How do I get a reference to a KEY value of a dictionary?

[Andy C]

Thanks, I think that is exactly what I'm looking for.  I guess if you do a =
intern(a), then if a is equivalent but not identical to something in the
global string table, it will assign a to reference the object in the global
string table, and thus the old "a" value can be garbage collected.

Well, for a C/C++ programmer, they ARE references, and I can't imagine
thinking otherwise.  They are implemented as references/pointers in the
interpreter.  Thinking in value semantics can simplify things, but there
will always be some cases where it doesn't work.

I think my example is a good one.  If you have a graph in an adjacency list
representation, you don't want the nodes to be copies of each other.  They
should just point to a global list of nodes.  This can definitely matter...
the worst case is when you have a complete graph, where you would have
n(n-1)/2 (or n(n-1) for a complete directed graph) node objects where you
only need n.  Instead of 100,000 nodes (not an unreasonable size), you could
have 10 billion (totally unreasonable).

But it is a good question whether the garbage collection will make the
difference worth it, in most cases.  I could reassign my references, but I
don't know when the old objects get deleted.  That is, I can do a =
intern(a), but the old value of a could hang around in memory for who knows
how long.  If anyone could shed some light on this, I would appreciate it.


"Terry Reedy" <tjreedy at udel.edu> wrote in message
news:Qf-cnYqDQejEJbSiXTWJjw at comcast.com...
>
> "Andy C" <andychup at yahoo.com> wrote in message
> news:645db655.0307311636.71923378 at posting.google.com...
> > I am new to python, so please bear with me if I am making some
> > conceptual error.
> >
> > Basically I want to create a graph with an adjacency list
> > representation, but I don't want any of the adjacency lists to have
> > duplicate strings when it is avoidable.  I have a function
> createEdge
> > that adds an edge to the graph.  The arguments will be distinct
> since
> > they are read from text files.  But basically I want to use the
> > dictionary as a string pool, and if the argument string equals
> > something in the pool already, don't use the argument string, just a
> > use a reference to something in the string pool already.
>
> Thinking in terms of 'references' can both help and hinder.  (There
> have been some long recent discussion.)
>
> Are you familiar with this?
>
> >>> help('intern')
>
> Help on built-in function intern:
>
> intern(...)
>     intern(string) -> string
>
>     ``Intern'' the given string.  This enters the string in the
> (global)
>     table of interned strings whose purpose is to speed up dictionary
> lookups.
>     Return the string itself or the previously interned string object
> with the
>     same value.
>
> TJR
>
>