Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

It would be the sugar syntactic to add '??' operator to D. Consider the example
code:

string error_message = getErrorMessage() ?? "no errors";
A a = x.getPreparedAObject() ?? y.getPreparedAObject() ?? new A();


the first non-null value or to null if all rvalues contains null. This operator
might be extended to '???' to evaluate to a value that conforms to some
conditions. The lvalue gets the first rvalue that evaluates to true (evaluation
goes from left to right). If no conditions evaluates to true then the lvalue
stays unchanged. If no conditions are given explicitly then those conditions
evaluates to true (so, the best place for them is at the end of a sequence).

For instance:

int codeval = getValue1() > 5 ??? getValue2() >= 4 ??? getValue3() != 0 ??? 1;
// if no conditions are satisfied then codeval = 1 as a default value at last.
There are no sane reasons to place it in the middle of the sequence.

int code = 7;
code = 5 < 10 ??? 0 != 0; // this way, code = 7 (no conditions are satisfied)

"Vladimir Panteleev" <thecybershadow gmail.com> writes:

On Sun, 23 Sep 2007 18:34:09 +0300, Arlen Albert Keshabyan <arlen.albert=
 gmail.com> wrote:

[regarding ???]
 The lvalue gets the first rvalue that evaluates to true (evaluation go=

es from left to right). =


So the result can only be boolean?
In this case, I don't understand how
  a =3D b ??? c ??? d;
is different from
  a |=3D b || c || d;

If b, c and d are false then a is unchanged.

-- =

Best regards,
 Vladimir                          mailto:thecybershadow gmail.com

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Vladimir Panteleev Wrote:

 On Sun, 23 Sep 2007 18:34:09 +0300, Arlen Albert Keshabyan
<arlen.albert gmail.com> wrote:
 
 [regarding ???]
 The lvalue gets the first rvalue that evaluates to true (evaluation goes from
left to right). 

 
 So the result can only be boolean?
 In this case, I don't understand how
   a = b ??? c ??? d;
 is different from
   a |= b || c || d;
 
 If b, c and d are false then a is unchanged.
 
 -- 
 Best regards,
  Vladimir                          mailto:thecybershadow gmail.com


Consider this:

int v = 15;
int v2 = 30;

int a = v > 20 ??? v = 17 ??? v2 < 25 ??? 5;

in this case 'a' evaluates to 5 because none of the conditions evaluate to true
except the last one. if you remove the last condition (??? 5) then 'a' stays 0
in this case.

If v2 = 23 then 'a' evaluates to 23;
If v = 27 then 'a' evaluates to 27;

etc.

It means that it takes a sequential rvalue, tests it against a condition and if
the condition evaluates to true then it assigns that rvalue to lvalue. It does
not mean you are restricted to boolean values only.

Chris Nicholson-Sauls <ibisbasenji gmail.com> writes:

Arlen Albert Keshabyan wrote:
 Vladimir Panteleev Wrote:
 
 On Sun, 23 Sep 2007 18:34:09 +0300, Arlen Albert Keshabyan
<arlen.albert gmail.com> wrote:

 [regarding ???]
 The lvalue gets the first rvalue that evaluates to true (evaluation goes from
left to right). 

 So the result can only be boolean?
 In this case, I don't understand how
   a = b ??? c ??? d;
 is different from
   a |= b || c || d;

 If b, c and d are false then a is unchanged.

 -- 
 Best regards,
  Vladimir                          mailto:thecybershadow gmail.com

 
 
 Consider this:
 
 int v = 15;
 int v2 = 30;
 
 int a = v > 20 ??? v = 17 ??? v2 < 25 ??? 5;
 
 in this case 'a' evaluates to 5 because none of the conditions evaluate to
true except the last one. if you remove the last condition (??? 5) then 'a'
stays 0 in this case.
 
 If v2 = 23 then 'a' evaluates to 23;
 If v = 27 then 'a' evaluates to 27;
 
 etc.
 
 It means that it takes a sequential rvalue, tests it against a condition and
if the condition evaluates to true then it assigns that rvalue to lvalue. It
does not mean you are restricted to boolean values only.

Or in other words this:

int a = v   > 20
     ??? v  == 17
     ??? v2  < 25
     ??? 5
;

Is supposed to be shorthand for this?

int a;
if (v  > 20 ||v == 17)
     a = v;
else if (v2 < 25)
     a = v2;
else
     a = 5;


...its an interesting thought, but it just feels more like a scripting language
feature 
than that of a systems language.  The other '??' version that just skips nulls
could be 
useful at times, though.  Hmm.

-- Chris Nicholson-Sauls

"Vladimir Panteleev" <thecybershadow gmail.com> writes:

On Sun, 23 Sep 2007 20:47:08 +0300, Chris Nicholson-Sauls
<ibisbasenji gmail.com> wrote:

 ...its an interesting thought, but it just feels more like a scripting
language feature than that of a systems language.  The other '??' version that
just skips nulls could be useful at times, though.  Hmm.

I think this is possible to implement via templates - or, if not, it will be
with macros (in the form of "firstNonNull(b, c, d)"). 

-- 
Best regards,
 Vladimir                          mailto:thecybershadow gmail.com

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Vladimir Panteleev Wrote:

 On Sun, 23 Sep 2007 20:47:08 +0300, Chris Nicholson-Sauls
<ibisbasenji gmail.com> wrote:
 
 ...its an interesting thought, but it just feels more like a scripting
language feature than that of a systems language.  The other '??' version that
just skips nulls could be useful at times, though.  Hmm.

 
 I think this is possible to implement via templates - or, if not, it will be
with macros (in the form of "firstNonNull(b, c, d)"). 
 
 -- 
 Best regards,
  Vladimir                          mailto:thecybershadow gmail.com

Well, maybe you're right.

Jari-Matti =?ISO-8859-1?Q?M=E4kel=E4?= <jmjmak utu.fi.invalid> writes:

Chris Nicholson-Sauls wrote:

 Or in other words this:
 
 int a = v   > 20
      ??? v  == 17
      ??? v2  < 25
      ??? 5
 ;
 
 Is supposed to be shorthand for this?
 
 int a;
 if (v  > 20 ||v == 17)
      a = v;
 else if (v2 < 25)
      a = v2;
 else
      a = 5;
 

D could have if and case expressions instead of statements for handling
these. Something like Haskell guards and cases:

http://haskell.org/tutorial/patterns.html#sect4.3
http://haskell.org/tutorial/patterns.html#sect4.1

Or if the coming metaprogramming capabilities will be strong enough, those
could be possible to implement on the library level, which would be nice
too (yes, mixin+ctfe makes them already possible, but is ugly).

 The other '??' version that just skips nulls could be
 useful at times, though.

But it's limited to this one tiny purpose.

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Chris Nicholson-Sauls Wrote:

 Arlen Albert Keshabyan wrote:
 Vladimir Panteleev Wrote:
 
 On Sun, 23 Sep 2007 18:34:09 +0300, Arlen Albert Keshabyan
<arlen.albert gmail.com> wrote:

 [regarding ???]
 The lvalue gets the first rvalue that evaluates to true (evaluation goes from
left to right). 

 So the result can only be boolean?
 In this case, I don't understand how
   a = b ??? c ??? d;
 is different from
   a |= b || c || d;

 If b, c and d are false then a is unchanged.

 -- 
 Best regards,
  Vladimir                          mailto:thecybershadow gmail.com

 
 
 Consider this:
 
 int v = 15;
 int v2 = 30;
 
 int a = v > 20 ??? v = 17 ??? v2 < 25 ??? 5;
 
 in this case 'a' evaluates to 5 because none of the conditions evaluate to
true except the last one. if you remove the last condition (??? 5) then 'a'
stays 0 in this case.
 
 If v2 = 23 then 'a' evaluates to 23;
 If v = 27 then 'a' evaluates to 27;
 
 etc.
 
 It means that it takes a sequential rvalue, tests it against a condition and
if the condition evaluates to true then it assigns that rvalue to lvalue. It
does not mean you are restricted to boolean values only.

 
 Or in other words this:
 
 int a = v   > 20
      ??? v  == 17
      ??? v2  < 25
      ??? 5
 ;
 
 Is supposed to be shorthand for this?
 
 int a;
 if (v  > 20 ||v == 17)
      a = v;
 else if (v2 < 25)
      a = v2;
 else
      a = 5;
 
 
 ...its an interesting thought, but it just feels more like a scripting
language feature 
 than that of a systems language.  The other '??' version that just skips nulls
could be 
 useful at times, though.  Hmm.
 
 -- Chris Nicholson-Sauls


Yes. Exactly.

"Janice Caron" <caron800 googlemail.com> writes:

On 9/23/07, Arlen Albert Keshabyan <arlen.albert gmail.com> wrote:
 Consider this:

 int v = 15;
 int v2 = 30;

 int a = v > 20 ??? v = 17 ??? v2 < 25 ??? 5;

 in this case 'a' evaluates to 5 because none of the conditions evaluate to
true except the last one. if you remove the last condition (??? 5) then 'a'
stays 0 in this case.

 If v2 = 23 then 'a' evaluates to 23;
 If v = 27 then 'a' evaluates to 27;

 etc.

 It means that it takes a sequential rvalue, tests it against a condition and
if the condition evaluates to true then it assigns that rvalue to lvalue. It
does not mean you are restricted to boolean values only.

I assume you meant v == 17, not v = 17 there, since v = 17 would be an
assignment. In any case

int a = v > 20 ??? v == 17 ??? v2 < 25 ??? 5;

Looks to me the same as

int a = v > 20 ? v : (v == 17 ? v : (v2 < 25 ? v2 : 5));

Could be just me, but the latter seems readable and more expressive.

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Janice Caron Wrote:

 On 9/23/07, Arlen Albert Keshabyan <arlen.albert gmail.com> wrote:
 Consider this:

 int v = 15;
 int v2 = 30;

 int a = v > 20 ??? v = 17 ??? v2 < 25 ??? 5;

 in this case 'a' evaluates to 5 because none of the conditions evaluate to
true except the last one. if you remove the last condition (??? 5) then 'a'
stays 0 in this case.

 If v2 = 23 then 'a' evaluates to 23;
 If v = 27 then 'a' evaluates to 27;

 etc.

 It means that it takes a sequential rvalue, tests it against a condition and
if the condition evaluates to true then it assigns that rvalue to lvalue. It
does not mean you are restricted to boolean values only.

 
 I assume you meant v == 17, not v = 17 there, since v = 17 would be an
 assignment. In any case
 
 int a = v > 20 ??? v == 17 ??? v2 < 25 ??? 5;
 
 Looks to me the same as
 
 int a = v > 20 ? v : (v == 17 ? v : (v2 < 25 ? v2 : 5));
 
 Could be just me, but the latter seems readable and more expressive.

Well, maybe you're right.

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

and...

A a = APointer1 ?? APointer2 ?? APointer3;

the line above is equal to the line below:

A a = APointer1 != null ??? APointer2 != null ??? APointer3 != null ??? null;

"Stewart Gordon" <smjg_1998 yahoo.com> writes:

"Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message 
news:fd68lm$f4r$1 digitalmars.com...
 and...

 A a = APointer1 ?? APointer2 ?? APointer3;

 the line above is equal to the line below:

 A a = APointer1 != null ??? APointer2 != null ??? APointer3 != null ??? 
 null;

Not by the way you specified it.  The first would assign to a an actual 
object reference; the second would try to assign a boolean value.

Stewart.

-- 
My e-mail address is valid but not my primary mailbox.  Please keep replies 
on the 'group where everybody may benefit. 

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Stewart Gordon Wrote:

 
 "Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message 
 news:fd68lm$f4r$1 digitalmars.com...
 and...

 A a = APointer1 ?? APointer2 ?? APointer3;

 the line above is equal to the line below:

 A a = APointer1 != null ??? APointer2 != null ??? APointer3 != null ??? 
 null;

 
 Not by the way you specified it.  The first would assign to a an actual 
 object reference; the second would try to assign a boolean value.
 
 Stewart.
 
 -- 
 My e-mail address is valid but not my primary mailbox.  Please keep replies 
 on the 'group where everybody may benefit. 
 

No. The code lines do the same thing anyway.

"Stewart Gordon" <smjg_1998 yahoo.com> writes:

"Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message 
news:fd6ee0$nd4$1 digitalmars.com...
 Stewart Gordon Wrote:
 "Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message
 news:fd68lm$f4r$1 digitalmars.com...
 A a = APointer1 ?? APointer2 ?? APointer3;

 the line above is equal to the line below:

 A a = APointer1 != null ??? APointer2 != null ??? APointer3 != null ???
 null;

 Not by the way you specified it.  The first would assign to a an actual
 object reference; the second would try to assign a boolean value.


<snip>
 No. The code lines do the same thing anyway.

Which same thing - evaluating to a boolean or evaluating to an object 
reference?  Are you going to supply a corrected version of your proposal?

Stewart.

-- 
My e-mail address is valid but not my primary mailbox.  Please keep replies 
on the 'group where everybody may benefit. 

"Stewart Gordon" <smjg_1998 yahoo.com> writes:

"Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message
news:fd611h$4co$1 digitalmars.com...
 It would be the sugar syntactic to add '??' operator to D.
 Consider the example code:

 string error_message = getErrorMessage() ??  "no errors";
 A a = x.getPreparedAObject() ??  y.getPreparedAObject() ??  new A();


 evaluates to the first non-null value or to null if all rvalues
 contains null.

So effectively, it works like || in JavaScript and the like.  I guess the 
return type of a ??  expression would be determined by the same rules that 
govern that of a ConditionalExpression.  The expression would evaluate to 
the first subexpression whose value when implicitly converted to a boolean 
is true.  Otherwise ...  to the .init of the return type?

 This operator might be extended to '???' to evaluate to a value
 that conforms to some conditions.  The lvalue gets the first rvalue
 that evaluates to true (evaluation goes from left to right).  If no
 conditions evaluates to true then the lvalue stays unchanged.  If
 no conditions are given explicitly then those conditions evaluates
 to true (so, the best place for them is at the end of a sequence).

<snip>

I don't really like this:
- It would cause the semantics of the = operator to depend on the form of 
the RHS.
- What if the ??? expression isn't the RHS of an = operator?
- The null case of this operator doesn't match semantically as they're 
normally expected to.  (I'm not sure if there's any better way to word 
this.)  To see what I mean, compare the meanings of

a = b ??? c ??? d;
a = b ??? c;
a = b;

Stewart.

-- 
My e-mail address is valid but not my primary mailbox.  Please keep replies 
on the 'group where everybody may benefit. 

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Stewart Gordon Wrote:

 "Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message
 news:fd611h$4co$1 digitalmars.com...
 It would be the sugar syntactic to add '??' operator to D.
 Consider the example code:

 string error_message = getErrorMessage() ??  "no errors";
 A a = x.getPreparedAObject() ??  y.getPreparedAObject() ??  new A();


 evaluates to the first non-null value or to null if all rvalues
 contains null.

 
 So effectively, it works like || in JavaScript and the like.  I guess the 
 return type of a ??  expression would be determined by the same rules that 
 govern that of a ConditionalExpression.  The expression would evaluate to 
 the first subexpression whose value when implicitly converted to a boolean 
 is true.  Otherwise ...  to the .init of the return type?
 
 This operator might be extended to '???' to evaluate to a value
 that conforms to some conditions.  The lvalue gets the first rvalue
 that evaluates to true (evaluation goes from left to right).  If no
 conditions evaluates to true then the lvalue stays unchanged.  If
 no conditions are given explicitly then those conditions evaluates
 to true (so, the best place for them is at the end of a sequence).

 <snip>
 
 I don't really like this:
 - It would cause the semantics of the = operator to depend on the form of 
 the RHS.
 - What if the ??? expression isn't the RHS of an = operator?
 - The null case of this operator doesn't match semantically as they're 
 normally expected to.  (I'm not sure if there's any better way to word 
 this.)  To see what I mean, compare the meanings of
 
 a = b ??? c ??? d;
 a = b ??? c;
 a = b;
 
 Stewart.
 
 -- 
 My e-mail address is valid but not my primary mailbox.  Please keep replies 
 on the 'group where everybody may benefit. 
 

If you'd read my first post carefully you'd never ask the question like this.

 a = b ??? c ??? d;
 a = b ??? c;
 a = b;

the conditions must be explicit (!). Inexplicit conditions always evaluates to
TRUE despite types involved in conditions. So, you will got the following:

 a = b ??? c ??? d;

a = b; //any type, even boolean


 a = b ??? c;

a = b;

 a = b;

a = b; //so no ??? ternary operator. Treated just like no (?:) operator :)



bool a = true
bool b = true
bool c = false
bool d = true;

a = b != true ??? c != false ??? d == false ??? false;

now a == false because no conditions evaluates to true except for the last one
'false'. Yes, 'false' evaluates to true :) so it's assigned to the 'a' variable.

"Stewart Gordon" <smjg_1998 yahoo.com> writes:

"Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message 
news:fd7qu5$2t7e$1 digitalmars.com...
<snip>
 If you'd read my first post carefully you'd never ask the question like 
 this.

 a = b ??? c ??? d;
 a = b ??? c;
 a = b;

 the conditions must be explicit (!). Inexplicit conditions always 
 evaluates to TRUE despite types involved in conditions. So, you will got 
 the following:

I meant the symbols to denote general expressions rather than variables.  I 
must've blinked and missed the bit about the semantics of a ??? expression 
depending on the form of the operands.

But anyway, what is the definition of an "explicit" or "inexplicit" 
condition"?  Just whether the operator in the next level below the ??? 
happens to be == != is !is < <= > >= !<>= !<> <> <>= !> !>= !< !<= in?

And if you're going to do this, why _allow_ "inexplicit" conditions anywhere 
except the end?  ISTM it can only be a mistake, and one that's far too easy 
to make.

 a = b ??? c ??? d;

 a = b; //any type, even boolean

 a = b ??? c;

 a = b;

 a = b;

 a = b; //so no ??? ternary operator. Treated just like no (?:) operator :)

<snip>

You miss the point.

Generally, when a binary operator is used n times in sequence, it denotes 
combining the n+1 operands in some fashion or another, which continues to 
hold when n == 0.  For instance, x can be thought of as a sum of one value, 
just as x + y is a sum of two values, x + y + z is a sum of three values, 
and so on.  Your ??? operator would break this principle:

n = a == x ??? b == y ??? c == z;
    takes, of the three conditions, the first condition to be true and 
assigns it to n, otherwise leaves n unchanged

n = a == x ??? b == y;
    takes, of the two conditions, the first condition to be true and assigns 
it to n, otherwise leaves n unchanged

n = a == x;
    would by extension take, of the one condition, the first condition to be 
true and assign it to n, otherwise leave n unchanged.  (In other words, 
assign the value of a == x to n only if a == x is true.)  But it doesn't....

Whether this matters, I'm not sure.  At least

    n = a == x ??? n;

would do the equivalent for a single condition.  But still....

Stewart.

-- 
My e-mail address is valid but not my primary mailbox.  Please keep replies 
on the 'group where everybody may benefit. 

renoX <renosky free.fr> writes:

Arlen Albert Keshabyan a écrit :
 It would be the sugar syntactic to add '??' operator to D. Consider
 the example code:
 
 string error_message = getErrorMessage() ?? "no errors"; A a =
 x.getPreparedAObject() ?? y.getPreparedAObject() ?? new A();
 

 evaluates to the first non-null value or to null if all rvalues
 contains null.[cut]


I feel as if the ?? operator is a workaround for the real issue, the 
fact that types reference can be null by default.

In Nice by default, types are not nullable so you don't need to put null
tests everywhere and there's a nice syntax for nullable type: just 
append a '?' to your type name i.e:

Type x = ...; // Here you're sure that x can never be null.
Type? x; // x may be null.


This is a big change, but it feels "more right" to me.

renoX

Frank Benoit <keinfarbton googlemail.com> writes:

 
 Type x = ...; // Here you're sure that x can never be null.
 Type? x; // x may be null.
 
 
 This is a big change, but it feels "more right" to me.
 
 renoX
 

I second that.

Derek Parnell <derek psych.ward> writes:

On Sun, 23 Sep 2007 11:34:09 -0400, Arlen Albert Keshabyan wrote:

 For instance:
 
 int codeval = getValue1() > 5 ??? getValue2() >= 4 ??? getValue3() != 0 ??? 1; 

How would this evaluate in "long hand" code?

 int a = b() > c() ??? d() >= e() ??? f() != g();

-- 
Derek Parnell
Melbourne, Australia
skype: derek.j.parnell

"Stewart Gordon" <smjg_1998 yahoo.com> writes:

"Derek Parnell" <derek psych.ward> wrote in message 
news:1lomy00la8rsl.1fa6pft3qx480$.dlg 40tude.net...
<snip>
 How would this evaluate in "long hand" code?

 int a = b() > c() ??? d() >= e() ??? f() != g();

From what I can make out:

int a = b() > c();
if (!a) a = (d() >= e());
if (!a) a = (f() != g());

This is slightly simpler than usual because it's an initialiser (and because 
int.init happens to be 0).  But in the general case where it's being 
assigned after declaration, you'd need a bit more:

int temp = b() > c();
if (!temp) temp = (d() >= e());
if (!temp) temp = (f() != g());
if (temp) a = temp;

Stewart.

-- 
My e-mail address is valid but not my primary mailbox.  Please keep replies 
on the 'group where everybody may benefit. 

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Stewart Gordon Wrote:

 "Derek Parnell" <derek psych.ward> wrote in message 
 news:1lomy00la8rsl.1fa6pft3qx480$.dlg 40tude.net...
 <snip>
 How would this evaluate in "long hand" code?

 int a = b() > c() ??? d() >= e() ??? f() != g();

 
 From what I can make out:
 
 int a = b() > c();
 if (!a) a = (d() >= e());
 if (!a) a = (f() != g());
 
 This is slightly simpler than usual because it's an initialiser (and because 
 int.init happens to be 0).  But in the general case where it's being 
 assigned after declaration, you'd need a bit more:
 
 int temp = b() > c();
 if (!temp) temp = (d() >= e());
 if (!temp) temp = (f() != g());
 if (temp) a = temp;
 
 Stewart.
 
 -- 
 My e-mail address is valid but not my primary mailbox.  Please keep replies 
 on the 'group where everybody may benefit. 
 

int a = b() > c() ??? d() >= e() ??? f() != g();

can be represented like this:

if(b() > c())
   a = b();
else
   if(d() >= e())
      a = d();
   else
      if(f() != g())
         a = f();

OR compiler must optimize it like this:

{
int temp = b();
if(temp > c())
   a = temp;
else
   {
      temp = d();
      if(temp >= e())
         a = temp;
      else
       {
          temp = f();
          if(temp != g())
            a = temp;
       }
   }
}

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Derek Parnell Wrote:

 On Sun, 23 Sep 2007 11:34:09 -0400, Arlen Albert Keshabyan wrote:
 
 For instance:
 
 int codeval = getValue1() > 5 ??? getValue2() >= 4 ??? getValue3() != 0 ??? 1; 

 
 How would this evaluate in "long hand" code?
 
  int a = b() > c() ??? d() >= e() ??? f() != g();
 
 -- 
 Derek Parnell
 Melbourne, Australia
 skype: derek.j.parnell

if(b() > c())
   a = b();
else
   if(d() >= e())
      a = d();
   else
      if(f() != g())
         a = f();

Robert Fraser <fraserofthenight gmail.com> writes:

Arlen Albert Keshabyan Wrote:

 It would be the sugar syntactic to add '??' operator to D. Consider the
example code:
 
 string error_message = getErrorMessage() ?? "no errors";
 A a = x.getPreparedAObject() ?? y.getPreparedAObject() ?? new A();
 

the first non-null value or to null if all rvalues contains null. This operator
might be extended to '???' to evaluate to a value that conforms to some
conditions. The lvalue gets the first rvalue that evaluates to true (evaluation
goes from left to right). If no conditions evaluates to true then the lvalue
stays unchanged. If no conditions are given explicitly then those conditions
evaluates to true (so, the best place for them is at the end of a sequence).
 
 For instance:
 
 int codeval = getValue1() > 5 ??? getValue2() >= 4 ??? getValue3() != 0 ??? 1;
// if no conditions are satisfied then codeval = 1 as a default value at last.
There are no sane reasons to place it in the middle of the sequence.
 
 int code = 7;
 code = 5 < 10 ??? 0 != 0; // this way, code = 7 (no conditions are satisfied)
 

I have mixed feelings about the ternary operator. Nobody ever told me what it
did, and I never saw it in any code when I was first learning to program, so
consequently, when I finally saw it for the first time, I was baffled by it.
Now that I understand it, I was using it quite often, but recently another
developer (with 10 years more experience than I) was reviewing my code before a
check-in and had never seen the conditional expression used outside a call
expression/parameter list.

Anyways, enough about my life. I think the ?? option should be added... I use
that sort of thing in scripting languages all the time, and now that I know the
dangers of the ?:, I might even use it more frequently than that. I don't think
it should be limited just to object references; as nulls evaluate to false, it
could be extended to other types with a logical false such as integers or
boolean expressions:

int x = 0 ?? 10; // Evaluates to 10
int y = 5 ?? 10; // Evaluates to 5
int z = 0 ?? 0; // Evaluates to 0
bool a = false ?? true; // Evaluates to true
bool b = false ?? false; // Evaluates to false

 ??? seems very weird to me, though, so I'm against that one.

Nathan Reed <nathaniel.reed gmail.com> writes:

Robert Fraser wrote:
 Anyways, enough about my life. I think the ?? option should be added... I use
that sort of thing in scripting languages all the time, and now that I know the
dangers of the ?:, I might even use it more frequently than that. I don't think
it should be limited just to object references; as nulls evaluate to false, it
could be extended to other types with a logical false such as integers or
boolean expressions:
 
 int x = 0 ?? 10; // Evaluates to 10
 int y = 5 ?? 10; // Evaluates to 5
 int z = 0 ?? 0; // Evaluates to 0
 bool a = false ?? true; // Evaluates to true
 bool b = false ?? false; // Evaluates to false
 
  ??? seems very weird to me, though, so I'm against that one.

Is there any reason that || couldn't be overloaded to do this?  It does 
exactly the same thing except that the result is always a bool.

a || b || c || ...

The expressions are evaluated one by one starting from the left, and the 
first one that is true when implicitly converted to a bool is returned. 
  If b is 'true' (meaning, it's a non-null reference, a non-zero number, 
etc), then c is never evaluated.  If all the expressions are 'false' 
(null reference, 0, etc) then the last one is returned.

Similiarly, we could define && such that

a && b && c && ...

results in the expressions being evaluated one by one starting from the 
left, and the first one that is 'false' is returned, or if they are all 
'true', the last one is returned.  If b is 'false', then c is never 
evaluated.

This seems like it could be useful, wouldn't break existing code (as the 
results of these expressions would still be implicitly converted to bool 
when necessary), and don't introduce new operators into the language.

Thanks,
Nathan Reed

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Robert Fraser Wrote:

 Arlen Albert Keshabyan Wrote:
 
 It would be the sugar syntactic to add '??' operator to D. Consider the
example code:
 
 string error_message = getErrorMessage() ?? "no errors";
 A a = x.getPreparedAObject() ?? y.getPreparedAObject() ?? new A();
 

the first non-null value or to null if all rvalues contains null. This operator
might be extended to '???' to evaluate to a value that conforms to some
conditions. The lvalue gets the first rvalue that evaluates to true (evaluation
goes from left to right). If no conditions evaluates to true then the lvalue
stays unchanged. If no conditions are given explicitly then those conditions
evaluates to true (so, the best place for them is at the end of a sequence).
 
 For instance:
 
 int codeval = getValue1() > 5 ??? getValue2() >= 4 ??? getValue3() != 0 ??? 1;
// if no conditions are satisfied then codeval = 1 as a default value at last.
There are no sane reasons to place it in the middle of the sequence.
 
 int code = 7;
 code = 5 < 10 ??? 0 != 0; // this way, code = 7 (no conditions are satisfied)
 

 
 I have mixed feelings about the ternary operator. Nobody ever told me what it
did, and I never saw it in any code when I was first learning to program, so
consequently, when I finally saw it for the first time, I was baffled by it.
Now that I understand it, I was using it quite often, but recently another
developer (with 10 years more experience than I) was reviewing my code before a
check-in and had never seen the conditional expression used outside a call
expression/parameter list.
 
 Anyways, enough about my life. I think the ?? option should be added... I use
that sort of thing in scripting languages all the time, and now that I know the
dangers of the ?:, I might even use it more frequently than that. I don't think
it should be limited just to object references; as nulls evaluate to false, it
could be extended to other types with a logical false such as integers or
boolean expressions:
 
 int x = 0 ?? 10; // Evaluates to 10
 int y = 5 ?? 10; // Evaluates to 5
 int z = 0 ?? 0; // Evaluates to 0
 bool a = false ?? true; // Evaluates to true
 bool b = false ?? false; // Evaluates to false
 
  ??? seems very weird to me, though, so I'm against that one.

?? operator should handle only object references, I think. Nothing more than
that.
??? ternary operator should handle any value types with condition expressions.

The ?? operator should stop evaluating on the first null reference it
encounters in a sequence.
The ??? operator should stop evaluating on the first condition that evaluates
to true EXPLICITLY. Any value in a sequence that has not explicit conditions
should be treated like it evaluates to true despite any implicit conversions.
So, the value of 0 should be evaluated to true in your case even if it might be
implicitly converted to false.

"Rioshin an'Harthen" <rharth75 hotmail.com> writes:

"Arlen Albert Keshabyan" <arlen.albert gmail.com> kirjoitti viestissä 
news:fd7rir$2u8a$1 digitalmars.com...
 ?? operator should handle only object references, I think. Nothing more 
 than that.

I can't begin to count how many times I've written code like

    long id = data.GetLong("ID") ?? 0;

or

    long? id = data.GetLong("ID");

and then later accessing the value with

    id ?? 0



The ?? operator, as well as the nullable ? types, are quite nice to have 
when interfacing with the database. 

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Rioshin anHarthen Wrote:

 "Arlen Albert Keshabyan" <arlen.albert gmail.com> kirjoitti viestissä 
 news:fd7rir$2u8a$1 digitalmars.com...
 ?? operator should handle only object references, I think. Nothing more 
 than that.

 
 I can't begin to count how many times I've written code like
 
     long id = data.GetLong("ID") ?? 0;
 
 or
 
     long? id = data.GetLong("ID");
 
 and then later accessing the value with
 
     id ?? 0
 

 
 The ?? operator, as well as the nullable ? types, are quite nice to have 
 when interfacing with the database. 
 


Agreed partly.

"Janice Caron" <caron800 googlemail.com> writes:

On 9/24/07, Rioshin an'Harthen <rharth75 hotmail.com> wrote:
 I can't begin to count how many times I've written code like

     long id = data.GetLong("ID") ?? 0;


make any sense of it. What is the return type of GetLong()? What is
the type of the variable named data?


     long? id = data.GetLong("ID");

What does the question mark after long mean? What is the type of the
variable named id? Is is "long" or "long?" ?

"Rioshin an'Harthen" <rharth75 hotmail.com> writes:

"Janice Caron" <caron800 googlemail.com> kirjoitti viestissä 
news:mailman.289.1190665392.16939.digitalmars-d puremagic.com...
 On 9/24/07, Rioshin an'Harthen <rharth75 hotmail.com> wrote:
 I can't begin to count how many times I've written code like

     long id = data.GetLong("ID") ?? 0;


 make any sense of it. What is the return type of GetLong()? What is
 the type of the variable named data?

Certainly. :)

GetLong, as the name suggests, is used in this example to return a long 
(signed 64-bit value) from the database column ID, with the current database 
row referenced through the variable data (which, e.g. could be of type 
MySqlDataReader). GetLong may, if the column (in the row accessed) doesn't 
contain a value, return null, to signify this, since 0 (or any other value) 
may be a legal value in the database.

     long? id = data.GetLong("ID");

 What does the question mark after long mean? What is the type of the
 variable named id? Is is "long" or "long?" ?

The question mark after a type name means the type is nullable: basically, 
in addition to its normal range may contain the value null. A nullable type 
can be built using a struct or class, and as far as I'm able to tell from 


If you wish further information on nullable types, I'd recommend reading 


http://www.ecma-international.org/publications/files/ECMA-ST/Ecma-334.pdf 

"Janice Caron" <caron800 googlemail.com> writes:

On 9/25/07, Rioshin an'Harthen <rharth75 hotmail.com> wrote:
 GetLong, as the name suggests, is used in this example to return a long
 (signed 64-bit value) from the database column ID, with the current database
 row referenced through the variable data (which, e.g. could be of type
 MySqlDataReader). GetLong may, if the column (in the row accessed) doesn't
 contain a value, return null, to signify this, since 0 (or any other value)
 may be a legal value in the database.

You contradict yourself. Clearly it does /not/ return a ulong, since a
ulong cannot contain null (or null is indistinguishable from zero,
depending on your point of view).

So what you really mean is that the type ulong? (with a question mark)
is basically equivalent to

struct
{
    ulong n;
    bool isNull;
}

which implicitly casts to its n member, and that ?? tests the isNull field.

I'm sure this could be done with templates. That is:

    Nullable!(long) id = data.GetNullableLong("ID");

    if (isNulled(id)) { /* ... */ }

    long n = id; // implicit cast

Or, all in one step, your

   long id = data.GetLong("ID") ?? 0;

would become

    long id = isNulled(auto temp = data.GetLong("ID")) ? 0 : temp;

"Janice Caron" <caron800 googlemail.com> writes:

Of course, the /easy/ way to do this is to have GetLong return a long
*, and then do

long * temp;
long id = (temp = data.getLong("ID")) ? *temp : 0;

"Janice Caron" <caron800 googlemail.com> writes:

On 9/25/07, Janice Caron <caron800 googlemail.com> wrote:
 Of course, the /easy/ way to do this is to have GetLong return a long
 *, and then do

 long * temp;
 long id = (temp = data.getLong("ID")) ? *temp : 0;

Gosh, that word-wrapped in an inconvenient place! I meant that GetLong
should return a pointer to long.

Max Samukha <samukha voliacable.com.removethis> writes:

On Tue, 25 Sep 2007 11:39:51 +0300, "Rioshin an'Harthen"
<rharth75 hotmail.com> wrote:

"Janice Caron" <caron800 googlemail.com> kirjoitti viestiss? 
news:mailman.289.1190665392.16939.digitalmars-d puremagic.com...
 On 9/24/07, Rioshin an'Harthen <rharth75 hotmail.com> wrote:
 I can't begin to count how many times I've written code like

     long id = data.GetLong("ID") ?? 0;


 make any sense of it. What is the return type of GetLong()? What is
 the type of the variable named data?

Certainly. :)

GetLong, as the name suggests, is used in this example to return a long 
(signed 64-bit value) from the database column ID, with the current database 
row referenced through the variable data (which, e.g. could be of type 
MySqlDataReader). GetLong may, if the column (in the row accessed) doesn't 
contain a value, return null, to signify this, since 0 (or any other value) 
may be a legal value in the database.

     long? id = data.GetLong("ID");

 What does the question mark after long mean? What is the type of the
 variable named id? Is is "long" or "long?" ?

The question mark after a type name means the type is nullable: basically, 
in addition to its normal range may contain the value null. A nullable type 
can be built using a struct or class, and as far as I'm able to tell from 


If you wish further information on nullable types, I'd recommend reading 


http://www.ecma-international.org/publications/files/ECMA-ST/Ecma-334.pdf 



additional bool member. ? is just a syntactic sugar for the templated
System.Nullable struct. Here is a quickly hacked together example of
how nullable types could be implemented in D using templates.

struct Nullable(T)
{
	alias typeof(*this) Type;

	T value;
	bool isNull = true;

	Type opAssign(T v)
	{
		value = v;
		isNull = false;
		return *this;
	}

	static Type opCall(T v)
	{
	    Type ret;
	    ret.value = v;
	    ret.isNull = false;
	    return ret;
	}

	/*
	Hacks to allow null keyword in assignments/comparisons
	if you really want it
	*/

	Type opAssign(void* v)
	{
	    assert(v == null);
	    isNull = true;
	    return *this;
	}

	int opEquals(void* v)
	{
	    assert(v == null);
	    return isNull;
	}

	T getValueOrDefault()
	{
	    return isNull ? T.init : value;
	}

	T ifNull(T v)
	{
	    return isNull ? v : value;
	}

	/*
	etc.
	opImplicitCasts and struct interfaces will be useful here
	*/
}

class DataReader
{
    Nullable!(T) get(T)(char[] field)
    {
        Nullable!(T) ret;
        return ret;
    }
}

void main()
{
    auto data = new DataReader;

    auto id = data.get!(long)("id");
    assert(id == null);

    id = 20;
    assert(id != null);

    id = null;

    // a replacement for 'long x = id ?? -1'
    long x = id.ifNull(-1);
    assert (x == -1);
}

I'm sure DB people already use something like that.

Reiner Pope <some address.com> writes:

Max Samukha wrote:
 On Tue, 25 Sep 2007 11:39:51 +0300, "Rioshin an'Harthen"
 <rharth75 hotmail.com> wrote:
 
 "Janice Caron" <caron800 googlemail.com> kirjoitti viestiss? 
 news:mailman.289.1190665392.16939.digitalmars-d puremagic.com...
 On 9/24/07, Rioshin an'Harthen <rharth75 hotmail.com> wrote:
 I can't begin to count how many times I've written code like

     long id = data.GetLong("ID") ?? 0;


 make any sense of it. What is the return type of GetLong()? What is
 the type of the variable named data?

 Certainly. :)

 GetLong, as the name suggests, is used in this example to return a long 
 (signed 64-bit value) from the database column ID, with the current database 
 row referenced through the variable data (which, e.g. could be of type 
 MySqlDataReader). GetLong may, if the column (in the row accessed) doesn't 
 contain a value, return null, to signify this, since 0 (or any other value) 
 may be a legal value in the database.

     long? id = data.GetLong("ID");

 What does the question mark after long mean? What is the type of the
 variable named id? Is is "long" or "long?" ?

 The question mark after a type name means the type is nullable: basically, 
 in addition to its normal range may contain the value null. A nullable type 
 can be built using a struct or class, and as far as I'm able to tell from 


 If you wish further information on nullable types, I'd recommend reading 


 http://www.ecma-international.org/publications/files/ECMA-ST/Ecma-334.pdf 

 
 

 additional bool member. ? is just a syntactic sugar for the templated
 System.Nullable struct. Here is a quickly hacked together example of
 how nullable types could be implemented in D using templates.
 
 struct Nullable(T)
 {
 	alias typeof(*this) Type;
 
 	T value;
 	bool isNull = true;
 
 	Type opAssign(T v)
 	{
 		value = v;
 		isNull = false;
 		return *this;
 	}
 
 	static Type opCall(T v)
 	{
 	    Type ret;
 	    ret.value = v;
 	    ret.isNull = false;
 	    return ret;
 	}
 
 	/*
 	Hacks to allow null keyword in assignments/comparisons
 	if you really want it
 	*/
 
 	Type opAssign(void* v)
 	{
 	    assert(v == null);
 	    isNull = true;
 	    return *this;
 	}
 
 	int opEquals(void* v)
 	{
 	    assert(v == null);
 	    return isNull;
 	}
 
 	T getValueOrDefault()
 	{
 	    return isNull ? T.init : value;
 	}
 
 	T ifNull(T v)
 	{
 	    return isNull ? v : value;
 	}
 
 	/*
 	etc.
 	opImplicitCasts and struct interfaces will be useful here
 	*/
 }
 
 class DataReader
 {
     Nullable!(T) get(T)(char[] field)
     {
         Nullable!(T) ret;
         return ret;
     }
 }
 
 void main()
 {
     auto data = new DataReader;
 
     auto id = data.get!(long)("id");
     assert(id == null);
 
     id = 20;
     assert(id != null);
 
     id = null;
 
     // a replacement for 'long x = id ?? -1'
     long x = id.ifNull(-1);
     assert (x == -1);
 }
 
 I'm sure DB people already use something like that.

Looks good. And we can overload the bitwise OR operator to make it look 
a little better (in my opinion). We just add

         Nullable!(T) opOr(Nullable!(T) other)
         {
             return isNull ? other : Type(value);
         }

         T opOr(T other)
         {
             return isNull ? other : value;
         }

to your struct, and then as a new main, we get

void main()
{
     auto data = new DataReader;

     auto id = data.get!(long)("id");
     assert(id == null);

     auto id2 = data.get!(long)("id2");
     id2 = 25;

     auto y = id | id2; // y is a Nullable!(T)
     assert(y != null);

     long x = y | -1;
     assert (x == 25);

     id = 20;
     assert(id | id2 | -1 == 20);
}


Voila!

Unfortunately, I couldn't get short-circuiting to work via using lazy on 
parameters types to opOr. This is because | is parsed 
left-associatively, so that doesn't work. I wonder if some 
expression-templatey work could fix this...

     -- Reiner

Max Samukha <samukha voliacable.com.removethis> writes:

On Tue, 25 Sep 2007 22:33:15 +1000, Reiner Pope <some address.com>
wrote:

Looks good. And we can overload the bitwise OR operator to make it look 
a little better (in my opinion). We just add

         Nullable!(T) opOr(Nullable!(T) other)
         {
             return isNull ? other : Type(value);
         }

         T opOr(T other)
         {
             return isNull ? other : value;
         }

to your struct, and then as a new main, we get

void main()
{
     auto data = new DataReader;

     auto id = data.get!(long)("id");
     assert(id == null);

     auto id2 = data.get!(long)("id2");
     id2 = 25;

     auto y = id | id2; // y is a Nullable!(T)
     assert(y != null);

     long x = y | -1;
     assert (x == 25);

     id = 20;
     assert(id | id2 | -1 == 20);
}


Voila!

Unfortunately, I couldn't get short-circuiting to work via using lazy on 
parameters types to opOr. This is because | is parsed 
left-associatively, so that doesn't work. I wonder if some 
expression-templatey work could fix this...

     -- Reiner

I like the way it looks but changing the meaning of overloaded
operators is not recommended by the dogmata. It seems like ?? operator
would be a useful addition to the language.

"Janice Caron" <caron800 googlemail.com> writes:

I think I've solved the problem. This works!


    int main()
    {
        int[string] aa;
        aa["B"] = 42;

        int n = firstOf("A" in aa)("B" in aa)("C" in aa)(0);
        writefln(n);

        return 0;
    }

As you'd expect, it prints 42. But wait - there's more! Just to prove
that it's not doing unnecessary work, watch this:

    int * lookup(string s, int[string] aa)
    {
        writef("(%s) ",s);
        return s in aa;
    }

    int main()
    {
        int[string] aa;
        aa["B"] = 42;

        int m = firstOf(lookup("A",aa))(lookup("B",aa))(lookup("C",aa))(0);
        writefln(m);

        return 0;
    }

This prints
(A) (B) 42.

So the third lookup is not done. Woo hoo! And all with no ?? operator.
How's it done? Well, I'll tell you. It's done like this:


    SnazzyFunctor firstOf(lazy int * p)
    {
        return new SnazzyFunctor(p);
    }

    class SnazzyFunctor
    {
        this(int * p)
        {
            result = p;
        }

        SnazzyFunctor opCall(lazy int * p)
        {
            if (result == null) result = p;
            return firstOf(result);
        }

        int opCall(lazy int n)
        {
            if (result != null) return *result;
            return n;
        }

        int * result;
    }

(I'm sure worthier minds than mine could templatise this)

"Janice Caron" <caron800 googlemail.com> writes:

On 9/26/07, Janice Caron <caron800 googlemail.com> wrote:
     SnazzyFunctor firstOf(lazy int * p)
     {
         return new SnazzyFunctor(p);
     }

Come to think of it, it should work with one fewer lazy:

   SnazzyFunctor firstOf(int * p)
   {
       return new SnazzyFunctor(p);
   }

since you're always going to want the first one to execute, and the
recursive call has already been evaluated. So, even simpler then!

"Janice Caron" <caron800 googlemail.com> writes:

On 9/25/07, Janice Caron <caron800 googlemail.com> wrote:
 Of course, the /easy/ way to do this is to have GetLong return a
 long *, and then do

 long * temp;
 long id = (temp = data.getLong("ID")) ? *temp : 0;

Or chaining lots of tests together

long id = *( get("A") || get("B") || get("C") || get("D") || &default );

"Janice Caron" <caron800 googlemail.com> writes:

 Or chaining lots of tests together

 long id = *( get("A") || get("B") || get("C") || get("D") || &default );

Oh wait! That won't work, because you can't overload ||. Still, it was good try.

I think it would be kinda cool if || could be overloaded to accept
pointer inputs. That is, if the LHS and RHS were both of type T* then
the expression (lhs || rhs) could also have type T*, and would mean
((lhs !is null) ? lhs : rhs).

That would be much nicer than introducing new operators. (It's still a
change to the language though).

"Janice Caron" <caron800 googlemail.com> writes:

How about this..

T maybe(T)(T * ptr, lazy T default)
{
	if (ptr !is null) return *ptr;
	return default;
}

long p = maybe(get("A"), maybe(get("B"), maybe(get("C"),
maybe(get("D"), maybe(get("E"), 0 )))));

Arlen Albert Keshabyan <arlen.albert gmail.com> writes:

Stewart Gordon Wrote:

 "Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message 
 news:fd6ee0$nd4$1 digitalmars.com...
 Stewart Gordon Wrote:
 "Arlen Albert Keshabyan" <arlen.albert gmail.com> wrote in message
 news:fd68lm$f4r$1 digitalmars.com...
 A a = APointer1 ?? APointer2 ?? APointer3;

 the line above is equal to the line below:

 A a = APointer1 != null ??? APointer2 != null ??? APointer3 != null ???
 null;

 Not by the way you specified it.  The first would assign to a an actual
 object reference; the second would try to assign a boolean value.


 <snip>
 No. The code lines do the same thing anyway.

 
 Which same thing - evaluating to a boolean or evaluating to an object 
 reference?  Are you going to supply a corrected version of your proposal?
 
 Stewart.
 
 -- 
 My e-mail address is valid but not my primary mailbox.  Please keep replies 
 on the 'group where everybody may benefit. 
 


It evaluates to an object reference.

Bill Baxter <dnewsgroup billbaxter.com> writes:

Arlen Albert Keshabyan wrote:
 It would be the sugar syntactic to add '??' operator to D. Consider the
example code:
 
 string error_message = getErrorMessage() ?? "no errors";
 A a = x.getPreparedAObject() ?? y.getPreparedAObject() ?? new A();
 


evaluates to the first non-null value or to null if all rvalues contains
null. This operator might be extended to '???' to evaluate to a value
that conforms to some conditions. The lvalue gets the first rvalue that
evaluates to true (evaluation goes from left to right). If no conditions
evaluates to true then the lvalue stays unchanged. If no conditions are
given explicitly then those conditions evaluates to true (so, the best
place for them is at the end of a sequence).
 
 For instance:
 
 int codeval = getValue1() > 5 ??? getValue2() >= 4 ??? getValue3() != 0 ??? 1;
// if no conditions are satisfied then codeval = 1 as a default value at last.
There are no sane reasons to place it in the middle of the sequence.
 
 int code = 7;
 code = 5 < 10 ??? 0 != 0; // this way, code = 7 (no conditions are satisfied)
 

It would be great to have a way to express

   complicated_expression > 0 ? complicated_expression : default

Without having to duplicate complicated_expression, while still 
remaining an expression overall.

But I find the "a op b ??? default" syntax illogical.  It looks like 'a 
op b' should be evaluated, becoming a boolean.

How about Something like:
    complicated_expression ?> 0 : -1

    evaluates to "complicated_expression if it's greater than zero, 
otherwise -1"

[Of course there are a dozen other bugs and features I'd rather see 
handled before this ;-)]

--bb

"Janice Caron" <caron800 googlemail.com> writes:

On 9/24/07, Bill Baxter <dnewsgroup billbaxter.com> wrote:
 It would be great to have a way to express

    complicated_expression > 0 ? complicated_expression : default

 Without having to duplicate complicated_expression, while still
 remaining an expression overall.

(x = complicated_expression) > 0 ? x : default

You'd have to declare x first, but I don't see that as a major
problem, and it does seem more sensible to me than creating new
operators. (If there's one thing worse than keyword bloat, it's
operator bloat).

Bill Baxter <dnewsgroup billbaxter.com> writes:

Janice Caron wrote:
 On 9/24/07, Bill Baxter <dnewsgroup billbaxter.com> wrote:
 It would be great to have a way to express

    complicated_expression > 0 ? complicated_expression : default

 Without having to duplicate complicated_expression, while still
 remaining an expression overall.

 
 (x = complicated_expression) > 0 ? x : default
 
 You'd have to declare x first, but I don't see that as a major
 problem, 

It *is* a problem in a template where you can't use the 
name-matches-return-value trick if there's any extraneous stuff defined 
in the template.  Like so

template ComputeAConstant(Foo)
{
    const ComputeAConstant =
       complicated_expression(Foo) > 0 ?
         complicated_expression(Foo) :
           default;
}

if you stick an x in there you can no longer use it as just 
ComputeAConstant(float).

Of course that should just be fixed anyway, but it's also not so nice to 
have to declare something first in other situations.  Like switch cases 
or nested if-elseifs.

 and it does seem more sensible to me than creating new
 operators. (If there's one thing worse than keyword bloat, it's
 operator bloat).

At least these operators (?> ?< ?==) wouldn't require any new opBlah 
overloads.

--bb

Alexander Panek <a.panek brainsware.org> writes:

Arlen Albert Keshabyan wrote:
 It would be the sugar syntactic to add '??' operator to D. Consider the
example code:
 
 string error_message = getErrorMessage() ?? "no errors";
 A a = x.getPreparedAObject() ?? y.getPreparedAObject() ?? new A();
 

the first non-null value or to null if all rvalues contains null. This operator
might be extended to '???' to evaluate to a value that conforms to some
conditions. The lvalue gets the first rvalue that evaluates to true (evaluation
goes from left to right). If no conditions evaluates to true then the lvalue
stays unchanged. If no conditions are given explicitly then those conditions
evaluates to true (so, the best place for them is at the end of a sequence).
 
 For instance:
 
 int codeval = getValue1() > 5 ??? getValue2() >= 4 ??? getValue3() != 0 ??? 1;
// if no conditions are satisfied then codeval = 1 as a default value at last.
There are no sane reasons to place it in the middle of the sequence.
 
 int code = 7;
 code = 5 < 10 ??? 0 != 0; // this way, code = 7 (no conditions are satisfied)
 

if (auto result_of_complex_expression = 
complex_expression_or_function()) var = result_of_complex_expression;
// else var = null;