www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - struct members not default initialized?

reply Ali Cehreli <acehreli yahoo.com> writes:
(I must be missing something here; this is such a fundamental operation and I
can't find a bug report on it.)

I've been under the impression that struct members would always be initialized.
Having seen D an init-happy language compared to C and C++, that's what I would
have expected. :)

Shouldn't the members of s be int.init and double.init below?

struct S
{
    int i;
    double d;
}

void main()
{
    S s;
}

They are not; at least for dmd versions from 2.034 to 2.037.

Also, leaving out some of the initializers do not init the corresponding
members either:

    S s = { 42 };

There, s.d is not double.init.

And static initializers don't help either:

struct S
{
    int i     = 1;
    double d  = 2.3;
}

Again, s.d is not initialized:

    S s = { 42 };

Is the above expected? (Not by me... :) )

Ali
Dec 07 2009
next sibling parent reply bearophile <bearophileHUGS lycos.com> writes:
This prints 42 0.000000 with the latest dmd2 and 42 nan with an older dmd1. It
can be a bug:

import std.c.stdio: printf;

struct S {
    int i;
    double d;
}

void main() {
    S s = { 42 };
    printf("%d %f\n", s.i, s.d);
}

Bye,
bearophile
Dec 07 2009
next sibling parent Ali Cehreli <acehreli yahoo.com> writes:
bearophile Wrote:


 This prints 42 0.000000 with the latest dmd2 and 42 nan with an older dmd1. It
can be a bug:


It prints garbage here with 2.037. :/

As I've responded to my own post :), I think the initialization is meant only
for static objects.


 
 import std.c.stdio: printf;
 
 struct S {
     int i;
     double d;
 }
 
 void main() {
     S s = { 42 };
     printf("%d %f\n", s.i, s.d);
 }
 
 Bye,
 bearophile


Ali
Dec 07 2009
prev sibling parent =?UTF-8?B?QWxpIMOHZWhyZWxp?= <acehreli yahoo.com> writes:
bearophile wrote:

 This prints 42 0.000000 with the latest dmd2 and 42 nan with an older 


dmd1.

You are unlucky then... ;) The following program leaves random values in 
s.d with dmd 2.037. Here is one output:

42 -0.008821


 It can be a bug:


Yes it is! :)

   http://d.puremagic.com/issues/show_bug.cgi?id=2485

Sorry for not finding it earlier; and thanks for you help.


 import std.c.stdio: printf;

 struct S {
     int i;
     double d;
 }

 void main() {
     S s = { 42 };
     printf("%d %f\n", s.i, s.d);
 }

 Bye,
 bearophile


Ali
Dec 28 2009
prev sibling parent Ali Cehreli <acehreli yahoo.com> writes:
Ali Cehreli Wrote:


 (I must be missing something here


Yes I am! :)


 I've been under the impression that struct members would always be
initialized. Having seen D an init-happy language compared to C and C++, that's
what I would have expected. :)


I still think that it should be the default behavior unless no-initialization
was specifically requested, like the case for uninitialized arrays.


 struct S
 {
     int i;
     double d;
 }


[...]

 Also, leaving out some of the initializers do not init the corresponding
members either:
 
     S s = { 42 };
 
 There, s.d is not double.init.


The spec says "Members not specified in the initializer list are default
initialized" for *static* objects. So this works:

    static S s = { 42 };
    dout.writefln(s.d);       // prints nan

This check fails though:

    assert(s.d == double.nan);

and it is probably not meant to work anyway; because I see that the spec at

    http://www.digitalmars.com/d/2.0/float.html

has this to say (probably recently modified):

   x == x → true 	not valid if x is a NaN


 struct S
 {
     int i     = 1;
     double d  = 2.3;
 }
 



     S s = { 42 };


Good: That does set s.d... :)

Ali
Dec 07 2009