www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - D sandbox?

reply davidl <davidl 126.com> writes:
currently DParser passes some dstress test cases. I'm in the mood of  
glueing
it with DMD backend.

With current C++ interface improvements in D, I think DParser is possible  
to glue with DMD backend. I hope Walter can provide an interface , and a  
backend library. Though it's a great bunches of work :D

It's still possible to glue with LLVM and take advantage from the work  
done by lindquist. But a general backend interface not only glueing with  
one backend is better for me in some sense.

1. This benefits several scripting langs. Those lang can take advantage  
 from DMD backend to generate native code.

2. The work on DParser can therefore attract more people to help. And thus  
more bug fixes can be done. DMD C++ frontend might get some patches from  
those bug fixes.

3. DParser itself can create branches as sandboxes to give the community  
some experiments and coding in D is relatively easier than C++ in the  
original DMD frontend for the community. So community can really make some  
experiments on certain branches, and as sandboxes. And Walter can view  
this sandbox and maybe lucky enough to pick one good feature integrated in  
DMD.

4. This gives us a bit elegance. D compiles D.

This is *not* an urgent request. Cause DParser compare to DMD frontend is  
still not relatively stable. I'm trying to fix those bugs. Thanks to  
dstress. I can get a good many test cases in it.
Currently I guess about 1/3 test cases pass.

I'd love to listen how CONs think and how PROs think.

DavidL
Oct 16 2007
parent reply BCS <ao pathlink.com> writes:
Reply to davidl,

 currently DParser passes some dstress test cases. I'm in the mood of
 glueing
 it with DMD backend.
 With current C++ interface improvements in D, I think DParser is
 possible  to glue with DMD backend. I hope Walter can provide an
 interface , and a  backend library. Though it's a great bunches of
 work :D
 
 It's still possible to glue with LLVM and take advantage from the work
 done by lindquist. But a general backend interface not only glueing
 with  one backend is better for me in some sense.
 
 1. This benefits several scripting langs. Those lang can take
 advantage
 from DMD backend to generate native code.
 2. The work on DParser can therefore attract more people to help. And
 thus  more bug fixes can be done. DMD C++ frontend might get some
 patches from  those bug fixes.
 
 3. DParser itself can create branches as sandboxes to give the
 community  some experiments and coding in D is relatively easier than
 C++ in the  original DMD frontend for the community. So community can
 really make some  experiments on certain branches, and as sandboxes.
 And Walter can view  this sandbox and maybe lucky enough to pick one
 good feature integrated in  DMD.
 
 4. This gives us a bit elegance. D compiles D.
 
 This is *not* an urgent request. Cause DParser compare to DMD frontend
 is
 still not relatively stable. I'm trying to fix those bugs. Thanks to
 dstress. I can get a good many test cases in it.
 Currently I guess about 1/3 test cases pass.
 I'd love to listen how CONs think and how PROs think.
 
 DavidL
 

Walter has a standing offer to /sell/ licenses to the backend.
Oct 16 2007
parent DavidL <DavidL 126.com> writes:
BCS Wrote:

 Reply to davidl,
 
 currently DParser passes some dstress test cases. I'm in the mood of
 glueing
 it with DMD backend.
 With current C++ interface improvements in D, I think DParser is
 possible  to glue with DMD backend. I hope Walter can provide an
 interface , and a  backend library. Though it's a great bunches of
 work :D
 
 It's still possible to glue with LLVM and take advantage from the work
 done by lindquist. But a general backend interface not only glueing
 with  one backend is better for me in some sense.
 
 1. This benefits several scripting langs. Those lang can take
 advantage
 from DMD backend to generate native code.
 2. The work on DParser can therefore attract more people to help. And
 thus  more bug fixes can be done. DMD C++ frontend might get some
 patches from  those bug fixes.
 
 3. DParser itself can create branches as sandboxes to give the
 community  some experiments and coding in D is relatively easier than
 C++ in the  original DMD frontend for the community. So community can
 really make some  experiments on certain branches, and as sandboxes.
 And Walter can view  this sandbox and maybe lucky enough to pick one
 good feature integrated in  DMD.
 
 4. This gives us a bit elegance. D compiles D.
 
 This is *not* an urgent request. Cause DParser compare to DMD frontend
 is
 still not relatively stable. I'm trying to fix those bugs. Thanks to
 dstress. I can get a good many test cases in it.
 Currently I guess about 1/3 test cases pass.
 I'd love to listen how CONs think and how PROs think.
 
 DavidL
 

Walter has a standing offer to /sell/ licenses to the backend.

Ahh, yet, LLVM, GCC are possible options. And I personally want to make it a script too.
Oct 16 2007