www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - compile time output

reply Trass3r <mrmocool gmx.de> writes:
Is there any way to output information at compile time other than 
pragma(msg?
pragma is driving me crazy, the following doesn't work:

auto members = __traits(allMembers, typeof(this));
foreach(m; members)
{
	pragma(msg, m);
}

-> Error: string expected for message, not 'm'

Though the docs clearly state:
allMembers: "An array of string literals is returned"
Also checked that, m is of type invariant(char)[].


Any ideas? -.-
Jan 20 2009
next sibling parent reply Bill Baxter <wbaxter gmail.com> writes:
On Wed, Jan 21, 2009 at 12:36 AM, Trass3r <mrmocool gmx.de> wrote:

 Is there any way to output information at compile time other than
 pragma(msg?
 pragma is driving me crazy, the following doesn't work:

 auto members = __traits(allMembers, typeof(this));
 foreach(m; members)
 {
        pragma(msg, m);
 }

 -> Error: string expected for message, not 'm'

 Though the docs clearly state:
 allMembers: "An array of string literals is returned"
 Also checked that, m is of type invariant(char)[].


 Any ideas? -.-


try indexing explicitly or using ref:

foreach(i,m; members)
{
    pragma(msg, members[i]);
}

foreach(ref m; members)
{
    pragma(msg, m);
}

Latter one may not be useful.  I can't recall.
--bb
Jan 20 2009
parent Trass3r <mrmocool gmx.de> writes:
Bill Baxter schrieb:

 try indexing explicitly or using ref:
 
 foreach(i,m; members)
 {
     pragma(msg, members[i]);
 }
 
 foreach(ref m; members)
 {
     pragma(msg, m);
 }
 
 Latter one may not be useful.  I can't recall.


Neither one works for me :(
Jan 20 2009
prev sibling next sibling parent Ary Borenszweig <ary esperanto.org.ar> writes:
Trass3r wrote:

 Is there any way to output information at compile time other than 
 pragma(msg?
 pragma is driving me crazy, the following doesn't work:
 
 auto members = __traits(allMembers, typeof(this));


Kind of offtopic, but I tried this (with typeof(Foo) and Foo is defined 
in the same module) in Descent and I get a NullPointerException. :-(

I thought the port to Java was wrong so I compared to dmd's code.

In traits.c I can see:

---
else if (ident == Id::allMembers || ident == Id::derivedMembers)
     {
	if (dim != 1)
	    goto Ldimerror;
	Object *o = (Object *)args->data[0];
	Dsymbol *s = getDsymbol(o);
---

and in template.c getDsymbol says:

---
Dsymbol *getDsymbol(Object *oarg)
{
     Dsymbol *sa;
     Expression *ea = isExpression(oarg);
     if (ea)
     {
         // (snip)
     }
     else
     {   // Try to convert Type to symbol
	Type *ta = isType(oarg);
	if (ta)
	    sa = ta->toDsymbol(NULL);
	else
	    sa = isDsymbol(oarg);	// if already a symbol
     }
     return sa;
}
---

and of course oarg is a type, it's a TypeTypeof, so 
TypeTypeof::toDsymbol is invoked. Note that in this point sc is NULL.

---
Dsymbol *TypeTypeof::toDsymbol(Scope *sc)
{
     Type *t;

     t = semantic(0, sc);
     if (t == this)
	return NULL;
     return t->toDsymbol(sc);
}
---

and finally...

---
Type *TypeTypeof::semantic(Loc loc, Scope *sc)
{   Expression *e;
     Type *t;

	sc->intypeof++; // sc is NULL!!
	exp = exp->semantic(sc);
	sc->intypeof--;

    // (snip)
}
---

and that's why I get a NPE. But compiling with dmd works fine. 
Unfortunately I don't have with me the necessary stuff to debug dmd... 
does anyone know what's going on?

Thanks,
Ary
Jan 20 2009
prev sibling next sibling parent reply BCS <ao pathlink.com> writes:
Reply to Trass3r,


 Is there any way to output information at compile time other than
 pragma(msg?
 pragma is driving me crazy, the following doesn't work:
 auto members = __traits(allMembers, typeof(this));
 foreach(m; members)
 {
 pragma(msg, m);
 }
 -> Error: string expected for message, not 'm'
 
 Though the docs clearly state:
 allMembers: "An array of string literals is returned"
 Also checked that, m is of type invariant(char)[].
 Any ideas? -.-
 


I don't do 2.0 but it looks to me like your mixing runtime and compile time 
stuff. A foreach over an array is a runtime foreach.
Jan 20 2009
parent reply Trass3r <mrmocool gmx.de> writes:
BCS schrieb:

 I don't do 2.0 but it looks to me like your mixing runtime and compile 
 time stuff. A foreach over an array is a runtime foreach.
 


Do you know how I could make it run at compile-time?
Jan 20 2009
parent reply BCS <ao pathlink.com> writes:
Reply to Trass3r,


 BCS schrieb:
 

 I don't do 2.0 but it looks to me like your mixing runtime and
 compile time stuff. A foreach over an array is a runtime foreach.
 


 Do you know how I could make it run at compile-time?
 


template Tpl(T...)
{
  alias T Tpl;
}

template Range(int l, int u)
{
  static if(l<u)
    alias Tpl!(l, Range!(l+1,u)) Range;
  else
    alias Tpl!(l) Range;
}

const char[][] set = ["Hello"[], "world" ];

void main()
{
  foreach(str; Range!(0,set.length-1)) // compile time foreach over the numbers 
from 0 to set.length-1
    pragma(msg, set[str]);
}
Jan 20 2009
parent reply Trass3r <mrmocool gmx.de> writes:
BCS schrieb:

 template Tpl(T...)
 {
  alias T Tpl;
 }
 
 template Range(int l, int u)
 {
  static if(l<u)
    alias Tpl!(l, Range!(l+1,u)) Range;
  else
    alias Tpl!(l) Range;
 }
 
 const char[][] set = ["Hello"[], "world" ];
 
 void main()
 {
  foreach(str; Range!(0,set.length-1)) // compile time foreach over the 
 numbers from 0 to set.length-1
    pragma(msg, set[str]);
 }
 


Still doesn't work for this code (used with a mixin):
template classMixin()
{
static this()
{
	auto members = __traits(allMembers, typeof(this));
	foreach(m; Range!(0, members.length-1))
	{
		pragma(msg, members[m]);
		static if (is (typeof(__traits(getMember, this, members[m])) F == 
function))
		{
			pragma(msg, "function");
		}
	}
}
}

-> Error: string expected for message, not 'members[0u]'

Leaving out the first pragma, compiles but doesn't output anything.
Replacing members[m] with a valid method name, such as "__ctor" it 
outputs function correctly :(
Jan 20 2009
parent reply BCS <ao pathlink.com> writes:
Reply to Trass3r,


 auto members = __traits(allMembers, typeof(this));


try switching that to 


 const[][] members = __traits(allMembers, typeof(this));


if that doesn't fix it try dropping this part (it might make it clearer what's 
going on)


 static if (is (typeof(__traits(getMember, this, members[m])) F ==
 function))
 {
 pragma(msg, "function");
 }


Looking at the rest of the thread, I think you might be looking at a bug.
Jan 20 2009
parent reply Trass3r <mrmocool gmx.de> writes:
This works:

template classMixin()
{
static this()
{
	const string[] members = __traits(allMembers, typeof(this));
	pragma(msg, foo(members));
}
}

string foo(const string[] members)
{
	string result;
	foreach(m; members)
		result ~= m ~ " ";
	return result;
}


But there must exist a better solution to this.
Jan 20 2009
parent reply Trass3r <mrmocool gmx.de> writes:
Trass3r schrieb:

 This works:
 
 string foo(const string[] members)
 {
     string result;
     foreach(m; members)
         result ~= m ~ " ";
     return result;
 }
 


Furthermore there seems to be no way to use these member names in a 
__traits(getMember, class, m) call.

This is driving me crazy.
Jan 20 2009
parent Bill Baxter <wbaxter gmail.com> writes:
On Wed, Jan 21, 2009 at 8:25 AM, Trass3r <mrmocool gmx.de> wrote:

 Trass3r schrieb:

 This works:

 string foo(const string[] members)
 {
    string result;
    foreach(m; members)
        result ~= m ~ " ";
    return result;
 }


 Furthermore there seems to be no way to use these member names in a
 __traits(getMember, class, m) call.

 This is driving me crazy.


Well, it does start with two underscores.  That may be trying to tell
you something. :-)

If you find a real bug definitely report it.   Also note that your bug
reports will be treated much more seriously if you can provide a
compelling use case which is difficult or impossible to work around
short of fixing the bug.  "If you fix this then I will be able create
cool thing X" seems to increase your chances of getting a fix
significantly.

--bb
Jan 20 2009
prev sibling parent reply Sergey Gromov <snake.scaly gmail.com> writes:
Tue, 20 Jan 2009 16:36:09 +0100, Trass3r wrote:


 Is there any way to output information at compile time other than 
 pragma(msg?
 pragma is driving me crazy, the following doesn't work:
 
 auto members = __traits(allMembers, typeof(this));
 foreach(m; members)
 {
 	pragma(msg, m);
 }
 
 -> Error: string expected for message, not 'm'
 
 Though the docs clearly state:
 allMembers: "An array of string literals is returned"
 Also checked that, m is of type invariant(char)[].
 
 Any ideas? -.-


Weird.  The following code does not compile:

class Cls
{
  int bar;
  char[] baz;
}
string foo()
{
  auto members = __traits(allMembers, Cls);
  return "";
}
pragma(msg, foo());


dmd -c test.d


test.d(11): Error: cannot evaluate foo() at compile time
test.d(11): pragma msg string expected for message, not 'foo()'

Comment out the traits and it compiles.  Traits are supposed to be
compile-time.  How's that possible for them to prevent compile-time
evaluation?
Jan 20 2009
next sibling parent reply Trass3r <mrmocool gmx.de> writes:
Sergey Gromov schrieb:
  > class Cls

 {
   int bar;
   char[] baz;
 }
 string foo()
 {
   auto members = __traits(allMembers, Cls);
   return "";
 }
 pragma(msg, foo());
 

 dmd -c test.d


 test.d(11): Error: cannot evaluate foo() at compile time
 test.d(11): pragma msg string expected for message, not 'foo()'
 
 Comment out the traits and it compiles.  Traits are supposed to be
 compile-time.  How's that possible for them to prevent compile-time
 evaluation?


Seeing this really simple example crashing makes me think that this has 
to be a bug.
Jan 20 2009
parent reply BCS <ao pathlink.com> writes:
Reply to Trass3r,


 Sergey Gromov schrieb:
 

 auto members = __traits(allMembers, Cls);


 Seeing this really simple example crashing makes me think that this
 has to be a bug.
 


that auto might be mucking it up (if so it would be a bug).
Jan 20 2009
parent Sergey Gromov <snake.scaly gmail.com> writes:
Tue, 20 Jan 2009 21:12:18 +0000 (UTC), BCS wrote:


 Reply to Trass3r,
 

 Sergey Gromov schrieb:
 

 auto members = __traits(allMembers, Cls);


 Seeing this really simple example crashing makes me think that this
 has to be a bug.
 


 
 that auto might be mucking it up (if so it would be a bug).


Bingo. Replacing it with

  string[] members = __traits(allMembers, Cls);

makes it compile.  Therefore 

class Cls
{
  int bar;
  char[] baz;
}
string foo()
{
  string[] members = __traits(allMembers, Cls);
  string result;
  foreach(m; members)
    result ~= m ~ " ";
  return result;
}
pragma(msg, foo());

works.
Jan 20 2009
prev sibling parent Christopher Wright <dhasenan gmail.com> writes:
Sergey Gromov wrote:

 Comment out the traits and it compiles.  Traits are supposed to be
 compile-time.  How's that possible for them to prevent compile-time
 evaluation?


It's the amazing powers of the DMD CTFE engine! And it's why I don't use 
d2 these days.

I think I'll dust off some old code I have that worked with dmd 2.009 
(and failed with 2.010 and 2.011) and try making it work. I don't think 
the bugs that I filed in reference to that code ever got fixed, though.
Jan 20 2009