std.container: SList linearRemove confusion

Steven Schveighoffer schveiguy at
Thu Oct 7 10:09:51 PDT 2010

On Thu, 07 Oct 2010 12:44:48 -0400, Johannes Pfau <spam at> wrote:

> Hi,
> I'm currently trying to replace the array in my signal implementation
> with the SList from std.container. I need to remove elements from the
> list, but I can't seem to get it right.
> According to the documentation, I expect linearRemove to remove the
> contents of the passed range, but linearRemove always removes all
> elements in the range and _after the range_ .
> For linearRemove(Range r) it seems to be expected? At least the code
> obviously behaves like that:
> -------------------------------------------------------------------------
> auto n = findNode(_root, r._head); //find beginning of range in list
> n._next = null;               //ignore all elements after that element
> return Range(null);
> -------------------------------------------------------------------------
> Is this really the expected behavior for linearRemove?

Yes, an SList range is terminated by a null pointer, so there is no limit  
to how far it will iterate.

Try to construct an SList range (without using Take) that doesn't iterate  
the rest of the list.

> The linearRemove(Take!Range r) version doesn't work either, but this
> time I'm sure it's a bug: it seems to happen if I use a Take!Range with
> the first element being the SList root node. looking at the code:
> -------------------------------------------------------------------------
> Range linearRemove(Take!Range r)
> {
>     auto orig = r.original;
>     // We have something to remove here
>     if (orig._head == _root)
>     {
>         // remove straight from the head of the list
>         for (; !orig.empty; orig.popFront())          <----- _why
> orig.empty and orig.popFront? should be r.empty and r.popFront_
>         {
>             removeFront();
>         }
>         return this[];
>     }
>     [...]
> }

Yes, I think you are right.

> I guess I should file a bug report, but this really is a showstopper for
> me. Is there any way to get this fixed faster? The fix is really trivial.

Yes, fix it in your local copy :)  It is a seldom used construct, so there  
should be no real compiled code in Phobos that will be affected.  Just  
change the source and everything will be good.

But do also file a bug please.


More information about the Digitalmars-d-learn mailing list