www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - preprocessor pass equivalent?

reply "Jay Norwood" <jayn prismnet.com> writes:
Is there some option, similar to -E gcc option, that would 
generate the analogous listing for D?  What I mean is that I'd 
like to see all the code in version blocks gone, all the mixin 
strings expanded.

It is probably too much to wish for having the infered auto's 
expanded...
Mar 15 2012
next sibling parent Dmitry Olshansky <dmitry.olsh gmail.com> writes:
On 15.03.2012 12:35, Jay Norwood wrote:

 Is there some option, similar to -E gcc option, that would generate the
 analogous listing for D? What I mean is that I'd like to see all the
 code in version blocks gone, all the mixin strings expanded.

 It is probably too much to wish for having the infered auto's expanded...


Not every auto could be expanded, even simple stuff like:

auto add(T1,T2)(T1 t1, T2 t2){ return t1+t2; }

Expanding mixins could be useful, yet in certain circumstances I'd 
rather not see the resulting code :)

-- 
Dmitry Olshansky
Mar 15 2012
prev sibling next sibling parent reply Jacob Carlborg <doob me.com> writes:
On 2012-03-15 09:35, Jay Norwood wrote:

 Is there some option, similar to -E gcc option, that would generate the
 analogous listing for D? What I mean is that I'd like to see all the
 code in version blocks gone, all the mixin strings expanded.

 It is probably too much to wish for having the infered auto's expanded...


The Eclipse plugin, Descent, has a view that does something like this. 
Although I don't know how well it works for D2.

-- 
/Jacob Carlborg
Mar 15 2012
parent reply Jacob Carlborg <doob me.com> writes:
On 2012-03-15 11:04, Jacob Carlborg wrote:

 On 2012-03-15 09:35, Jay Norwood wrote:

 Is there some option, similar to -E gcc option, that would generate the
 analogous listing for D? What I mean is that I'd like to see all the
 code in version blocks gone, all the mixin strings expanded.

 It is probably too much to wish for having the infered auto's expanded...


 The Eclipse plugin, Descent, has a view that does something like this.
 Although I don't know how well it works for D2.


It expands mixins, string mixins, replaces scope statements with 
try/catch/finally and other things. It also has a compile time debugger.

-- 
/Jacob Carlborg
Mar 15 2012
parent reply "Jay Norwood" <jayn prismnet.com> writes:
On Thursday, 15 March 2012 at 10:09:25 UTC, Jacob Carlborg wrote:

 The Eclipse plugin, Descent, has a view that does something 
 like this.
 Although I don't know how well it works for D2.


 It expands mixins, string mixins, replaces scope statements 
 with try/catch/finally and other things. It also has a compile 
 time debugger.


Wow, that's pretty impressive.  Does it do aliases also?

I read a little more last night and saw that there is an option 
for json output  from dmd, and tried it out.  Is that meant to 
serve the purpose of the gccxml program?


I also found mention of a c++ compiler that someone states can 
compile c++ and output C.  Kind of weird, but maybe compiling D 
to expanded D would not be such a stretch.

http://stackoverflow.com/questions/1139793/c-template-preprocessor-tool
Mar 15 2012
next sibling parent =?utf-8?Q?Simen_Kj=C3=A6r=C3=A5s?= <simen.kjaras gmail.com> writes:
On Thu, 15 Mar 2012 14:34:19 +0100, Jay Norwood <jayn prismnet.com> wrote:


 I also found mention of a c++ compiler that someone states can compile  
 c++ and output C.  Kind of weird, but maybe compiling D to expanded D  
 would not be such a stretch.


That's how early C++ compilers worked. Actually, many compilers work/have
worked that way due to the ease of finding a C compiler for just about
any given platform.
Mar 15 2012
prev sibling parent Jacob Carlborg <doob me.com> writes:
On 2012-03-15 14:34, Jay Norwood wrote:

 On Thursday, 15 March 2012 at 10:09:25 UTC, Jacob Carlborg wrote:

 The Eclipse plugin, Descent, has a view that does something like this.
 Although I don't know how well it works for D2.


 It expands mixins, string mixins, replaces scope statements with
 try/catch/finally and other things. It also has a compile time debugger.


 Wow, that's pretty impressive. Does it do aliases also?


I think so and showing the actual type for type inference as well.


 I read a little more last night and saw that there is an option for json
 output from dmd, and tried it out. Is that meant to serve the purpose of
 the gccxml program?


 I also found mention of a c++ compiler that someone states can compile
 c++ and output C. Kind of weird, but maybe compiling D to expanded D
 would not be such a stretch.


I think LLVM can do this, convert C++ to C, via its byte code.

-- 
/Jacob Carlborg
Mar 15 2012
prev sibling next sibling parent Jonathan M Davis <jmdavisProg gmx.com> writes:
On Thursday, March 15, 2012 09:35:46 Jay Norwood wrote:

 Is there some option, similar to -E gcc option, that would
 generate the analogous listing for D?  What I mean is that I'd
 like to see all the code in version blocks gone, all the mixin
 strings expanded.
 
 It is probably too much to wish for having the infered auto's
 expanded...


http://d.puremagic.com/issues/show_bug.cgi?id=5051

- Jonathan M Davis
Mar 15 2012
prev sibling next sibling parent reply "Adam D. Ruppe" <destructionator gmail.com> writes:
On Thursday, 15 March 2012 at 08:35:48 UTC, Jay Norwood wrote:

 Is there some option, similar to -E gcc option, that would 
 generate the analogous listing for D?


You could add one to the compiler in just
a few lines; there's already a function that
does it, but it isn't called from anywhere.

Open up mars.c, and find this comment:
"Do not attempt to generate output files if errors"

It is line 1358 in my copy.


Right under that if(), add this:

         for (size_t i = 0; i < modules.dim; i++)
         {
             m = modules[i];
	    m->gensymfile();
	}

compile your new compiler.

BACK UP YOUR FILES because this function overwrites
the original .d file!


$ cat test10.d
void main() {
    auto a = 0;
}


lost!
$ cp test10.d test10_original.d


you bring in a dmd.conf...
$ d/dmd2/src/dmd/dmd -Id/dmd2/src/druntime/import 
-Id/dmd2/src/phobos -L-Ld/dmd2/linux/lib32/    test10.d

$ cat test10.d
// Sym file generated from 'test10.d'
import object;
void main()
{
int a = 0;
return 0;
}





It gets ugly if you use a lot of features because this
outputs the dmd translations - so foreach becomes for
and other lowerings.

But you can see basically what the compiler is going
to generate for final code.
Mar 15 2012
next sibling parent "H. S. Teoh" <hsteoh quickfur.ath.cx> writes:
On Thu, Mar 15, 2012 at 02:51:33PM +0100, Adam D. Ruppe wrote:

 On Thursday, 15 March 2012 at 08:35:48 UTC, Jay Norwood wrote:

Is there some option, similar to -E gcc option, that would
generate the analogous listing for D?


 
 You could add one to the compiler in just
 a few lines; there's already a function that
 does it, but it isn't called from anywhere.


[...]

 $ cat test10.d
 void main() {
    auto a = 0;
 }
 

 lost!
 $ cp test10.d test10_original.d
 

 bring in a dmd.conf...
 $ d/dmd2/src/dmd/dmd -Id/dmd2/src/druntime/import
 -Id/dmd2/src/phobos -L-Ld/dmd2/linux/lib32/    test10.d
 
 $ cat test10.d
 // Sym file generated from 'test10.d'
 import object;
 void main()
 {
 int a = 0;
 return 0;
 }


[...]

Whoa! This is cool. I might take a stab at making this a compiler option
(and writing the output to stdout instead of overwriting the original
source). I think this will be VERY useful when learning the language,
and also when debugging a compiler bug.


T

-- 
EMACS = Extremely Massive And Cumbersome System
Mar 15 2012
prev sibling parent reply "H. S. Teoh" <hsteoh quickfur.ath.cx> writes:
On Thu, Mar 15, 2012 at 02:51:33PM +0100, Adam D. Ruppe wrote:

 On Thursday, 15 March 2012 at 08:35:48 UTC, Jay Norwood wrote:

Is there some option, similar to -E gcc option, that would
generate the analogous listing for D?


 
 You could add one to the compiler in just
 a few lines; there's already a function that
 does it, but it isn't called from anywhere.


[...]

Alright, so I hacked dmd a little to make this a runtime option: I
changed the default symfile extension from .d to .ds so that it won't
overwrite the original source files, and added the option -sym to turn
it on/off. Check it out:

	https://github.com/quickfur/dmd/tree/gensymfile

Playing around with it a little, I notice that it doesn't output
template bodies, even though calls to template functions are left
intact. It also doesn't expand [] into opIndex/opIndexAssign, for
example. So you can't actually compile the output if you're using
templates, AFAICT.

I'm guessing that by this point the compiler has stored all templates in
a symbol table somewhere, so it only has the bare template declarations
left.


T

-- 
The peace of mind---from knowing that viruses which exploit Microsoft system
vulnerabilities cannot touch Linux---is priceless. -- Frustrated system
administrator.
Mar 16 2012
parent reply "Adam D. Ruppe" <destructionator gmail.com> writes:
On Saturday, 17 March 2012 at 00:13:33 UTC, H. S. Teoh wrote:

 Playing around with it a little, I notice that it doesn't output
 template bodies, even though calls to template functions are 
 left intact. It also doesn't expand [] into 
 opIndex/opIndexAssign,


I didn't play with this yet.. but it might not have to do this.

On arrays, [] is a primitive, not a function. opIndex should
only show up on custom types.

For the templates, I'd be surprised if they weren't there
somewhere. My D -> JS code works in the same location
(after running semantic) and templates just work in there.

The instantiations might be in a weird place though.
Mar 16 2012
parent "H. S. Teoh" <hsteoh quickfur.ath.cx> writes:
On Sat, Mar 17, 2012 at 01:21:56AM +0100, Adam D. Ruppe wrote:

 On Saturday, 17 March 2012 at 00:13:33 UTC, H. S. Teoh wrote:

Playing around with it a little, I notice that it doesn't output
template bodies, even though calls to template functions are left
intact. It also doesn't expand [] into opIndex/opIndexAssign,


 
 I didn't play with this yet.. but it might not have to do this.
 
 On arrays, [] is a primitive, not a function. opIndex should
 only show up on custom types.


Yes, but I was testing with my custom AA implementation.



 For the templates, I'd be surprised if they weren't there
 somewhere. My D -> JS code works in the same location
 (after running semantic) and templates just work in there.
 
 The instantiations might be in a weird place though.


Could be. I didn't really do much, just copied the code from your
previous email as-is. I don't know enough about dmd internals to be able
to do much more (though I'll have to get up to speed soon, if my AA
implementation ends up going into druntime).


T

-- 
IBM = I'll Buy Microsoft!
Mar 16 2012
prev sibling parent Trass3r <un known.com> writes:
There's a pull request to help with debugging string mixins:
https://github.com/D-Programming-Language/dmd/pull/426
Mar 15 2012