Re: Reducing memory by emulating IDL's temporary() function

[Judd Storrs]

On Tue, Apr 29, 2008 at 2:17 AM, John W. Eaton <address@hidden> wrote:

On 25-Apr-2008, Judd Storrs wrote:

| On Fri, Apr 25, 2008 at 1:43 AM, Jaroslav Hajek <address@hidden> wrote:
|
| > 1. there can be an explicit means of "give a value and deallocate".
| > Instead of creating a stand-alone `temporary', however, I suggest
| > extending `clear' to return the cleared value when called with nargout
| > = 1.
| >
|
| One challenge of changing clear like this is that the function form of clear
| requires a string argument.

The "command" form of clear is also passing a character string
argument to the internal "clear" function.  The difference is just in
the way the command/function is parsed.

| The demonstration code I posted also requires a
| string also but I considered that a bug in my implementation.

Why?  Clear works on variable names in the symbol table, not on
values.

The reason I considered it a bug is that I was following the IDL temporary() function definition. I'm still mulling it over but the string idea (i.e. symbol name instead of value) has really grown on me. A pitfall of the IDL model is that you can do things that are completely useless like

A = temporary(B(:))
A = temporary(temporary(B))

This is silly in IDL because B(:) creates an intermediate copy of B and the original B can't be released. The second one is just a waste of time. The explicit symbol name approach avoids that mess and even makes attemting it look like incompetant coding:

A = clear("B(:)")
A = clear("clear(\"B\")");

Should we safeguard against clearing a variable that doesn't have a local representation? i.e. don't allow clear with return value on a varible that is inherited? Think:

function B = test(C)
  B = clear("A")
end

--judd