www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - cppfront (Herb Sutter's idea in CppCon)

reply XavierAP <n3minis-git yahoo.es> writes:
Anyone watched it already? What do you think?

https://youtu.be/ELeZAKCN4tY?t=22m05s

https://github.com/hsutter/cppfront

Quite an origial idea, while at the same time it looks partially 
self-defeating in terms of defending C++ (against competition 
from Rust, which seems to be the motivation).

I'm particularly surprised (to be honest, put off) by Herb's 
choice to make declarations in this new language of his 
unnecessarily different from C++ in order to make them look like 
Rust.
Nov 13 2022
next sibling parent Nick Treleaven <nick geany.org> writes:
On Sunday, 13 November 2022 at 17:14:15 UTC, XavierAP wrote:

 Anyone watched it already? What do you think?


This was discussed here in September:
https://forum.dlang.org/post/tfzncabvtbanpzyvufpf forum.dlang.org
Nov 13 2022
prev sibling parent reply norm <norm.rowtree gamil.com> writes:
On Sunday, 13 November 2022 at 17:14:15 UTC, XavierAP wrote:

 Anyone watched it already? What do you think?

 https://youtu.be/ELeZAKCN4tY?t=22m05s

 https://github.com/hsutter/cppfront

 Quite an origial idea, while at the same time it looks 
 partially self-defeating in terms of defending C++ (against 
 competition from Rust, which seems to be the motivation).

 I'm particularly surprised (to be honest, put off) by Herb's 
 choice to make declarations in this new language of his 
 unnecessarily different from C++ in order to make them look 
 like Rust.


and the Rust marketing machine cogs turn once more...Rust did not 
invent that syntax yet people now see it and think of Rust.

I think this approach can work for C++. I really like C++ (post 
C++17) and want to see it improve, same as I really like D and 
would love to see it take off.
Nov 13 2022
parent reply XavierAP <n3minis-git yahoo.es> writes:
On Sunday, 13 November 2022 at 22:07:41 UTC, norm wrote:

 and the Rust marketing machine cogs turn once more...Rust did 
 not invent that syntax yet people now see it and think of Rust.

 I think this approach can work for C++. I really like C++ (post 
 C++17) and want to see it improve, same as I really like D and 
 would love to see it take off.


It looks to me that (competing against) Rust must be the reason 
why this syntax is chosen. I didn't say and I didn't think Rust 
invented it -- though I don't really know where it comes from 
originally, or how to call it. Disclaimer: I don't use Rust.

It occurs to me the syntax may make it easier for cppfront to 
identify parts of the code that it needs to compile and others to 
leave alone. In any case requiring it to handle mixed source with 
cpp and cpp2 in the same file, as it does now for the few 
features it supports, sounds really challenging in the long term.

Otherwise as I said, choosing this syntax looks to me very 
questionable, and likelier to push people away from C++ than 
attract them. Not to mention pushing away existing C++ 
programmers. It's like admitting defeat, that C-like syntax is 
worse.

Of course you want to automate and get rid of const auto & const 
overloads, copy-move Hell etc, this is a different topic and the 
main benefit that Herb seems to be going after. But as for 
teaching C++ programmers to declare name: type = instead of type 
name = etc, is it a good idea? While at the same time claiming 
that this is meant so that the generated cpp can still be read 
and maintained on its own if necessary.

Anyway other than the syntax, I'm also curious if this approach 
can work, both in the short and the long terms. If you follow the 
logic, comparing this, as Herb does, to the first C++ compiler by 
Bjarne S which was a trans-compiler to C, then the mixed C/C++ 
period, until the time when C and C++ have become completely 
separate languages, is this the success that Herb would be going 
after in the best of cases, to create a new, different language 
(and one that looks so different)?
Nov 14 2022
parent Olivier Pisano <olivier.pisano laposte.net> writes:
On Monday, 14 November 2022 at 13:57:34 UTC, XavierAP wrote:

 Of course you want to automate and get rid of const auto & 
 const overloads, copy-move Hell etc, this is a different topic 
 and the main benefit that Herb seems to be going after. But as 
 for teaching C++ programmers to declare name: type = instead of 
 type name = etc, is it a good idea?


It enables syntax for functions to be the same as for lambdas, 
eliminating a special case, so yes, it is a good idea.
Nov 14 2022