http://d.puremagic.com/issues/show_bug.cgi?id=7864

           Summary: Structs with class members can't be sorted
           Product: D
           Version: unspecified
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: regression
          Priority: P2
         Component: DMD
        AssignedTo: nobody@puremagic.com
        ReportedBy: jmdavisProg@gmx.com


--- Comment #0 from Jonathan M Davis <jmdavisProg@gmx.com> 2012-04-08 17:15:12 PDT ---
This compiles with 2.058, but not with the latest head (which I assume is the
same as 2.059 beta)

import std.algorithm;
import std.string;

class C
{
    string str;

    this(string str)
    {
        this.str = str;
    }

    override int opCmp(Object o) const
    {
        auto rhs = cast(C)o;
        assert(rhs);
        return cmp(str, rhs.str);
    }
}

struct S
{
    C c;

    this(C c)
    {
        this.c = c;
    }

    int opCmp(S rhs) const
    {
        if(c < rhs.c)
            return -1;
        if(c > rhs.c)
            return 1;
        return 0;
    }

    int opCmp(const ref S rhs) const
    {
        if(c < rhs.c)
            return -1;
        if(c > rhs.c)
            return 1;
        return 0;
    }
}

void main()
{
    S[] stuff;
    sort(stuff);
}


/home/jmdavis/dmd2/linux/bin/../../src/phobos/std/algorithm.d(6802): Error:
static assert  "Invalid predicate passed to sort: a < b"
q.d(52):        instantiated from here: sort!("a < b",cast(SwapStrategy)0,S[])



The same occurs if you use a SysTime instead of C, so the problem occurs whether the class is directly or indirectly insed of S. This strikes me as being similar to issue# 7808 (though not identical). That was fixed however, and this isn't.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=7864


Kenji Hara <k.hara.pg@gmail.com> changed:

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


--- Comment #1 from Kenji Hara <k.hara.pg@gmail.com> 2012-04-08 19:17:46 PDT ---
(In reply to comment #0)
> This compiles with 2.058, but not with the latest head (which I assume is the
> same as 2.059 beta)
> 
[snip]
> 
> /home/jmdavis/dmd2/linux/bin/../../src/phobos/std/algorithm.d(6802): Error:
> static assert  "Invalid predicate passed to sort: a < b"
> q.d(52):        instantiated from here: sort!("a < b",cast(SwapStrategy)0,S[])

This is wrong opCmp overload problem, not compiler regression. If you change opCmp signatures to:

    int opCmp(const S rhs) const
    int opCmp(const ref S rhs) const
    // may work in general
or:
    int opCmp(const S rhs) const
    // may work in general if S allows lvalue copying
or:
    int opCmp(S rhs) const
    // In this case, sorted elements is always mutable so it works
or:
    int opCmp(const ref S rhs) const
    // In this case, current std.algorithm.sort implementation only requires
    // lvalue comparison, so it works)

Then you can compile it.

> The same occurs if you use a SysTime instead of C, so the problem occurs whether the class is directly or indirectly insed of S. This strikes me as being similar to issue# 7808 (though not identical). That was fixed however, and this isn't.

Bug 7808 was alias this type resolution problem. There is no matter.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=7864


bearophile_hugs@eml.cc changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bearophile_hugs@eml.cc


--- Comment #2 from bearophile_hugs@eml.cc 2012-04-09 10:42:47 PDT ---
(In reply to comment #1)
> If you change opCmp signatures to:
> 
>     int opCmp(const S rhs) const
>     int opCmp(const ref S rhs) const
>     // may work in general

Maybe one opCmp is enough in general:

struct S {
    C c;

    int opCmp()(const auto ref S rhs) const {
        if (c < rhs.c) return -1;
        if (c > rhs.c) return 1;
        return 0;
    }
}

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------