Re: [Libmesh-devel] const Elem * const _parent

[John P. <pet...@cf...>]

Benjamin S. Kirk writes:
 > On Thu, 2006-03-23 at 12:52 -0600, John Peterson wrote:
 > > Yes, we ran into this same problem when reading in adapted grids.
 > > IIRC, we just delayed the creation of the the Elems until we found who
 > > their parents should be, but it would have been nice to be able to set the
 > > parent pointer later.
 > > 
 > > -J
 > 
 > 
 >  Now, correct me if I'm wrong ('cus const &  pointers has always been a
 > bit confusing to me...)

Apparently, it's also confusing to me ;)

 > but Roy's change should not address John's
 > problem, right?  Roy, you have no intention of redefining an element's
 > parent, but only to call non-const methods on it?

 > While a hassle, I prefer John's work-around to keep the pointer itself
 > unchangeable.  The parent-child relationship is pretty fundamental to
 > what we do and I would hate to accidentally overwrite the parent.

And honestly, it's not much of a hassle since when reading in the
elements we do it in level order, so a given element is guaranteed
to be created before any of its children.  Just for completeness, the
rule of thumb is to read

Foo const * const foo

from _right_ to _left_ as: "foo is a constant pointer to a
constant Foo object".  The location of the first "const" before
or after Foo is irrelevant.

-J