For some reason, the author of the MMap module didn't see fit to 
expose the "offset" parameter of the mmap syscall to python. It 
would be really nice if it had that. Currently, it's always set to 0.

    m_obj->data = mmap(NULL, map_size,
               prot, flags,
               fd, 0);

Logged In: YES 
user_id=341410

I agree.  Having access to the offset parameter would be
quite convenient, at least to some who use mmap in a
nontrivial fashion.

Logged In: YES 
user_id=11375

Would either of you care to provide a patch adding the
parameter?  I'll review it...

Logged In: YES 
user_id=341410

I would, but I don't know the slightest about the C-level
mmap internals on any platform, and spending the last hour
looking through Python's mmap.c hasn't made me any more
informed.

James (or anyone else who reads this), are you able?

Logged In: YES 
user_id=671362

There's already a patch for this request:

  http://www.python.org/sf/708374
  add offset to mmap

The rationale is same.
It's almost ready to be committed but has been left out for a 
year now. So give it a second chance.

Logged In: YES 
user_id=33168

The patch just needs some attention (testing and possible
fixes) on Windows.

I need map a file large than 4G in my program, But it is impossible in an 32bit system. So I need use offset parameter to map specified part of the file. I want to know when python will expose the offset parameter to us.

I need map a file large than 4G in my program, But it is impossible in an 32bit system. So I need use offset parameter to map specified part of the file. I want to know when python will expose the offset parameter to us.

This issue can be closed.  The issue is resolved in SVN.