http://d.puremagic.com/issues/show_bug.cgi?id=251

           Summary: foreach does not allow updating inside with block
           Product: D
           Version: 0.162
          Platform: PC
        OS/Version: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: DMD
        AssignedTo: bugzilla@digitalmars.com
        ReportedBy: ashleymedlock@yahoo.com


Array members cannot be updated inside a with block for either structs or
classes.
Sample code below:

import std.stdio;


struct Bar { float a= 1.0; }

struct Foo
{
  Bar[]   arr;
}


void main( char[][] args )
{
  Foo foo ;//= new Foo();
  foo.arr.length = 20;

  with(foo)
  {
    foreach( int n, Bar bar; arr ) bar.a = 100;
  }
  foreach( int n, Bar bar; foo.arr ) writefln("A=%s", bar.a );
}


-- 

http://d.puremagic.com/issues/show_bug.cgi?id=251





------- Comment #1 from shro8822@uidaho.edu  2006-07-13 18:46 -------
use inout

  with(foo)
  {
    foreach( int n, inout Bar bar; arr ) bar.a = 100;
  }

haven't tested this but...

INVALID I think


-- 

d-bugmail@puremagic.com wrote:

> http://d.puremagic.com/issues/show_bug.cgi?id=251
> 
> 
> 
> 
> 
> ------- Comment #1 from shro8822@uidaho.edu  2006-07-13 18:46 -------
> use inout
> 
>   with(foo)
>   {
>     foreach( int n, inout Bar bar; arr ) bar.a = 100;
>   }
> 
> haven't tested this but...
> 
> INVALID I think
> 
> 
Has this changed?

I know the opApply used to require inout arguments, and they were implied mutable on iteration.

David Medlock wrote:
> d-bugmail@puremagic.com wrote:
> 
>> http://d.puremagic.com/issues/show_bug.cgi?id=251
>>
>>
>>
>>
>>
>> ------- Comment #1 from shro8822@uidaho.edu  2006-07-13 18:46 -------
>> use inout
>>
>>   with(foo)
>>   {
>>     foreach( int n, inout Bar bar; arr ) bar.a = 100;
>>   }
>>
>> haven't tested this but...
>>
>> INVALID I think
>>
>>
> Has this changed?
> 
> I know the opApply used to require inout arguments, and they were implied mutable on iteration.

I think that a basic foreach on an array requiters inout to change things.

http://www.digitalmars.com/d/statement.html#foreach

The opApply does requirer the inout (I ran into this a week or so back). This seems like a problem to me. A non-inout version should be allowed so that read only access can be granted.

BCS wrote:
> David Medlock wrote:
> 
>> d-bugmail@puremagic.com wrote:
>>
>>> http://d.puremagic.com/issues/show_bug.cgi?id=251
>>>
>>>
>>>
>>>
>>>
>>> ------- Comment #1 from shro8822@uidaho.edu  2006-07-13 18:46 -------
>>> use inout
>>>
>>>   with(foo)
>>>   {
>>>     foreach( int n, inout Bar bar; arr ) bar.a = 100;
>>>   }
>>>
>>> haven't tested this but...
>>>
>>> INVALID I think
>>>
>>>
>> Has this changed?
>>
>> I know the opApply used to require inout arguments, and they were implied mutable on iteration.
> 
> 
> I think that a basic foreach on an array requiters inout to change things.
> 
> http://www.digitalmars.com/d/statement.html#foreach
> 
> The opApply does requirer the inout (I ran into this a week or so back). This seems like a problem to me. A non-inout version should be allowed so that read only access can be granted.

Agreed.

BCS wrote:

> I think that a basic foreach on an array requiters inout to change things.
> 
> http://www.digitalmars.com/d/statement.html#foreach
> 
> The opApply does requirer the inout (I ran into this a week or so back). This seems like a problem to me. A non-inout version should be allowed so that read only access can be granted.

I agree too.

But I would also like to have a warning or even completely disallow modifications of variables in foreach that are not declared as inout. The present situation is too error-prone:

-----------------------------------------------------

foreach(i,k; a) // just forgot to type "inout". code does nothing
    k=i+1;

-----------------------------------------------------

should be

-----------------------------------------------------
foreach(i,k; a)
    k=i+1; 	// "warning: k is not inout" or "error: k is immutable".

foreach(i,inout k; a)  // ok
    k=i+1;
-----------------------------------------------------  

or maybe just a simple solution is to make them inout by default?

-- 

Andrei Khropov wrote:
> 
> But I would also like to have a warning or even completely disallow
> modifications of variables in foreach that are not declared as inout.
> The present situation is too error-prone:
> 
[...]
> 
> or maybe just a simple solution is to make them inout by default?
> 

No!!! An error/warning would be enough.

http://d.puremagic.com/issues/show_bug.cgi?id=251


bugzilla@digitalmars.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




------- Comment #3 from bugzilla@digitalmars.com  2008-06-24 21:53 -------
This has nothing to do with it being in a with statement.

At one point, the 'key' part of the foreach was set to 'final', meaning it could not be reassigned. This fell afoul of all the tail const problems, so it was abandoned.

The current compiler is working as designed. The 'key' value is a mutable copy, unless it is declared as 'inout'.


--