because not once have i heard someone say programming in C++ is enjoyable. But maybe its just me, maybe im too young to know the olden days of the 80s/90s where C++ was a godsend compared to C. Or was it? I dont know, i wasnt alive...
Ironically, I think C++ has only gotten worse with age and I'm speaking compared to back when there were debates over functions were worth using. In terms of modern C++... I saw nothing of the 'beauty' of C++11 that everyone else somehow saw, all I saw was more cruft to sift through and more "rules" to learn because the language is fundamentally broken. At some point the C++ committee needs to accept that C++ is approaching unusable levels of complexity, but I guess everyone needs job security.
Jan 02 2016
prev sibling parent reply Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Sunday, 3 January 2016 at 04:25:55 UTC, israel wrote:
 But maybe its just me, maybe im too young to know the olden 
 days of the 80s/90s where C++ was a godsend compared to C. Or 
 was it? I dont know, i wasnt alive...
C++ was hyped up in the press and commercial sector because it provided abstraction mechanisms on top of the existing C infrastructure. But most academics viewed C++ as an ugly hack back then, so it was no God send. Many programmers viewed it as bloated and inefficient, and it kinda was because compilers could not beat hand optimized C code (or: computers were too slow to make good compilers practical). So C++ was more of an application level language than system level language. That is still true, but today CPUs are fast, bottle necks are more in the memory system/multi threading, CPU internals are less transparent and programmers have less optimization skills... Which paves the way for high level programming languages like Swift. Modern C++ is tolerable and fairly decent, but it will take a proficient programmer years to master. The big issue with C++ is that you also need to learn what _not_ to do, by trail and error... C++ is not a language that one can reasonable select for a one-off project, it's a language for experts only. Unlike Python, Ruby, Swift or Go. Maybe D2 is becoming a language for experts too at this point. A stronger focus on consistency and ergonomics would be beneficial.
Jan 03 2016
parent Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Sunday, 3 January 2016 at 09:55:14 UTC, Ola Fosheim Grøstad 
wrote:
 C++ was hyped up in the press and commercial sector because it 
 provided abstraction mechanisms on top of the existing C 
 infrastructure.
And just to give some context to this: in the 80s a large number of programmers were self-taught and had never been exposed to a disciplined language like Java as their first language, many started with BASIC and machine language. At best they had some experience with ad hoc structured programming in Pascal. In the 80s it was not uncommon to implement applications or portions of applications in assembly, and CISC CPUs like motorola 68k had a rather high level instruction set so it wasn't as unpleasant as it sounds. One guy I knew implemented a full fledged BBS system (Amiga BBS) in assembly because he didn't know C and thought he might as well do it in machine language rather than learning C... In that context C++/OOP provided a new market for educational books and magazine articles and lots of "propaganda". Which in turn made C++ fashionable... It created a sense of "you need to learn C++", yet in practice many just used it as C with inlining and overloading...
Jan 03 2016
prev sibling parent welkam <wwwelkam gmail.com> writes:
On Saturday, 2 January 2016 at 11:31:24 UTC, Dibyendu Majumdar 
wrote:
 * Do D compilers generate more efficient code compared to C++ 
 compilers? Are there performance benchmarks comparing D 
 implementation against C and C++?

 * Can D generics do everything C++ templates can (are they 
 Turing complete for instance) - and what additional 
 capabilities does D have that would make a significant 
 difference to the way templates are used in C++?

 * How does D's performance compare with Java when it comes to 
 concurrent applications or garbage collection?

 * How stable is D? Can one use it without fear that the next 
 version will significantly change the language and therefore 
 one would have to rewrite?

 I feel that the D website could do more to explain why someone 
 should move from C++ or Java to D - and this has to be 
 explained in terms of real world benefits in terms of:

 * Performance
 * Quality of implementation

 For someone like me - cool language features don't count that 
 much unless they lead to real world benefits.

 My post here was triggered after watching a video by Andrei 
 where he talked about how D generics are different from C++ 
 templates - I think stuff like that needs to be explained in 
 the front page of D website.

 Regards
 Dibyendu

 http://ravilang.org
 http://simpledbm.org
Are there performance benchmarks comparing D
 implementation against C and C++?
C style D code will run the same as C style C++ code and a C style C code because they all compile to the same machine code. There will be differences if you use higher level language features, but then you are trading performance for convenience and its up to you to decide if that trade off is worth it. Also keep in mind that D compilers do not apply all possible optimisation and your code next year my become faster.
 * How does D's performance compare with Java when it comes to 
 concurrent applications or garbage collection?
In this world there are two options: 1. You build language around GC sacrificing flexibility and performance if you dont use GC. 2. Or you add simple GC. D went with second option. D`s GC will never be as good as Java, without GC and reach C level of performance. You asked a lot about language performance, but application performance is mostly dependant on code you write and not the language you use. Benchmarking and tuning your code will give way more performance than changing languages.
Jan 05 2016