Catching results of an exec

[Greg Fortune]

Is there any way to "catch" the value produced by an exec?  I'm considering 
things like print statements that don't really assign a value, but do send 
a result to stdout.

For example, if I execute

exec('print dir()')

from an interpreter, there will be some stuff spit out on the console.  
What if I really wanted to display the list from dir() in a different way?  
I'm working on a really dinky server that will sit on a machine and execute 
arbitrary python code.  I want the results from the code execution to be 
returned to client.

Even just eval('dir()') exhibits the desired behavior, but only from an 
interactive interpreter.

Right now I've got a nasty hack that redirects stdout, uses an 
InteractiveInterpereter from the code module to execute the code, writes to 
an output file, reads the result from the file, restores stdout and then 
sends the result to the client.  Hopefully, that's a terrible way to do it. 
 It just occured to me that I can probably use a pipe in place of the 
output file, but still....

It would be great if exec and eval set something somewhere that held the 
results of the code execution.  I looked in the InteractiveInterpreter 
(which, at least, kicks the results of the code execution out to stdout), 
but all it does is

exec code in self.locals

Did I miss something?

Greg Fortune