D - typedef bug

"Jon Frechette" <jonf4 mindspring.com> Jul 29 2003

"Sean L. Palmer" <palmer.sean verizon.net> Jul 29 2003

"Jon Frechette" <jonf4 mindspring.com> Jul 30 2003

Burton Radons <loth users.sourceforge.net> Jul 30 2003

"Jon Frechette" <jonf4 mindspring.com> Jul 31 2003
"Walter" <walter digitalmars.com> Aug 01 2003

Burton Radons <loth users.sourceforge.net> Aug 01 2003

"Walter" <walter digitalmars.com> Aug 02 2003

Russ Lewis <spamhole-2001-07-16 deming-os.org> Jul 29 2003

"Jon Frechette" <jonf4 mindspring.com> Jul 30 2003

Russ Lewis <spamhole-2001-07-16 deming-os.org> Jul 30 2003

"Jon Frechette" <jonf4 mindspring.com> writes:

Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

This code is fine:

void foo( char [] word, out int[ char [] ] wordList )
{
  if ( word in wordList ){}
}

but this code:

typedef int[ char [] ] T;

void foo( char [] word, out T wordList )
{
  if ( word in wordList ){}
}

produces this error:

  "rvalue of in expression must be an associative array, not T"

Jul 29 2003

"Sean L. Palmer" <palmer.sean verizon.net> writes:

Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

Why "out"?  Surely you want to preserve the prior value of wordList?

Sean
  "Jon Frechette" <jonf4 mindspring.com> wrote in message =
news:bg5ulg$1lsh$1 digitaldaemon.com...
  This code is fine:

  void foo( char [] word, out int[ char [] ] wordList )
  {
    if ( word in wordList ){}
  }

  but this code:

  typedef int[ char [] ] T;=20

  void foo( char [] word, out T wordList )
  {
    if ( word in wordList ){}
  }

  produces this error:

    "rvalue of in expression must be an associative array, not T"

Jul 29 2003

"Jon Frechette" <jonf4 mindspring.com> writes:

Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

No, the parameter wordList needs to be modified. If I had left the code =
snippet closer to the original code this would have been clearer.

void addWordToList( char [] word, out int[ char [] ] wordList )
{
  word =3D tolower( word );
  wordList[ word ]++;
  // other stuff
}

  "Sean L. Palmer" <palmer.sean verizon.net> wrote in message =
news:bg659t$1sit$1 digitaldaemon.com...
  Why "out"?  Surely you want to preserve the prior value of wordList?

  Sean
    "Jon Frechette" <jonf4 mindspring.com> wrote in message =
news:bg5ulg$1lsh$1 digitaldaemon.com...
    This code is fine:

    void foo( char [] word, out int[ char [] ] wordList )
    {
      if ( word in wordList ){}
    }

    but this code:

    typedef int[ char [] ] T;=20

    void foo( char [] word, out T wordList )
    {
      if ( word in wordList ){}
    }

    produces this error:

      "rvalue of in expression must be an associative array, not T"

Jul 30 2003

Burton Radons <loth users.sourceforge.net> writes:

Jon Frechette wrote:

 No, the parameter wordList needs to be modified. If I had left the code 
 snippet closer to the original code this would have been clearer.
  
 void addWordToList( char [] word, out int[ char [] ] wordList )
 {
   word = tolower( word );
   wordList[ word ]++;
   // other stuff
 }


Then you'll want to use "inout" because "out" clears the parameter to 
its initial value at the start of the function.  Last I checked this 
wasn't in the prototype, and that's a bug.

As is the error you reported.  Russ was wrong - you have to explicitly 
cast to turn a base type into a typedef, but you can implicitly cast to 
turn a typedef into a base type.  If your problem wasn't a bug, then it 
would be impossible to access or modify a typedef'd associative array!

So, Walter, here's a bug report:

    typedef int [int] a;

    void foo ()
    {
        1 in a;
    }

Compiling this with DMD 0.68 fails with the error message "f.d(5): 
rvalue of in expression must be an associative array, not a".  It should 
compile.

Jul 30 2003

"Jon Frechette" <jonf4 mindspring.com> writes:

"Burton Radons" <loth users.sourceforge.net> wrote in message
news:bg862s$vur$1 digitaldaemon.com...
 Jon Frechette wrote:

 No, the parameter wordList needs to be modified. If I had left the code
 snippet closer to the original code this would have been clearer.

 void addWordToList( char [] word, out int[ char [] ] wordList )
 {
   word = tolower( word );
   wordList[ word ]++;
   // other stuff
 }


 Then you'll want to use "inout" because "out" clears the parameter to
 its initial value at the start of the function.  Last I checked this
 wasn't in the prototype, and that's a bug.


Something is wrong here because the addWordToList function is "working" as
expected in my program. The wordList is not cleared on each call to the
function. Is this a bug?

 As is the error you reported.  Russ was wrong - you have to explicitly
 cast to turn a base type into a typedef, but you can implicitly cast to
 turn a typedef into a base type.  If your problem wasn't a bug, then it
 would be impossible to access or modify a typedef'd associative array!

 So, Walter, here's a bug report:

     typedef int [int] a;

     void foo ()
     {
         1 in a;
     }

 Compiling this with DMD 0.68 fails with the error message "f.d(5):
 rvalue of in expression must be an associative array, not a".  It should
 compile.

Jul 31 2003

"Walter" <walter digitalmars.com> writes:

"Burton Radons" <loth users.sourceforge.net> wrote in message
news:bg862s$vur$1 digitaldaemon.com...
 So, Walter, here's a bug report:

     typedef int [int] a;

     void foo ()
     {
         1 in a;
     }

 Compiling this with DMD 0.68 fails with the error message "f.d(5):
 rvalue of in expression must be an associative array, not a".  It should
 compile.


It should not compile, since a is a type, not a value.

Aug 01 2003

Burton Radons <loth users.sourceforge.net> writes:

Walter wrote:

 "Burton Radons" <loth users.sourceforge.net> wrote in message
 news:bg862s$vur$1 digitaldaemon.com...
 
So, Walter, here's a bug report:

    typedef int [int] a;

    void foo ()
    {
        1 in a;
    }

Compiling this with DMD 0.68 fails with the error message "f.d(5):
rvalue of in expression must be an associative array, not a".  It should
compile.


 
 It should not compile, since a is a type, not a value.


My bad.  DMD 0.68 under Windows XP:

     typedef int [int] aType;
     aType a;

     void foo ()
     {
        1 in a;
     }

This fails compilation using "dmd f.d" with the error message "f.d(6): 
rvalue of in expression must be an associative array, not aType".  It 
should compile.

Aug 01 2003

"Walter" <walter digitalmars.com> writes:

"Burton Radons" <loth users.sourceforge.net> wrote in message
news:bge8fs$s2i$1 digitaldaemon.com...
 My bad.  DMD 0.68 under Windows XP:

      typedef int [int] aType;
      aType a;

      void foo ()
      {
         1 in a;
      }

 This fails compilation using "dmd f.d" with the error message "f.d(6):
 rvalue of in expression must be an associative array, not aType".  It
 should compile.


Ok, I can take care of that one <g>.

Aug 02 2003

Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

Jon Frechette wrote:
 This code is fine:
 
 void foo( char [] word, out int[ char [] ] wordList )
 {
   if ( word in wordList ){}
 }
  
 but this code:
  
 typedef int[ char [] ] T;
  
 void foo( char [] word, out T wordList )
 {
   if ( word in wordList ){}
 }
  
 produces this error:
 
   "rvalue of in expression must be an associative array, not T"


This behavior is intentional.  When you typedef in D, it creates a new 
type, and the new type does not have implicit conversion with the old 
type.  This is to prevent bad coding styles like this:

HandleToSomething foo();

void bar()
{
   int a = foo();
};

The person who calls foo() should store the return value as 
HandleToSomething, NOT as an int.  True, in the current version of 
foo()'s library, maybe it is an int...but it could change to some other 
implementation in the future.

If you have to do the conversion, then you can explicitly cast it:

int b = (int)foo();

If you want a C-style typedef (where there are implicit casts defined), 
then in D you use an alias:

alias int OtherHandleType;
OtherHandleType c;
...
int d = c;	// this is ok, because it was an alias, not a typedef

Jul 29 2003

"Jon Frechette" <jonf4 mindspring.com> writes:

"Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
news:bg6k33$2c5k$1 digitaldaemon.com...
 Jon Frechette wrote:
 This code is fine:

 void foo( char [] word, out int[ char [] ] wordList )
 {
   if ( word in wordList ){}
 }

 but this code:

 typedef int[ char [] ] T;

 void foo( char [] word, out T wordList )
 {
   if ( word in wordList ){}
 }

 produces this error:

   "rvalue of in expression must be an associative array, not T"


 This behavior is intentional.  When you typedef in D, it creates a new
 type, and the new type does not have implicit conversion with the old
 type.  This is to prevent bad coding styles like this:

 HandleToSomething foo();

 void bar()
 {
    int a = foo();
 };

 The person who calls foo() should store the return value as
 HandleToSomething, NOT as an int.  True, in the current version of
 foo()'s library, maybe it is an int...but it could change to some other
 implementation in the future.

 If you have to do the conversion, then you can explicitly cast it:

 int b = (int)foo();

 If you want a C-style typedef (where there are implicit casts defined),
 then in D you use an alias:

 alias int OtherHandleType;
 OtherHandleType c;
 ...
 int d = c; // this is ok, because it was an alias, not a typedef


OK, I guess I understand this now. But then when would one want to use
typedef  instead of alias?

Jul 30 2003

Russ Lewis <spamhole-2001-07-16 deming-os.org> writes:

You use typdef when you want to create a new type that inherits its 
internal impelmentation from something else.  So you want to use typedef 
for things like handles, dev_t, size_t, and other things.  These are all 
types that are probably integers (or pointers) in their internal 
implementation, but generally you don't want the users to think of them 
that way.

You use alias when you just want a shortcut to access a type.  Like many 
of us alias 'bit' to 'bool' since the compler doesn't come with a 
builtin bool type.

Jon Frechette wrote:
 "Russ Lewis" <spamhole-2001-07-16 deming-os.org> wrote in message
 news:bg6k33$2c5k$1 digitaldaemon.com...
 
Jon Frechette wrote:

This code is fine:

void foo( char [] word, out int[ char [] ] wordList )
{
  if ( word in wordList ){}
}

but this code:

typedef int[ char [] ] T;

void foo( char [] word, out T wordList )
{
  if ( word in wordList ){}
}

produces this error:

  "rvalue of in expression must be an associative array, not T"


This behavior is intentional.  When you typedef in D, it creates a new
type, and the new type does not have implicit conversion with the old
type.  This is to prevent bad coding styles like this:

HandleToSomething foo();

void bar()
{
   int a = foo();
};

The person who calls foo() should store the return value as
HandleToSomething, NOT as an int.  True, in the current version of
foo()'s library, maybe it is an int...but it could change to some other
implementation in the future.

If you have to do the conversion, then you can explicitly cast it:

int b = (int)foo();

If you want a C-style typedef (where there are implicit casts defined),
then in D you use an alias:

alias int OtherHandleType;
OtherHandleType c;
...
int d = c; // this is ok, because it was an alias, not a typedef


 
 OK, I guess I understand this now. But then when would one want to use
 typedef  instead of alias?

Jul 30 2003

D Programming

C/C++ Programming

Other

D - typedef bug