Python 2.4.1, Red Hat Linux 7.3.

Speeds up message parsing on files with large
attachments by approximately 4x, mostly by replacing
REs by direct string processing.

Logged In: YES 
user_id=88157

A first examinaation reveals no particular speedup on an
email with approximately 30 MB of attachments. Can the OP
perhaps provide some code and test data I could time to
verify the assertions of speedup? Otherwise I can't see much
point in applying the patch.

Logged In: YES 
user_id=12800

Here's a slightly better version, cleaned up for style and
applicable to Python 2.5 (which is the only place I'd feel
comfortable applying it).  I've verified that this provides
about a 3x speed up at least for some messages with really
big attachments.

Since this is a performance hack and is considerably invasive of the feedparser code (and needs updating), I'm deferring it to 3.3.

Test fails with stack overflow:

======================================================================
ERROR: test_pushCR_LF (email.test.test_email.TestIterators)
FeedParser BufferedSubFile.push() assumed it received complete
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/serhiy/py/cpython2.7/Lib/email/test/test_email.py", line 2585, in test_pushCR_LF
    bsf.push(il)
  File "/home/serhiy/py/cpython2.7/Lib/email/feedparser.py", line 140, in push
    parts = _splitlines(data)
  File "/home/serhiy/py/cpython2.7/Lib/email/feedparser.py", line 170, in _splitlines
    lines.extend(_splitlines(part))
...
  File "/home/serhiy/py/cpython2.7/Lib/email/feedparser.py", line 170, in _splitlines
    lines.extend(_splitlines(part))
RuntimeError: maximum recursion depth exceeded